*Article* **Influence of Temperature and Lignin Concentration on Formation of Colloidal Lignin Particles in Solvent-Shifting Precipitation**

**Johannes Adamcyk \*, Sebastian Serna-Loaiza, Stefan Beisl, Martin Miltner and Anton Friedl**

Institute of Chemical, Environmental and Bioscience Engineering, TU Wien, 1060 Vienna, Austria; sebastian.serna@tuwien.ac.at (S.S.-L.); stefan.beisl@tuwien.ac.at (S.B.); martin.miltner@tuwien.ac.at (M.M.); anton.friedl@tuwien.ac.at (A.F.)

**\*** Correspondence: johannes.adamcyk@tuwien.ac.at

**Abstract:** Colloidal lignin particles offer a promising route towards material applications of lignin. While many parameters influencing the formation of these particles in solvent-shifting precipitation have been studied, only a small amount of research on the influence of temperature has been conducted so far, despite it being a major influence parameter in the precipitation of colloidal lignin particles. Temperature influences various other relevant properties, such as viscosity, density, and lignin solubility. This makes investigation of both temperature and lignin concentration in combination interesting. The present work investigates the precipitation at different temperatures and initial lignin concentrations, revealing that an increased mixing temperature results in smaller particle sizes, while the yield is slightly lowered. This effect was strongest at the highest lignin concentration, lowering the hydrodynamic diameter of the particles from 205 to 168 nm. Decreasing the lignin concentration resulted in significantly smaller particles (from 205 to 121 nm at 20 ◦C mixing temperature) but almost no change in particle yield (between 81.2 and 84.6% at 20 ◦C mixing temperature). This opens up possibilities for the process control and optimization of lignin precipitation.

**Keywords:** lignin; colloidal particles; biorefinery; organosolv; precipitation; self-assembly; solvent shifting

#### **1. Introduction**

Lignin is an abundantly available biopolymer, which is currently mostly used for energy production and thus underutilized as a material. Several authors have stressed the importance of lignin valorization into high added-value products for the sustainability of biorefineries [1–3]. Therefore, evaluating the material uses of lignin becomes more relevant, especially considering the multiple properties and functionalities that lignin has (biodegradability, biocompatibility, UV-resistance, and low toxicity) [4]. In recent years, investigations of colloidal lignin particles (CLPs) for material applications have shown some promising results [4–8]. Due to the high specific surface area, lignin nanomaterials have shown improved qualities compared to bulk materials [9], which makes them interesting for applications such as sunscreens [8], food packaging [5], and emulsifiers [7], among others [4].

Organosolv pretreatment is a well-known method for the extraction of lignin suitable for high-value applications, using organic solvents such as aqueous ethanol, organic acids, or acetone [10,11]. There are several methods to produce CLPs from the extracted lignin (ideally directly from the liquor), such as pH shifting, solvent shifting, or polymerization [12]. Among these methods, solvent shifting (lignin precipitation by mixing lignin solution with an antisolvent) has been intensively investigated and has shown promising results [12] but has the downside of high solvent consumption and often low precipitation yields [13]. This leads to a low overall process efficiency, since solvents need to be recovered

**Citation:** Adamcyk, J.; Serna-Loaiza, S.; Beisl, S.; Miltner, M.; Friedl, A. Influence of Temperature and Lignin Concentration on Formation of Colloidal Lignin Particles in Solvent-Shifting Precipitation. *Sustainability* **2022**, *14*, 1219. https://doi.org/10.3390/su14031219

Academic Editors: Oz Sahin and Edoardo Bertone

Received: 6 December 2021 Accepted: 18 January 2022 Published: 21 January 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

downstream, and substantial amounts of lignin stay in solution after precipitation. Thus, methods to improve the efficiency of CLP production need to be investigated.

In previous works [13], the influence of the antisolvent, ratio, and flowrate in the mixer were investigated, which indicated that the flow regime in the mixer has a strong impact on particle formation. The flow regime in a tube can be characterized by the Reynolds number, which is influenced by fluid viscosity and density, both temperature dependent properties. Additionally, the degree of local supersaturation is known to influence particle formation [14], which is influenced by the temperature-dependent solubility of the precipitated compound. In summary, the temperature influences several properties that are critical for the formation of colloidal lignin particles. This makes predicting the influence of temperature on particle formation difficult and requires experimental research, which has not been investigated much, despite being a major process parameter [15].

Varying the precipitation temperature also opens up interesting possibilities from a process perspective. On a laboratory scale, lignin precipitations are usually conducted at ambient temperature [13,16–18]. This is not necessarily representative of precipitation in a biorefinery process, since temperatures in the range of 150–210 ◦C are commonly applied in organosolv pretreatment [19,20]. Hence, cooling the liquor to ambient temperature after pretreatment implicates an energy demand that should be fulfilled only if necessary, which is currently a matter of uncertainty. If higher liquor temperatures still result in colloidal particles, this could be a step towards the optimization of a biorefinery process producing CLPs. Additionally, since lignin solubility increases with the temperature [21], the process efficiency could be further improved if the lignin concentration can be increased at higher temperatures. This would reduce the amount of solvent needed per the number of CLPs produced.

To summarize, the influence of the temperature in the solvent-shifting precipitation of lignin has not been sufficiently investigated so far, despite being a major process parameter with high relevance in scaled-up processes. Since the temperature influences lignin solubility, which in turn impacts process efficiency, it makes sense to investigate temperature and lignin concentration simultaneously. In this work, we therefore varied the temperature of the lignin solution and the mixer, as well as the initial lignin concentration in the precipitation of a commercial organosolv-lignin solution using a T-mixer. The resulting suspensions were characterized by particle diameter and yield to gain knowledge on the impacts on product quality and process efficiency. Selected samples were analyzed regarding their molecular weight distribution.

#### **2. Materials and Methods**

#### *2.1. Materials*

The lignin used in this work was a commercial organosolv-lignin from an annual plant supplied by ChemicalPoint (Oberhaching, Germany), which had a lignin content of 93.7 wt%, ash content of 1.6 wt%, and carbohydrate content of 0.2 wt%. For the solutions, 99.9 wt% undenatured ethanol (Chem-Lab, Zedelgem, Belgium), and ultrapure water (produced with a Sartorius arium pro system at 18 MΩ/cm2) were used.

#### *2.2. Preparation of Lignin Solutions*

Commercial organosolv-lignin and 60 wt% aqueous ethanol where mixed in a ratio of 15 g dry lignin per 1 L of aqueous ethanol and stirred at 20 ◦C for 24 h. After that, the liquids and solids were separated by filtration using cellulose nitrate filters (Whatman, Maidstone, Great Britain) with a pore size of 0.1 μm. The particle-free filtrate was used as lignin stock solution in the precipitation experiments. The concentration of this stock solution was 12 g/kg, which was determined by drying a sample of the stock solution in a drying oven at 105 ◦C until it reached a constant weight.

#### *2.3. Precipitation*

For the precipitation experiments, the lignin concentration of the solution was set to three different concentrations by volumetric addition of 60 wt% aqueous ethanol. The concentration levels used were that of the lignin stock solution (12 g/kg), a dilution to 75% (9 g/kg), and a dilution to 50% (6 g/kg) of that concentration. The lignin was precipitated in a T-mixer with ultrapure water as the antisolvent, as described by Beisl et al. [22] and depicted in Figure 1. The volumetric flow in the mixer was set to 112.5 mL/min and the volumetric ratio of extract to antisolvent was kept at 1:5 for all experiments. The temperature of the lignin solution, the antisolvent, and the mixer was controlled with water baths. Antisolvent and lignin solution were overheated to compensate for heat losses during syringe filling and pumping. The temperature was checked at the mixer inlet before the precipitation, in the bath, and in the tempered beaker where the suspension was collected. The temperature was varied over three different values, 20, 40, and 60 ◦C. Three precipitations were carried out for each experimental condition; the presented results are averages and standard deviations of the three repetitions. Two sets of precipitation experiments were conducted. In the first set, the temperature of lignin solution, antisolvent, and mixer was always set to the same value to facilitate a better understanding of the temperature's influence. In the second set, only the lignin solution's temperature was varied, while the other temperatures were always set to 20 ◦C, to simulate conditions closer to an industrial process.

**Figure 1.** Schematic of the precipitation setup.

#### *2.4. Analytics*

#### 2.4.1. Particle Size

The hydrodynamic diameter of the lignin particles was determined with an Anton Paar Litesizer 500 (Graz, Austria). The suspensions were diluted 1:150 with ultrapure water before the measurements. The refractive index of the particles was set to 1.53 and the absorbance to 0.1. For each precipitation, the average of three measurements was calculated.

#### 2.4.2. Yield

The particle yield of the precipitations was determined by filtering (filtrate) the suspensions through a hydrophilic polyethersulfone membrane with a 30 kDa cutoff (supplied by Nadir®) and comparing the dry matter content of the filtrate to that of the suspensions, as shown in Equation (1). The dry matter content was determined by drying the samples in a drying oven at 105 ◦C until they reached a constant weight.

$$\text{Particle Yield} = \text{(DMS} - \text{DMF}) / \text{DMS} \times 100\%\_{\text{t}} \tag{1}$$

where DMS is the dry matter of the suspension and DMF is the dry matter of the particlefree filtrate. Preliminary tests showed that this filtration method leads to similar but more consistent results compared to the ultracentrifugation used in previous works [13], with lower operational and material expenditure.

#### 2.4.3. Molecular Weight Distribution

The molecular weight distribution was determined through high-performance size exclusion chromatography (HP–SEC), with 10 mM NaOH as eluent with three TSK-Gel columns in series at 40 ◦C (PW5000, PW4000, PW3000; TOSOH Bioscience, Darmstadt, Germany) using an Agilent 1200 HPLC system (flow rate: 1 mL/min, DAD detection at 280 nm). The pH of liquid samples was adjusted to that of the eluent with NaOH for analysis. Polystyrene sulfonate reference standards (PSS GmbH, Mainz, Germany) with molar mass peak Maxima at 78,400 Da, 33,500 Da, 15,800 Da, 6430 Da, 1670 Da, 891 Da, and 208 Da were used for calibration.

#### **3. Results and Discussion**

#### *3.1. First Experimental Plan: Variation of Mixing Temperature*

Both the lignin concentration and the temperature were varied to three different levels (6 to 12 g/kg and 20 to 60 ◦C, respectively). For the first set of experiments, the temperature for the lignin solution and the antisolvent was set to the same temperature for each precipitation. As can be seen in Figure 2, increasing the mixing temperature in the lignin precipitation led to smaller particle sizes, while an increased lignin concentration increased the particle size. This means that the increase in particle size at higher lignin concentrations can be at least partially compensated for by increasing the mixing temperature. In the temperature and concentration ranges used in this work, the influence of the initial lignin concentration was stronger than that of the mixing temperature. Increasing the lignin concentration is advantageous for process efficiency, since less solvent and antisolvent are needed per the number of CLPs produced. However, this also leads to larger particles, which can be partially compensated for by increasing the mixing temperature, as the results indicate. Significantly higher lignin concentrations than those used in this study have been reported in the literature (Goldmann et al. [23] achieved a kraft lignin concentration of 235.89 g/L in 60 wt% aqueous ethanol). However, the temperature cannot be increased much higher than the maximum temperature applied in this work without potentially significantly altering the lignin molecular structure. Thus, the results indicate that the possibility to compensate for higher concentrations by higher mixing temperatures is limited. Additionally, the influence of the temperature is stronger for higher lignin concentrations, while the particle diameter stays similar for all temperatures at the lowest concentration. This indicates that the particle size converges towards a lower limit with increasing precipitation temperatures.

The increase in particle size with the increasing concentration of lignin is a welldocumented phenomenon [24,25] and can be explained by a larger number of solubilized molecules available for particle growth after primary nucleation [26]; however, contrasting results have also been reported [27]. So far, little research has been conducted on the influence of temperature on lignin precipitation [15]. According to Elnashaie et al. [14], an elevated temperature tends to lead to larger particles due to it hindering primary nucleation.

However, in the case examined in this work, other effects seem to have had a stronger influence, since the particle size decreased with the mixing temperature.

**Figure 2.** Influence of mixing temperature (**a**) and initial lignin concentration (**b**) on particle diameter.

In this work, the particle size was only determined by dynamic light scattering (DLS), which cannot differentiate between single particles and aggregates. However, pictures obtained by scatter electron microscopy (SEM) published in previous works [13,22] showed that the particle diameters as determined by DLS are in the same order of magnitude as those determined by SEM.

A previous work [13] showed that increasing the volumetric flow in the T-mixer up to a certain point leads to decreasing particle sizes. This was explained by improved mixing at higher flowrates, which can be expressed as higher Reynolds (Re) numbers in the mixer. Therefore, a possible explanation for the decreasing particle sizes with increasing mixing temperature could be the increase in the Reynolds number in the mixer from 396 to 1017 due to the decrease in the viscosity with the temperature (from 1.6 to 0.6\*10<sup>3</sup> Pas). The correlation of particle size with Re in the mixer and the viscosity of the mixture are shown in Figure 3. The density and dynamic viscosity of the suspension were approximated using literature data from Belda et al. [28]. Due to the proportional increase in the Re with the temperature in the mixer, the correlation of particle size and Re is similar to that of particle size and temperature (Figure 3a). On the other hand, Figure 3b shows that the particle size linearly increased with the viscosity. These correlations suggest that the particle size is primarily determined by the mixing quality, which improves with an increased mixing temperature due to the decrease in viscosity. These results are in agreement with results from other works correlating smaller particle sizes with improved mixing quality [29,30].

Another relevant factor for the particle formation is the lignin supersaturation in the mixture, the driving force for precipitation. The solubility of lignin increases with increasing temperature [21], which should lower the driving force for precipitation and thus lead to higher particle diameters due to the lower precipitation speed [14]. While the lower supersaturation might still be a factor, the results indicate that the improved mixing quality outweighs its influence and leads to smaller particle diameters.

Since the experiments were carried out with ethanol–water mixtures, it is not certain whether the results can be transferred to other solvents commonly used for solvent-shifting precipitation, such as acetone [25] or tetrahydrofuran (THF) [31]. However, the increase in particle size with increasing concentration is well-established for precipitation [24,25]. As stated earlier, information from the literature on the influence of temperature on lignin precipitation is limited. Based on the explanation that the particle size decreases due to improved mixing with decreasing viscosity, it should be possible to also transfer the results to other solvents, since the same principles should apply. This could be investigated in future works to confirm or refute the explanation given in the present work.

**Figure 3.** Influence of Re (**a**) and dynamic viscosity (**b**) in the mixer on the hydrodynamic diameter of CLPs.

The particle yield of the precipitations ranged from 76.7 to 85.8% (Figure 4), which can be considered high compared to the results of previous works [13]. When comparing the results from different conditions, there was a slight trend to lower yields at higher temperatures, and higher yields at higher initial lignin concentrations. The latter result matches findings by Xiong et al. [24], who precipitated enzymatic hydrolysis lignin from THF at initial lignin concentrations ranging from 0.5 to 2 mg/mL; however, the changes in the yield were small in the present work, especially considering the deviations of the experiments at the same conditions. Since the yield was determined gravimetrically, the influence of random errors was higher at lower dry matter concentrations, which explains the tendentially higher deviations for yield and filtrate dry matter at lower lignin concentrations (e.g., Figure 4).

**Figure 4.** Influence of mixing temperature (**a**) and lignin concentration (**b**) on particle yield.

The decreased yield at higher temperatures could be explained by the increasing solubility of lignin, leading more lignin to stay in solution. The yield was determined at ambient temperature, which would suggest that nonprecipitated lignin stays in solution even after cooling down.

Since the precipitation is assumed to be solubility-driven, it is noteworthy that there was only a slight decrease in the yields with decreasing initial lignin concentrations (Figure 4b). If the solubility limit of lignin at a certain ethanol content is assumed to be constant, the concentration of solubilized lignin after the precipitation should be constant. This would result in a direct correlation between initial lignin concentration and

particle yield. For the particle yield determination, the lignin particles were removed using membrane filtration (30 kDa cutoff), and the dry matter contents of the particle-free filtrates were calculated. These dry matter contents are plotted in Figure 5, showing that the dry matter content of the filtrates from the two lower lignin concentrations was significantly lower than that of the highest concentration. This indicates that the concentration of solubilized lignin after precipitation depends on the initial lignin concentration, meaning that the amount of solubilized lignin is higher when more lignin is present.

**Figure 5.** Dry matter content of filtrate for different mixing temperatures (**a**) and lignin concentrations (**b**).

This could be explained by the polydispersity of lignin and the different solubility limits of lignins with different molecular structures and weights, which was also found by Buranov et al. [32]. Previous studies have also shown the fractionation of lignin by molecular weight in solvent-shifting precipitation, meaning that predominately highmolecular-weight lignin precipitates while low-molecular-weight lignin predominately stays in solution [33,34]. Figure 6 shows the molecular weight distributions of lignin in filtrates from experiments at different initial lignin concentrations. The non-normalized distributions (Figure 6a) show a significant increase in lignin concentration over the whole molecular weight spectrum from the lowest initial lignin concentration. Interestingly, the distribution of the filtrates from two higher initial concentrations are very similar, despite a difference in dry matter content; the main difference is an increase in the peak at 400 Da, which was found in earlier works to be influenced by *p*-hydroxycinnamic acids such as ferulic acid and *p*-coumaric acid, both monomeric and connected to lignin fragments [35]. This suggests that the difference in residual solubilized lignin between these samples is mostly influenced by this lignin fraction.

The area-normalized distributions (Figure 6b) show that the filtrate from the lowest initial concentration has the highest ratio of lignin at 400 Da, while the ratio of the highest molecular weight fraction is very low. The distributions of the two higher initial concentrations are very similar, the main difference being a higher ratio of the 400 Da fraction for the highest initial lignin concentration.

The results suggest that the higher-molecular-weight fractions soluble at 10 wt% ethanol (ethanol content after precipitation) are completely dissolved at 6 g/kg initial lignin concentration, but hit a solubility limit at 9 g/kg. The lignin fraction at 400 Da exhibits better solubility. The increase of this fraction from 9 to 12 g/kg suggests that the concentration of this fraction is limited by the concentration in the solution at 9 g/kg, not by solubility. Generally, the results demonstrate the high complexity of lignin solubility. Therefore, the solubility of different lignin fractions will be further investigated in future works.

From a process perspective, the observation that the decrease in yield was less than proportional with the decrease in lignin concentration has interesting implications, since it would allow the tailoring of the particle size by adjusting the lignin concentration of the solution with only a minor impact on the yield. For example, when the lignin concentration of the solution is reduced to 50% at 20 ◦C mixing temperature, the particle size is reduced from 205 ± 4 to 121 ± 2 nm (reduction by 41%), while the yield does not change significantly. However, the merits of this depend on the requirements and value of the final product, since a higher solvent consumption lowers process efficiency.

In the context of a biorefinery process producing CLPs by precipitation directly from the organosolv extract [13], the results of the first set of experiments lead to the conclusion that cooling the extract to ambient temperature after lignin extraction is not necessary. Higher mixing temperatures still result in colloidal particles and even lower particle diameters compared to ambient mixing temperature, while the yield does not change significantly.

#### *3.2. Second Experimental Plan—Variation of Lignin Solution Temperature*

While it may be advantageous to use warm extract from a process perspective, it is disadvantageous to use warm antisolvent, since this would result in increased energy demand for the precipitation process. While warming antisolvent and lignin solution to the same temperatures for the precipitations was necessary to facilitate a better understanding of the influence of the temperature, in an industrial process, different temperatures for the solution and antisolvent are more likely. Thus, a second experimental plan was conducted, in which the temperatures of the antisolvent and the mixer were kept at 20 ◦C, while the temperature of the lignin solution was varied.

Figure 7 shows that there are only small differences in the particle diameter for different solution temperatures. The ratio of extract to antisolvent was kept at 1:5 for all experiments, so the mixing temperature was influenced more by the temperature of the antisolvent and the mixer than that of the lignin solution. This meant there were only small temperature changes in the temperature of the suspension after mixing in this experimental plan. It is noteworthy that the experiments with the solution warmed to 60 ◦C consistently resulted in the smallest average particle sizes. As in the first set of experiments, the influence of temperature was most visible at the highest lignin concentration. This supports the findings from the first experimental plan, since a higher solution temperature results in higher temperatures in the mixer. These results suggest that the temperature and properties of the mixture are relevant for the precipitation, rather than the temperature and properties of the lignin solution or antisolvent. This could mean that the size of the final particles is decided after the solution and antisolvent are mixed.

**Figure 7.** Influence of solution temperature (**a**) and lignin concentration (**b**) on particle diameter with constant antisolvent temperature.

Similar to the particle sizes, the particle yield of the second experimental plan showed no strong dependency on the lignin solution temperature (Figure 8a). As in the first experimental plan, the correlation between the initial lignin concentration and the particle yield was less than proportional (Figure 8b). The results from the second experimental plan indicate that elevated temperatures of the lignin solution alone have a negligible influence on both the particle size and yield. In a biorefinery process, this would mean that it is not necessary to cool the liquor to ambient temperature after biomass pretreatment.

**Figure 8.** Influence of lignin solution temperature (**a**) and initial lignin concentration (**b**) on particle yield.

#### **4. Conclusions**

In the present work, the influence of the temperature and lignin concentration on the precipitation of CLPs was investigated. The particle size increased with an increasing concentration of lignin in the solution and decreased with increasing mixing temperatures. The particle yield was slightly lowered by increasing the mixing temperature, while lower initial lignin concentrations resulted in lower yields. The correlation between the particle yield and lignin concentration was underproportional, which was explained by the polydispersity of lignin and the different solubilities of different lignin fractions. In a second set of experiments, lignin solutions warmed to different temperatures were precipitated with antisolvent at ambient temperature, which had only a minor influence on the particle size and no significant influence on the yield. The lignin concentration showed the same influence on the particle size as in the first set of experiments.

From a process perspective, the results suggest that the precipitation of lignin particles directly from warm extract is not disadvantageous with respect to particle size and yield, but may in fact be advantageous for the formation of CLPs with smaller diameters. This finding could help to increase the efficiency of a biorefinery producing CLPs by removing the necessity to cool the liquor to ambient temperature after lignin extraction.

The changes in the molecular weights of lignin still dissolved after precipitation demonstrate the complexity of lignin solubility even in a relatively simple experimental plan and stress the necessity of further investigating this topic. The solubility and interaction of different lignin fractions will be investigated in future works, as well as downstreaming and valorization methods for the nonprecipitated lignin.

**Author Contributions:** Conceptualization, J.A.; methodology, J.A.; formal analysis, J.A., S.B. and M.M.; investigation, J.A.; writing—original draft preparation, J.A.; writing—review and editing, S.B., S.S.-L., M.M. and A.F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was partially funded by FFG Spin-off Fellowship project (FFG project no. 874260).

**Data Availability Statement:** Data supporting the results are included within the paper.

**Acknowledgments:** Open Access Funding by TU Wien. This paper builds on a paper on the same topic that was presented by the first author at the 2021 Sustainable Development of Energy, Water, and Environment Systems (SDEWES) Conference (10–15 October, Dubrovnik).

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

#### **References**


### *Article* **Development of a Bio-Digital Interface Powered by Microbial Fuel Cells**

**Jiseon You 1,\*, Arjuna Mendis 1, John Greenman 1, Julie Freeman 2, Stephen Wolff 2, Rachel Armstrong 3, Rolf Hughes <sup>3</sup> and Ioannis A. Ieropoulos 1,4,\***


**Abstract:** This paper reports the first relatable bio-digital interface powered by microbial fuel cells (MFCs) that was developed to inform the public and introduce the concept of using live microbes as waste processors within our homes and cities. An innovative design for the MFC and peripherals system was built as a digital data generator and bioreactor, with a custom-built energy-harvesting controller that was connected to the system to enable efficient system operation using adaptive dynamic cell reconfiguration and transmit data for the bio-digital interface. This system has accomplished multiple (parallel) tasks such as electricity generation, wastewater treatment and autonomous operation. Moreover, the controller demonstrated that microbial behaviour and consequent system operation can benefit from smart algorithms. In addition to these technical achievements, the bio-digital interface is a site for the production of digital art that aims to gain acceptance from a wider-interest community and potential audiences by showcasing the capabilities of living microorganisms in the context of green technologies.

**Keywords:** sustainable built environment; microbial fuel cell; bio-digital interface; adaptive dynamic cell reconfiguration

#### **1. Introduction**

Although the first discovery of microbial electricity generation was made over 100 years ago [1], attention towards microbial fuel cell (MFC) technology has only begun to grow fairly recently (1990s), and it is still unfamiliar to the public. Whether or not future developers have encountered a technology, either directly or indirectly, this is crucial in its development. In contrast to existing large-scale sewage treatment facilities, MFCs enable medium/small-scale, decentralised, remote, on-site sewage treatment, as well as energy generation [2,3]. While there are still several technical challenges to overcome to achieve full commercialisation, such as the relatively low power density and high initial investment costs [4], additional challenges that cannot be overlooked include the perception, public acceptability and usability of the technology, which do not have technological solutions and, therefore, require different approaches.

Microbial fuel cell technology is being developed for real-world implementation as a commercially viable product, with demonstrations in various settings such as public science events, music festivals [5] and field trials [6], as an essential tool for technology evaluation. One of the previous study investigated user perception of the technology where everyday use enabled end users to understand its benefits [7]. Such studies have

**Citation:** You, J.; Mendis, A.; Greenman, J.; Freeman, J.; Wolff, S.; Armstrong, R.; Hughes, R.; Ieropoulos, I.A. Development of a Bio-Digital Interface Powered by Microbial Fuel Cells. *Sustainability* **2022**, *14*, 1735. https://doi.org/ 10.3390/su14031735

Academic Editors: Oz Sahin and Edoardo Bertone

Received: 13 December 2021 Accepted: 8 January 2022 Published: 2 February 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

revealed that MFC technology could be highly acceptable and help tackle sanitation issues by providing not only electricity but also accompanying benefits such as feeling safe and improving toilet cleanliness. Despite very positive feedback received so far, the broader uptake of MFC technology into our daily lives requires further work. For example, although there is a growing social consensus on the need for wastewater reuse, the biggest challenge lies in changing the stakeholders' perception of the water cycle more than technological developments [8,9]. Therefore, in order to develop and disseminate a particular technology, efforts must be made in many respects, such as understanding the need, technical context of the solution and cultural acceptance/resistance to this.

The goal of this project is to develop a bio-digital interface that encompasses technological advancements in an art installation, to better articulate how the technology works and encourage public engagement through curiosity that assists its progress towards the market. By developing an attractive interface that enables people and microbes to interact, enhances the user experience and increases the probability of technology uptake. This approach may further result in the same system being utilised as an educational tool or become part of the emerging market of ecohomes [10]. Following the successful completion of the Living Architecture project [11], which demonstrated a selectively programmable bioreactor wall for future living spaces, development of the first relatable bio-digital interface powered by MFCs was set to inspire the public and to introduce the concept of using live microbes as processors of waste within our homes and cities. This 'sociable' interface is a first-generation bio-digital hardware and user experience that translates microbial activity into meaningful encounters with human audiences, establishing a trans-species communications platform. Beginning with an electronic interface powered by MFCs, it extracts data from sensors, which is translated into a lively, communicable display that is showcased in a range of social contexts—from art galleries to exhibition spaces and festivals. From a scientific perspective, the following objectives were pursued: (i) develop an innovative MFC design capable of generating sufficient levels of power; (ii) develop an energy-efficient, multi-functional electronic controller; (iii) better understand microbial behaviour in functioning MFCs when subjected to external stimuli from the electronic controller.

#### **2. Materials and Methods**

#### *2.1. System Design, Operation and Performance Measurement*

For this work, a total of 15 MFC units were built and placed in five cascade groups of three MFCs. The bioreactor chassis was 3D printed in ABS (acrylonitrile butadiene styrene) with dimensions of 193 × 100 × 80 mm (W × H × D). Each MFC bioreactor consists of two anode chambers and two cathode chambers alternately located. This is to increase the contact area between the anode and cathode chambers within the volume of a given individual MFC unit; the two anodes and two cathodes of each MFC were connected in parallel. Each chamber holds approximately 100 mL and 120 mL volumes of anolyte and catholyte, respectively. The anode electrodes were made from plain carbon veil (20 g/m<sup>2</sup> carbon loading, PRF Composite Materials, Poole, Dorset, UK) modified with activated carbon ink [12]. Each anode electrode had a macro surface area of 1,350 cm2 (30 × 45 cm). A U-shaped cathode made of thermally compressed activated carbon on stainless steel mesh [13] was placed inside each cathode chamber in contact with ceramic membranes. The size of the cathode (macro surface area) was 68 cm2 (17 × 4 cm). Custom-made ceramic membrane sheets (65 × 85 mm each, thickness of 3 mm) [14] were placed between the anode and cathode chambers. The MFCs were designed to be gravity fed (the effluent of the MFC above becomes the influent of the MFC below) in order to reduce the energy input requirements. The flow of both anolyte and catholyte of the MFC system is described in Figure 1.

**Figure 1.** Flow diagram of the system. The black- and orange-coloured squares represent the anode and cathode chambers, respectively.

MFC anodes were inoculated with activated sludge collected from a local sewage treatment plant (Wessex Water, Saltford, UK), after being cultivated for 24 h in artificial urine medium (AUM). AUM was also used as feedstock and supplied to the system at various flow rates ranging between 1.5 L/d and 16.0 L/d. The composition of feedstock was modified from a previous study [15], containing the following: 2.5 g/L peptone, 0.5 g/L yeast extract, 5 g/L urea, 5.2 g/L sodium chloride, 1.4 g/L sodium sulphate, 0.95 g/L potassium dihydrogen phosphate, 1.2 g/L di-potassium hydrogen phosphate. A day after inoculation, a fixed external resistance of 200 Ω was connected to each MFC. The value of external resistance was gradually decreased to 100 Ω before being connected to a bespoke electronic controller on day 14.

The voltage output levels of 15 MFCs in volts (V) were initially monitored against time using a data logger (34972A DAQ unit, Agilent Technologies, Santa Clara, CA, USA) every 5 min. Once the aforementioned electronic controller was connected to the system, MFC power performance, including voltage, current, power and information of electrical configurations, was monitored by the controller in different time intervals. Power density (PD) of an MFC unit or of the entire MFC system was normalised by the anolyte volume of 200 mL (per unit) or 3 L (for the whole system). In order to verify an individual MFC performance, periodic polarisation experiments were performed. For this, MFCs were left open circuit for at least 3 h before the run to reach stable open circuit voltages (OCVs). Then, various external resistances ranging from 1.2 kΩ to 4 Ω were loaded every 5 min and the potential between the anode and cathode was recorded every 30 s. For measuring chemical oxygen demand (COD), water samples were filtered using 0.45 μm syringe filters and then immediately analysed using a COD test kit (COD test tubes, Camlab, Cambridge, UK).

#### *2.2. Electrical Infrastructure*

A bespoke system, consisting of the electronic control system (ECS), power connections, external sensors, external actuators such as pumps and LEDs and a data-communication link, was developed to interface the MFC system located in the laboratory with the users' virtual experience. This infrastructure was autonomously harvesting electrical energy generated from the MFC array to measure, filter and collect the parameters of interest used for the live animation on the web user interface and transmit these collected data over a

secure link to the user interface application; this app allows the user to visualise the data in a truly unique fashion (Figure 2).

**Figure 2.** System high level functional block diagram.

The electrical infrastructure system is broken down into a number of functional subsystems, including energy harvesting, cell management (dynamic reconfiguration), selfregulation, measurement, data logging and transmission. The autonomous self-regulating system was designed, based on the EcoBot principle [16,17], to mimic a living organism in which actions are governed by energy availability (Figure 3).

**Figure 3.** Cell reconfiguration module design.

The ECS electronics consist of two microcontroller units (MCUs), which are software driven and thus require electronic firmware. The firmware consists of device drivers, diagnostics and applications. The first MCU consists of a power management module, which operates and manages the energy harvester, and other power-related functions. It operates in low-power (sleep) mode and is awakened periodically by a timer. It operates the main power switches that supply the ECS and monitors bulk capacitor voltage and bus power. The second MCU is responsible for all the applications and communications. This MCU is user programmable, and applications can be uploaded via a universal serial bus (USB) link. The applications consist of cell reconfiguration, measurement, data logging, external sensors and other external devices such as pumps and LEDs.

#### *2.3. Bio-Digital Interface*

The interaction infrastructure design aims to capture meaningful data from the MFC system and visualise it appropriately to inform the user experience. The interactive livestreamed animations and live system information are either measured or processed parameter data such as individual cell voltage, cell current, cell power, stack current, stack power, ambient temperature, humidity, rate of change of power, ECS internal voltage, bulk capacitor store voltage, cell configuration information and LED actuation. Data transmission infrastructure enabled the MFC system to be used for the user experience, which involved transferring data from the ECS to the user experience server. This was accomplished via the periodic publication of data by the ECS to a single board computer attached to the ECS, which in turn executes a program to save sensor data to a local MySQL (My Structured Query Language) based database, and via encrypted SSH (Secure Shell) to a remote MySQL database on the external server. The SSH connection is secured with a public/private RSA (Rivest–Shamir–Adleman) key pair, with the private key living on the single board computer and the public key on the server. Once the connection is made, an application saves the data on the server.

Figure 4 shows the established responsive feedback loop between microbial action mediated through the MFC array that generates live data for the user experience, which the participant responds to by interacting with the microbes through animations shaped by live data and, in doing so, generates new data. Participants can choose from several actions, such as feeding the microbes, changing both the ambient and fluid temperature (warming or cooling) and changing the supply rate of feedstock to the microbes. The data from these interactions will feed forward into the live data animations, so that a broader community of participants can see the online events unfolding.

**Figure 4.** Responsive feedback loop between MFC microbial metabolism and planned interactive user experience.

#### **3. Results and Discussion**

#### *3.1. Individual MFC Performance*

Figure 5 shows the average power-generating performance of the 15 MFCs. The polarisation sweep was performed on day 20, when the cells were assumed to have reached a quasi-steady state. The average open circuit (OCV) value was 663 ± 55 mV and the maximum power output was 5.3 ± 0.9 mW (power density: 26.5 W/m3) on average, reaching up to 6.5 mW (32.5 W/m3).

**Figure 5.** Polarisation and power curves of 15 individual matured MFCs measured on day 20. Each data point represents the average value of all MFCs (n = 15).

During the same period, when the MFCs were running individually without being electrically connected through the ECS, the organic content of the feedstock measured in COD decreased by 18.9 ± 2.3, 27.4 ± 7.5 and 19.6 ± 3.9% in the top, middle and bottom rows of the MFCs, respectively. Although the amount of organic consumption differs depending on the location within a cascade, there was no significant difference in the maximum power output. This means that this flow rate (4.0 L/d) and organic concentration (COD of 3.45 mg/L) were sufficient to feed the last cell of each cascade following a sequential cascade treatment.

Several strategies have been established to increase the wastewater treatment efficiency, such as increasing the hydraulic retention time (HRT) by reducing the flow rate or enlarging the cell footprint [18], increasing the operating temperature [19] or increasing the external load [20]. Among these, increased HRT and operating temperature can lead to higher operating costs unless properly optimised. External load changes can be relatively simple to apply at no significant additional cost. Several research groups have reported that appropriately changing the value of the external resistance over a certain period of time, can increase the power output [21,22], as it is related to the microbial metabolic rate. Dynamic reconfiguration, using smart algorithms, is therefore worth pursuing as an efficient MFC operation strategy and is discussed in more detail in the next section.

The MFC reactor design used in this study employed a partial open-to-air cathode, in which 1/2 of the cathode and 1/3 of the membrane (on the cathode side) are submerged in the catholyte. The cathode chamber was initially empty, and catholyte began to accumulate as the MFC started operating. A complete air cathode is a popular choice in the field due to its design simplicity, smaller footprint and lower material cost, but precipitation of salts on the cathode electrode has been reported [23,24], which can be problematic. However, in this design, the ceramic membrane can remain wet, and no significant reduction in power output was observed during the one-year operation period.

#### *3.2. MFC Stack Performance as a Digital Data Generator*

The ECS (electronic control system) is an energy management unit consisting of an energy harvester, power management unit and a mechanism to reconfigure cell connection topologies (cell reconfiguration). During operation, the ECS attempts to harvest the maximum energy while sustaining the MFCs for long-term operation. This is useful as the loading of the harvester on the MFC stack dynamically changes depending on the state of the MFCs, based on feedback [25].

The electrical connection of the 15 MFC units was dynamically changed by the ECS to the following four topologies: all in parallel (15P), all in series (15S), five cells in parallel with the three groups in series (5P3S) and three cells in parallel within the five groups in series (3P5S). Initially, the MFC stack operated at a fixed setting of 15P during inoculation and maturation. An 'adaptive gain' (AG) method, based on dynamic gain allocation and the rate of change of power with voltage clamping in the range of 350~450 mV, was used during subsequent operation. The energy harvester acted as a pseudo-time-varying nonlinear voltage-controlled current sink with a voltage range of 0.15~5 V and a max sink current of 70 mA.

The performance of the energy harvester and MFC stack was evaluated by observation of the mean instantaneous power, mean instantaneous voltage and by the number of actuations per day (Figure 6). The ECS checked energy levels approximately every two minutes and performed data logging and/or actuation depending on the energy acquisition. During the first 200 days of operation, the system harvested a mean instantaneous power of 28.36 ± 5.5 mW, with a minimum of 13.02 mW and a maximum of 36.34 mW; this was at a mean voltage of 339.63 ± 70.75 mV with a minimum voltage of 110.13 mV and a maximum voltage of 477.24 mV at each check. An LED configuration acted as the demonstrable actuation of the 15-MFC system (in addition to all the energy management and data communication) with a 20-s actuation/display; the mean number of 20-s actuations per day was 905, with a minimum of 6 and a maximum of 1578. Overall, the MFC system showed a consistently stable performance, except during days 90–105 when the feed rate was reduced to a minimum (0.7 L/d) due to the Christmas break. Figure 7 shows the MFC system installed in the laboratory actuating the LED lights (top left).

**Figure 6.** Harvester performance: mean instantaneous power variation, mean instantaneous voltage variation and actuations per day.

By utilising a cell reconfiguration mechanism, the system maintains the microbial activity within the MFCs as well as the power output at optimum levels [26,27]. External electrical loading can change the internal impedance and the impedance is varied by changing the connection topology. Here, we create four dynamic impedances which also have an effect on the loading of MFCs, drawing optimum levels of current depending on state of performance. By employing dynamic cell reconfiguration, the MFCs are maintained such that they increase in power output over time until a steady state is reached.

**Figure 7.** The MFC system in action, with the LED lights (located in upper left corner) switched on.

During normal operation, the ECS system was programmed to reconfigure every five minutes, with each cycle being subject to the availability of adequate energy. The control algorithm would adaptively reconfigure the MFCs based on the rate of change of power. Thus, as the rate of change of power increased, the topologies shifted towards a low Thévenin impedance and vice versa [28]. Analysing 200 days of data, a median power generation of 16.21 ± 0.77 mW was maintained with the configurations of series, parallel, 3P5S, and 5P3S at a daily active percentage of 0.17, 0.25, 0.33 and 0.25%, respectively (Figure 8). During low-power-output periods, a mean power level of 29.75 ± 0.35 mW was maintained at a daily activity percentage vector of (0.16, 0.26, 0.35, 0.22)%; likewise, at high power output, a mean power level of 34.77 ± 0.77 mW was maintained at a daily activity percentage vector of (0.1, 0.23, 0.35, 0.32)%. Based on Thévenin resistances of series and parallel networks of MFCs, for a given low impedance load, as the Thévenin resistance increases in a series network, unless the matched load decreases, the MFCs will be under stress to deliver equivalent power. The results corroborate this model, whereby the series topology indicates higher loading in comparison to the parallel one. As power increases, the time spent on parallel and series topologies is less. More time is spent on mid topologies, which enable a balance of loading. It was observed that the system selected the configuration 3S5P. Topologically, this configuration provided an apt balance between parallel and series configurations during peak power-harvesting periods. When power was lowest, the system maintained itself by choosing parallel configurations, which prevented suboptimal performance (also known as 'weakening'), which can result in weaker cells reversing their polarity, so any form of a series topology is harmful. These are findings that can nicely feed into novel AI strategies in further optimising MFC systems, and this will form part of our future work.

Performing as a digital data generator, the overall activity of the MFC-powered biodigital interface was digitally represented. All critical parameters (including, but not limited to, MFC voltages, power, current, system states, temperature) for monitoring the system were transmitted from the ECS to a remote server, where users can monitor and interact with the system.

**Figure 8.** Configuration percentages for low-, mid- and high-power levels.

#### *3.3. Bio-Digital Interface*

A bio-digital interface where users can virtually interact with live microbes in the real MFC system was developed, as shown in Figure 9. Visitors would remotely access a full-screen website, which presents an abstract world of animated microbes ('mobes') moving, feeding and colonising in response to the data streaming from the MFC array installed in the laboratory. Viewers can interactively engage with the mobes simulations and see real-time data, which provides the ability to understand the microbial behaviour affected by the interactions. Technical information about the live system is also available for viewing. A live video feed from the lab, which captures the real-time action of the MFC system that is generating the data informing the animations, is also accessible.

**Figure 9.** Exemplar interactive elements of the mobes user interaction interface (**a**) and exemplar user data visualisation application animating mobes (**b**). Images produced by Translating Nature.

The animated work considers the conceptual commonalities between data and microbes, including the different types of power they each harness (for example, MFCs generate electrical power; data are a power behind many critical decisions), and the social preconceptions which surround both data and microbes (for example: abundance, necessity, cleanliness). The algorithm used as a basis for the behaviours of the mobes in this work was

from [29]. A bacterial foraging optimisation algorithm (BFOA) with social communication is used as a framework for the animated microbes by providing each mobe with a set of rules that generate a random walk within a two-dimensional area. The BFOA is 'seeded' with parameters, such as the number of microbial communities, the population of each and a fitness value. The close-to-real-time fluctuating data from the MFCs then directly affects the mobility of each mobe (direction, speed, step distance) and their attraction or repulsion to the matrix or food. The form of each mobe is designed as a minimal scalable vector graphic (SVG) with flexible anchor points which animate a Bézier curve. As the anchor points change in response to data, the mobes appear to flap and rotate. The animated world which emerges shows the fluctuating nature of the MFC outputs and is an artwork which raises questions about how we translate microbial data as a visual microbial system. It aims to be an experience to convey life, rather than a typical data visualisation, using data as an art material [30] to 'power' the animation.

This participatory technology–art experience is currently available on the website (https://mobes.alice-interface.eu, accessed on 12 December 2021) and is based on the first bio-digital interface powered by MFCs. Providing an interesting and stimulating space for thorough creative encounters with microbes, it generates innovative learning opportunities and striking artistic experiences. This was also exhibited at the Digital Design Weekend (24–26 September 2021, V&A Museum, London, UK) and received a great deal of public interest (Figure 10).

**Figure 10.** Bio-digital interface technology-art experience at the V&A Museum as part of the Digital Design Weekend, London Design Festival, 2021.

#### **4. Conclusions**

The first human–microbial interactive bio-digital interface powered by MFCs was developed and continuously operated over a one-year period. The system operated as a selfsustainable digital data generator, which produced an average power of 28.36 ± 5.5 mW (9.45 W/m3). This level of power was sufficient to power data logging and data transmission, LEDs, system sensing and system control through adaptive dynamic cell reconfiguration. The dynamic reconfiguration allowed better maintenance of the microbial communities inside the MFCs, which resulted in establishing steady states and a consistent performance.

In addition to these technical achievements, the bio-digital interface as a site for digital art is presently scheduled for online public exhibitions. Enabling the visualisation of the otherwise invisible actions of microbes, the data art is expected to inspire users and the general public, thus helping them to understand how a complex biotechnology can be beneficial for society in the long term.

**Author Contributions:** Conceptualization, R.A., R.H., J.F., I.A.I. and J.G.; methodology, A.M., S.W. and J.Y.; software, A.M. and S.W.; formal analysis, A.M. and J.Y.; investigation, A.M. and J.Y.; writing—original draft preparation, A.M. and J.Y.; writing—review and editing, R.A., J.F. and I.A.I.; visualization, A.M., J.F. and J.Y.; funding acquisition, R.A., R.H., J.F., I.A.I. and J.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project has been funded by the European Union's Horizon 2020 Innovation Action Programme, grant no. 851246. IAI is a Bill & Melinda Gates Foundation grantee, grant no. INV-006499.

**Institutional Review Board Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Towards Energy-Positive Buildings through a Quality-Matched Energy Flow Strategy**

**Nick Novelli 1,\*, Justin S. Shultz 2, Mohamed Aly Etman 1, Kenton Phillips 2, Jason O. Vollen 3, Michael Jensen <sup>4</sup> and Anna Dyson 1,\***


**Abstract:** Current strategies for net-zero buildings favor envelopes with minimized aperture ratios and limiting of solar gains through reduced glazing transmittance and emissivity. This load-reduction approach precludes strategies that maximize on-site collection of solar energy, which could increase opportunities for net-zero electricity projects. To better leverage solar resources, a whole-building strategy is proposed, referred to as "Quality-Matched Energy Flows" (or Q-MEF): capturing, transforming, buffering, and transferring irradiance on a building's envelope—and energy derived from it—into distributed end-uses. A mid-scale commercial building was modeled in three climates with a novel Building-Integrated, Transparent, Concentrating Photovoltaic and Thermal fenestration technology (BITCoPT), thermal storage and circulation at three temperature ranges, adsorption chillers, and auxiliary heat pumps. BITCoPT generated electricity and collected thermal energy at high efficiencies while transmitting diffuse light and mitigating excess gains and illuminance. The balance of systems satisfied cooling and heating demands. Relative to baselines with similar glazing ratios, net electricity use decreased 71% in a continental climate and 100% or more in hot-arid and subtropical-moderate climates. Total EUI decreased 35%, 83%, and 52%, and peak purchased electrical demands decreased up to 6%, 32%, and 20%, respectively (with no provisions for on-site electrical storage). Decreases in utility services costs were also noted. These results suggest that with further development of electrification the Q-MEF strategy could contribute to energy-positive behavior for projects with similar typology and climate profiles.

**Keywords:** on-site net-zero electricity; energy-positive buildings; active integrated facades; thermal storage; distributed systems; exergy management

#### **1. Introduction**

Measures employed in the development of net-zero and net-generating building projects are determined through a design framework, whether implicit or explicit, and any such framework must address multiple criteria to develop traction and ultimately enjoy uptake. These criteria are occupant-related, physics-based, engineering and manufacturing, logistical, and economic in nature. One such framework, termed Quality-Matched Energy Flows (Q-MEF), is explored in this study, and focuses on maintaining and applying the inherent value of climatic energy resources as they pertain to the service demands of the built environment—namely, lighting, electricity supply, and heating and cooling. Specifically considering on-site solar resources, Q-MEF is explored through simulation with a specific generalized building type, in multiple climates. Measures proposed through the Q-MEF rubric comprise a novel Building-Integrated, Transparent, Concentrating Photovoltaic and

**Citation:** Novelli, N.; Shultz, J.S.; Aly Etman, M.; Phillips, K.; Vollen, J.O.; Jensen, M.; Dyson, A. Towards Energy-Positive Buildings through a Quality-Matched Energy Flow Strategy. *Sustainability* **2022**, *14*, 4275. https://doi.org/10.3390/su14074275

Academic Editors: Oz Sahin and Edoardo Bertone

Received: 26 December 2021 Accepted: 15 March 2022 Published: 4 April 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Thermal fenestration system (BITCoPT) coupled to an interconnected network of thermal storage and heat pumps (both thermally and electrically powered) which are distributed throughout the building. BITCoPT (investigated in earlier works through modeling, simulation, and experimentation [1–3]) uniquely enables the Q-MEF strategy by providing daylighting (illumination and glare control) in addition to generating electricity and collecting thermal energy from solar energy incidents on the building envelope. As Q-MEF and BITCoPT are architecturally grounded, they draw on multiple realms of expertise, and the context of these realms—and the state of their technology—is explored.

#### *1.1. Addressing Limits of Load-Reduction Strategies with Active Envelope and Thermal Distribution*

Traditionally, net-zero energy building (NZEB) strategies follow a three-tier approach to energy efficiency: load reduction is foremost, followed by the application of passive systems such as shading, and lastly the use of active systems such as energy harvesting [4]. Although load reduction is important, in contemporary architecture this approach begets façade designs that supply good daylighting and natural ventilation, but do not attempt energy harvesting [5,6]. The proliferation of on-site NZEB has been stalled by these legacy design frameworks, which remain grounded in efficient use of grid-supplied electricity and fuel, and employ envelope designs that offload solar energy gains through insulating, limiting apertures, shading, and glazing emissivity control. Moreover, prevalent buildingintegrated energy harvesting systems (such as roof-top and façade-integrated PV) exhibit sub-optimal exergetic efficiency and limited installable area, limiting the addressable fraction of a building's service demands, even though in many scenarios solar exergy available to a building is far greater than its own consumption.

Here, we propose an alternative approach, manipulating incident solar energy into daylighting (with minimal glare production) by intercepting that energy with an opticallyconcentrating photovoltaic fenestration system and hydronic mechanisms that capture and transfer "waste" heat. In this approach, solar gain is re-cast from a low-grade "load" to be remediated through energy intensive heating, ventilation, and air-conditioning (HVAC) systems into a high-grade (thermodynamically useful) energy source for cooling (and other) systems. This approach (Q-MEF) suggests re-distributing a building's systems out across its envelope to more-efficiently interface with the distributed solar resource. This distribution represents a revision of the traditional system topology that has evolved around grid-sourced energy supplies, of centralized plants and trunk-and-branch systems (Figure 1).

**Figure 1.** Transitioning from centralized (grid-source optimized) systems to distributed, integrated systems (climate-sourced optimized), matching resource (sunlight) to service demands (daylighting, power, cooling, heating).

If active control is thus shifted outward, the building envelope is empowered to redistribute the abundant light, heat, and power resources available, transferring solar energy towards the building's particular demands and, when necessary, transforming incident solar energy into consumable forms. In Q-MEF, this shift is implemented through integrating distributed coolth storage, (thermally driven) adsorption chillers, and a transparent, optically concentrating photovoltaic and thermal collecting fenestration system (BITCoPT). BITCoPT transmits diffuse irradiance for daylighting and views, and strips out direct normal irradiance, either transforming insolation into electricity or collecting it as thermal energy. Rather than exacerbating cooling demands, insolation can therefore be used directly, as electricity, or as thermal energy that supplies small, localized chillers. Local, distributed use of heat is important for avoiding losses from piping that energy to a central plant. The exergetic value of collected thermal energy is sensitive to such temperature losses, and the surface area-mass ratio of the system (potential for losses through insulation) is inherently high compared to centralized distribution networks with larger pipes. Generated heating and cooling capacities are stored as warm and cold water and circulated through the building via thermally massive hydronic circuits, redistributing these resources towards demands for cooling as well as heating and other process energy, maintaining small but useful temperature differences rather than immediately dispersing generated potential into the built environment. In concept, this integrated system of the multifunctional envelope collector and activated thermal mass distribution can uniquely maintain the quality of the solar resource (through daylighting), and its exergetic content, by collecting a high fraction of direct irradiance either as electricity or as thermal energy at elevated temperatures.

#### *1.2. Precedents: Active Facades*

The Q-MEF strategy builds on precedent and contemporary building-scale integrated strategies and is dependent on multifunctional envelope technologies. As such, a review of precedent active integrated envelopes and facades (AIFs) is relevant. Buildingintegrated technologies have been reviewed in such focus areas as photovoltaics [7], thermal collection and life cycle analysis [8], concentrating technologies [9,10], and architectural integration [11]. To the authors' knowledge, an investigation coupling a highconcentration/daylighting system such as BITCoPT to a full-building scale application and distribution strategy—designed specifically to leverage envelope-wide collection—has not yet been undertaken.

#### *1.3. Benefits of Distributed Systems*

Beyond deficits in energy harvesting capacity, the drawback of load-reduction NZEB strategies is in the utilization of harvested energy: Although high quality flows of electricity, light, and thermal energy can be developed with envelope-integrated systems, the building's mechanical infrastructure must be designed to apply these flows effectively. Specifically, generated electricity and thermal resources must be managed. DC microgrids can be employed to efficiently transfer envelope-generated electrical power to equipment and lighting, which often requires DC power natively, using available DC-DC conversion [12]. Harvested thermal resources can be distributed with hydronic systems if the network topology is designed to maintain the useful temperatures (and therefore exergy) that can be developed by a concentrating-type envelope collection system.

Traditionally, building mechanical systems are centrally organized as a logical result of the availability of grid electricity and fuels. Current NZEB strategies therefore involve centralized systems as well (save for active shading/daylighting and natural ventilation). Even on-site harvesting of electricity and heat is typically routed through the same wiring and plumbing that are optimized for the centralized resource-generating systems. To take advantage of envelope-generated energy resources, the mechanical processes that use these resources can be scaled down, parallelized, and located at the envelope, distributed in the same fashion as energy resources are harvested. This is especially relevant for thermal energy collected at high temperatures (beyond roughly 25 ◦C above the comfort range), which endures larger transport losses than electricity. Recent research into distributed ventilation and humidity control [13] and DC microgrids [14] indicates the inherent efficiency and demand–response benefits of co-locating processes with both their sources of power and the demands they service within a building. Analysis of coupling these end uses to a polygenerating envelope such as BITCoPT has not yet been done.

Many technological elements of load-reduction NZEB strategies can be successfully reconfigured to the distributed Q-MEF strategy. Radiant and mixed-mode heating and cooling, for example, is efficient at managing sensible thermal loads when coupled with effective ventilation and latent-controlled mechanisms such as dedicated outdoor air systems (DOAS) [15,16]. In standard practice, the heat and coolth these systems distribute are generated by centralized chillers and boilers. However, they integrate just as well with distributed thermal control and are employed in the Q-MEF framework.

#### *1.4. Thermal Energy Storage for Resource-Matching and Dispatchability*

In contrast to grid-sourced energy, which is available on-demand, climatic energy resources must be collected when they are available. To make these resources useful, they must be made dispatchable. Therefore, storing energy in a recoverable way is useful. The benefits of energy storage additionally extend to load factor reduction, grid efficiency and stability, primary fuel costs and pollution reduction, equipment longevity, and resiliency [17]. A range of technologies and materials are actively investigated to these ends [18].

On-site fluidic thermal energy storage (TES) at temperatures suitable for heating (over 30 ◦C) and cooling (under 18 ◦C) is attractive because of the potential for good round-trip efficiency: Losses can be minimized through proper insulation, and fluid pumps can be used to dispatch thermal power with little power consumption [19]. Although broadly applicable [20], on-site thermal storage can be limited because of space requirements as well as the relative unfamiliarity of Architecture, Engineering, Construction, and Operation (AECO) sector industries. The current commercial solution to this spatial limitation is to deploy ice storage [21]. Research into alternative phase-changing materials and systems that melt at higher temperatures than water [22–24] could improve system efficiencies (by reducing required temperature lifts and maintaining optimal temperatures for process efficiencies) and design desirability (by reducing system footprint). For the sake of conceptual clarity, in the present study, water is chosen as the heat transfer fluid and storage media. The media is well-characterized, but the system geometric configuration—with thermal mass distributed throughout the building's floorplate, along the fluid loop—is uniquely configured to distributed generation and, as such, adds a novel dimension to this investigation.

TES at three temperature ranges is important in the Q-MEF strategy, corresponding to the supply, re-cooling, and output circuits of thermally driven chillers. Because commercial buildings have significant cooling loads and electricity is typically used to create cooling, on-site coolth TES can reduce site energy use, as well as peak demands on the grid, by phase-shifting cooling [25].

In addition to the cool temperature range, TES at two higher ranges is important. In support of the envelope-integrated thermal collection, a buffering mechanism is required to match high-temperature collection to the operation of a thermally driven chiller. Hightemperature storage has been deployed for utility-scale thermal power cycles such as solar electric plants [26], but the lower temperatures useful for driving adsorption chillers (above 60 ◦C) are easier to maintain in the building context.

A mid-range TES (at temperatures above the range of thermal comfort but below 50 ◦C) is also required and considered. This storage can act as both a sink for heat rejected from the chillers' operation and a source for heating processes, either as pre-heating for service hot water (SHW) demands or as space heating. Additionally, it has been shown that by storing rejected heat until night-time, when the temperature of the atmospheric sink is lower, the rejection process (and overall energy use) can be more efficient [27].

#### *1.5. Entropic Efficiency of Thermal Cascade*

By defining multiple temperature (or quality) ranges for thermal storage, a cascade develops through which harvested thermal resources can be matched to a building's service demands (Figure 2). Besides the demands discussed here, other uses such as humidity control through desiccant regeneration have been investigated. This cascading framework has been recognized in other research, such as the Building as Power Plant concept [28], and in studies and optimizations of latent thermal storage in the building context [23,29]. However, to the authors' knowledge, including daylighting as a primary product and analysis through the mechanism of a multifunctional envelope has not been undertaken.

**Figure 2.** Cascade of thermal potentials for servicing built environment demands. Optimal exergy application performed by cascading heat through and into processes that require different grades (temperatures) of energy. In the current Q-MEF investigation, thermal storage in three discrete ranges (driving, rejection/heating, chilled) governs energy flow between processes.

#### *1.6. Valuing Different Forms of Energy*

To evaluate Q-MEF as a strategy, a rubric is devised for the value of the different energy flows—lighting, electricity, heat, and coolth—that can be gleaned from solar energy by BITCoPT. The relative values of these different flows must be grounded in reasonable goals for a given context [30].

Because well-distributed lighting with good color rendering is difficult to reproduce in the built environment, effective daylighting is a priority in a high-performance building. Diffuse sunlight is multi-directional, exhibits good color rendering, and has good luminous efficacy (relative to common tube fluorescents, although LED lamps are targeted to surpass [31,32]). Diffuse sunlight is easier to apply as daylighting than direct sunlight, which is typically strong and narrowly directional, causing overheating and glare through high contrasts and distracting reflections.

Of the resources into which direct irradiance can be transformed, electricity is more valuable than thermal energy, as it is easy to transduce into other forms of work, and its exergy content is not directly subject to the Carnot efficiency. DC electricity is objectively more efficient to use than AC power in many commercial building contexts, as most modern electrical equipment is DC-native, and DC-DC conversion efficiency has reached parity with AC-DC rectification. Although DC distribution is not yet common, standards are mature and gaining acceptance for integration of multiple sources, storage mechanisms, and use devices into sub-building scale microgrids [12].

Thermal energy stored as a heat transfer fluid (HTF) at high temperatures (here defined as over 60 ◦C, in the context of building systems) can be applied to drive processes such as adsorption chilling, so it can both do work and be applied as heat, in sequence. The higher the temperature developed in the working fluid (relative to an available thermal sink such as the atmosphere or earth), the greater the exergy content, if losses incurred in storage and transportation are managed.

In this study, cooling power is more valuable than heating power for two reasons: More cooling is typically required in medium-sized office buildings, and the coefficient of performance (COP) of a reversible heat pump acting as a chiller is lower than that of the same pump in heating mode.

#### *1.7. Exergy-Efficient Solar Collection Aligned with Façade Criteria*

Successful on-site energy strategies will internalize a large fraction of the solar exergy available to a building by integrating with the multiple objectives of a successful envelope objectives such as thermal regulation, daylighting, privacy, and design identity. The method proposed here for gleaning maximum utility from solar energy is represented by the thermal/energy cascade (Figure 2), and, to functionalize it within the Q-MEF framework, a model is incorporated of BITCoPT, which has previously undergone development [33], simulation [34], and prototype characterization [2,3] (Figure 3).

**Figure 3.** Multifunctional envelope-integrated solar collector: BITCoPT. Model of BITCoPT incorporated into Q-MEF modeling to represent spatially simultaneous daylighting and cogeneration from glazed envelope areas.

#### *1.8. Comparison to Alternative Building Energy Schema*

Existing schema cross over in purpose and methods with Q-MEF. Although wholebuilding polygeneration is typically considered component by component, as evidenced by review work on net-zero strategies [35], the impact of intersectional modeling is recognized [36]. Notably, high-performing built works achieve performance through combined strategies such as occupant-driven energy savings and multi-functional envelope structures. The Bullitt Center's canopy exemplifies this, reducing gains on the vertical envelope while increasing functional area for traditional photovoltaics [37]. Buildings such as Power Plants, mentioned earlier, emphasize the efficiency gains and building demand-matched resources from on-site cogeneration [28]. Exergy analysis provides a rubric for identifying appropriate quality matches between resources and demands [38] and for valuing both warming and cooling storage mechanisms according to temperature differences against an ambient state [39]. The Low Exergy (or LowEx) design emphasizes the application of thermal resources to the small-potential thermal demands of the built environment (heating, cooling, humidity control), the efficiency of heat pumps at maintaining small, useful temperature differences, and sequencing heat pumps to obtain larger temperature lifts [40,41]. LowEx has been extended to district scales [42]. Thermally Active Building Systems (TABS) emphasize the utility of hydronic thermal redistribution, activating thermal mass within the occupied regions of a building, and the efficiency benefits of transferring heat across the large surface areas that comprise indoor built spaces [43,44]. A Passive House is specific regarding the reduction of thermal loads, which can limit solar harvesting opportunities, although work is ongoing to integrate the strategy with on-site generation [45]. These frameworks can be employed to design systems that maximize wise use of available energy resources, but they do not explicitly tackle the provision of desirable lighting conditions in the built environment.

More holistically, Integrated Design emphasizes the consideration of an architectural design as a dynamic totality that shifts between states in a knowable fashion. Attention is paid to the thermal properties (particularly diffusivity) of a building design at the material, construction, and space scales [46,47]. Bioenergetic modeling emphasizes the pathways of potential, flow, and transformation for a broad category of resources, distributed sensing and response, and dynamic systems at the boundary between built and natural environments [48]. Furthermore, thermo-economic analysis is ongoing to ascertain the value of exergetic efficiency and on-site coupled generation and energy storage [49–53], and novel interpretive tools are developed towards these ends [54]. An intention of developing the Q-MEF framework is to leverage the insights of these other strategies to provide for a wide reach of building service demands, including quality lighting.

#### *1.9. Research Objective: Modeling Q-MEF Building Energy Behavior*

Overlapping with these complementary schemas, Q-MEF incorporates similar concepts. Q-MEF's combination of daylighting-primary operation, distributed high temperature buffering and active chilling, and distributed active thermal mass is potentially novel. Hence, this present work investigates Q-MEF's capacity to more fully leverage building-incident solar energy—and contribute to net-zero and energy-positive design through integrating BITCoPT with thermal management at multiple temperature ranges and controlled redistribution.

As a platform for this investigation, a representative medium-scale office building with Q-MEF-aligned systems was modeled and tested in diverse climates, by first generating a pair of baseline (BL) building energy models with both moderate and high window-wall ratios (representing a desirable condition in commercial architecture), and then revising the high-ratio configuration to represent a full Q-MEF implementation with experimentally validated transparent active solar façade energy collectors, distributed thermally driven chillers, sequential thermal redistribution, and thermal storage at three temperature ranges. A fourth building energy model was also generated and tested, wherein the geometry of the building was modified to harvest additional solar energy. The Energy Use Intensity (EUI), peak demands, thermal load profiles, and economic benefits (based on energy consumption) that resulted from this matrix of full-building-scale simulations were compared, and salient behaviors were highlighted.

#### **2. Methods**

To evaluate the effects of Q-MEF on a building's energy flows and consumption, a model was constructed by combining (through post-processing) precursor building energy models of a medium-scale commercial office building (generated in the EnergyPlus environment) with an analytical model of an active façade system (BITCoPT). In addition to these two primary precursors, the Q-MEF model includes separate functions representing: hydronic thermal storage and redistribution, thermally driven chillers, electrically driven auxiliary heat pumps, a deep-mullion curtain wall cavity, fan-powered air volume exchanges in that cavity, and reactive (not predictive) controls. The models and functions were integrated through post-processing in a time-step fashion. The block diagram of the Q-MEF model is presented in Figure 4, and is referred to subsequently.

**Figure 4.** Q-MEF model block diagram, showing interfaces between precursor models and additional functions.

The model was configured in four ways, with two baseline options (low and high glazing ratios) and two options including Q-MEF components, which were both highly glazed, but with either normal (vertical) or outward-tilted fenestration. Three climates were analyzed, for twelve total model configurations (Figure 5).

**Figure 5.** Building configurations and components from modeling environments.

#### *2.1. Systems for Thermal Energy Collection, Redistribution and Use*

In the Q-MEF rubric, solar energy on a building's glazed envelope areas is either transmitted as daylight, or concentrated and converted to electricity, or collected as thermal energy, where it is then passed through subsystems, applied to do work (such as chilling), and released as space or water heating, or directly out to climatic sinks. The modeled HVAC strategy (shared between all four configurations) is based on a four-pipe distribution that services all the building's conditioned zones and envelope collection areas (Figure 6). A chilled distribution loop is maintained at sub-ambient temperatures. A rejection-stage (heating) loop was maintained (through system control algorithms) at super-ambient or near-ambient temperatures, depending on the average demands for heating or cooling which varied seasonally.

Two parallel heat pump systems operate between the loops. In the Q-MEF configurations, small-capacity (<10 kW) adsorption chillers are distributed in the region of the building's envelope and driven by collected solar thermal energy, which is stored adjacently in small buffer tanks. Additionally, in all four configurations, electric water–water heat pumps act in parallel between the two loops.

In both the baseline and Q-MEF configurations, heating and cooling in a zone was modeled as transferring energy through baseboard units or chilled beams, from the rejectionrange loop or to the chilled loop, respectively.

**Figure 6.** Plumbing schematic showing chilled and rejection-range distribution. Storage or distribution applied at three temperature ranges. Distributed adsorption chillers and auxiliary chiller operate in parallel between heating/rejection loop and chilled loop to maintain operating temperatures.

In the Q-MEF model configurations, thermal energy collected by BITCoPT was both used to drive adsorption chilling, and as heat. The glazed areas of the precursor BEMs were defined as BITCoPT collectors [34]. In post-processing, each collector fed a thermal buffer tank, which in turn was plumbed to a chiller ("Buffer/Chiller Operation" block in Figure 4). When a buffer's temperature (*Tbuffer*) rose high enough to drive the connected chiller at a reasonable COP (over 0.5, requiring *Tbuffer* ≥ 60 ◦C, depending on rejection loop temperature), the chiller engaged, pumping heat from the chilled loop to the rejection loop. If the chilled loop became too warm, or the rejection loop cooled off too much, auxiliary heat pumps engaged, pumping heat in parallel to the thermally driven chillers.

Pre-heating for service hot water demands and the DOAS air intake were modeled to occur through water–water heat exchangers, subtracting energy from the rejection-range loop balance according to the difference between the supply temperature reported in the BEM and the loop temperature, less a 2.0 ◦C approach (Figure 6). The quantity of pre-heating was subtracted from boiler usage of the precursor BEM on a time step basis.

By modeling envelope-integrated solar thermal collection with distributed thermal systems and circulating thermal mass, it was possible to represent the application of thermal energy both as a driving force (for adsorption chillers, at high temperatures) and as useful heat (for zone demands and pre-heating), taking advantage of both the exergy and energy value of the collected energy.

#### *2.2. Inputs to Q-MEF Model*

Inputs to the overall Q-MEF analysis are described in this section, including precursor models (building energy model baselines, the BITCoPT model, and daylighting method) and boundary conditions (climates).

#### 2.2.1. Climates Considered

The Q-MEF model was analyzed in three climates, exploring a range of system response patterns (Table 1, where *Tdb,outd* is outdoor dry bulb temperature). New York City (NYC, LaGuardia Airport TMY3 data) is a fluctuating climate: continental and seasonally humid, with varying weather and strong seasonal swings. Although the city is characteristically dense, and sites are typically shaded by adjacent structures, because this study parametrized climate context, not site, no external shading was defined in NYC or elsewhere. Phoenix (PHX) is an arid subtropical desert with a strong solar resource, high mean temperatures, and large diurnal swings. Mountain View (MTV) is in a semi-arid Mediterranean climate with mild temperatures and a significant solar resource.


**Table 1.** Climates used for analysis.

#### 2.2.2. Building Energy Models as Baselines

Four building energy models were developed (using the Open Studio [55] interface for EnergyPlus [56]), and used in two ways: directly, as two baseline configurations; and as precursors for two Q-MEF configurations. Each energy model, whether it was used directly as a baseline or as a precursor to the Q-MEF model, was configured for three distinct climates (Table 1) according to ASHRAE 90.1 (2013) building standards, resulting in a set of twelve analyzed configurations. Configurations are summarized in Table 2.


**Table 2.** Model

configuration

 parameters.

The first baseline configuration (BL40) was a building energy model with no postprocessing, based on the United States Department of Energy's (DOE) Medium Office Commercial Reference, with ASHRAE 90.1 (2013) specifications [57]. Default operation and occupancy schedules were used, including 5-day work weeks throughout the year, with reduced occupancy and thermostat setbacks outside working hours. The window-wall ratio was 40%. BL40 s HVAC systems were assigned per ASHRAE 189.1: hydronic thermal distribution, with zoned chilled beams, baseboard heating, and DOAS. These updates slightly reduced energy use, relative to the DOE reference.

Representing architectural trends towards highly glazed facades, the second baseline configuration (BL95) had a 95% window–wall ratio. The glazing specifications and systems behaviors were held constant, though systems were re-sized (through the auto-sizing algorithms in EnergyPlus). BL95 was the precursor to the Q-MEF configuration.

For the precursor BEM to the Q-MEF + Tilt configuration, the morphology of the Q-MEF BEM was revised to increase the solar gain on the building. At each floor, facades were swung upward at 20◦ (see Figure 5). The chosen angle maximized acquirable irradiance by maximizing solar flux through the exterior glazing on a per-area basis, while limiting the shading of a floor by the floor above.

#### 2.2.3. BITCoPT Envelope Cavity Model

A model of the BITCoPT façade collector [34] was integrated into the Q-MEF model, contributing electrical generation, thermal collection, and direct solar gain reduction based on inputs of climate data and envelope orientation. A constant efficiency (*ηconv,Egen* = 0.96) was applied to the model's electrical output to represent the transformation of the collector's variable-voltage DC output to a useful form (constant-voltage DC for tying into to zonelevel microgrids, or AC for tying into building distribution).

The thermal collection output was likewise modified to represent the heat transfer fluid inlet and envelope cavity temperatures in the Q-MEF model, which were distinct from those values in the BITCoPT simulation. Because the cavity temperatures (*Tcav*) in the Q-MEF model were held lower (*Toutd,db* < *Tcav* < 38 ◦C, less than 45 ◦C in the precursor model, to be realistic for mechanical control components), the Q-MEF-modeled array lost more heat, and thermal collection efficiency was lower than in the precursor model.

#### 2.2.4. Daylighting Modeling Method

All baseline and precursor BEMs employed daylighting controls. Daylighting algorithms in EnergyPlus modified interior lighting schedules according to illuminance measured by virtual sensors. In a prior study [58], a threshold was noted in the room depth for which sufficient work-plane illuminance could be achieved when BITCoPT was introduced, but the perimeter zone depth here (4.57 m) was shallower than that threshold.

Precursor BEMs employed in this study did not include skylights, in keeping with the referenced DOE models. Daylighting from skylights would reduce the baseline lighting electrical loads in the core of the third floor (which comprises 20% of the building's floor area) but would likely contribute to net cooling loads due to solar gains. Although it is surmised that the zones' daylighting and thermal circumstances would improve with glazed roof expanses and installed BITCoPT, in this study the roof collection was treated as stand-alone equipment which did not interact (through lighting, or thermally) with the adjacent zones (in contrast to the vertical glazing expanses, for which light and thermal interactions were modeled).

#### *2.3. Q-MEF Model Components and Functions*

To represent the impacts of the Q-MEF strategy, components were incorporated with the collection of inputs through additional modeled functions.

#### 2.3.1. Thermal Energy Storage Elements Model

Two types of thermal energy storage elements (TES) were modeled: high-range buffer tanks, and distribution loops. One buffer was defined for each solar collector. One cooling loop and one rejection-range loop were defined to service the building. Both TES types were defined to be well-mixed, rectangular water tanks with a fixed height and width. (Although higher-capacity thermal storage media such as PCMs might increase the overall performance of the Q-MEF system, water was modeled in this study for the sake of conceptual clarity.) As indicated in Figure 4 ("Buffer/Chiller ... " and "Distribution Loops" modules) the TES exchanged heat with the building's systems (controlled) and with the interior environment (uncontrolled, due to losses through their insulated surface areas). Losses were modeled as the heat transferred across the TES boundary (tank walls) assuming insulation of 100 mm of polyisocyanurate foam, and negligible film coefficients on both interior and exterior surfaces, for an effective thermal resistance of *RTES,wall* = 0.4 W/m2-K. The heat lost from the buffers transferred to the curtain wall cavity in which BITCoPT was installed. The heat lost from the distribution loops transferred to the occupied zones. Pumping power required to equalize the temperature between the storage elements in each loop was defined as equivalent to the power used by the equivalent circulation pumps in the contributing building energy model. Buffer TESs were each sized (through iterative testing) to maximize the yearly output of their paired chillers. The distribution loop capacities were sized to minimize whole-building electrical use (Table 3).

**Table 3.** Thermal capacities of modeled thermal energy storage (example: NYC climate).


It can be noted that in an earlier (unreported) configuration of the Q-MEF model, a high-temperature-range distribution loop was considered as well, to simplify the work extraction from that source. However, the energy losses through the system insulation to indoor ambient temperatures were too high to justify, so a distributed configuration and 4-pipe distribution (reported here) were adopted, with paired buffers and adsorption chillers.

#### 2.3.2. Deep-Mullion Cavity Energy Balance

The building envelopes for the baseline and precursor models were modeled natively in EnergyPlus. For the Q-MEF configurations, the deep-mullion curtain wall cavity (into which BITCoPT integrated) was represented by an energy balance in post-processing. Energy transfers across the glazing determined in the precursor BEMs were replaced by equivalent transfers as determined by this energy balance. The function did not account for variations of temperature or fluid movement within (or external to) the cavity control volume, or non-homogeneous masses. Overall conductivity values between the cavity and adjacent environments were constant, whereas these relationships were variable in the precursor BEM. More accurate heat transfers would be expected from a cavity model that included convection, surface emissivity, material diffusivity, and thermal bridging effects, but one-dimensional, steady state assumptions were deemed sufficient to contrast behavior between baseline and Q-MEF model configurations.

The energy balance comprised: the transmittance of direct and diffuse solar energy into and out from the cavity; the thermal and electrical energy generated by BITCoPT; the heat transferred (via conduction and convection) across glazing surfaces with the two

adjoining environments; the heat extracted from the cavity by a flushing function; and the heat lost from high-range buffers (Figure 4: "Cavity balance" block). The balance was

$$\begin{aligned} \left(\sum m\_{\text{cav}} \mathbf{c}\_{p, \text{cav}} \frac{dT}{dt}\right) \\ &= \mathbf{G}\_{\text{DN}, \text{cav}} + \mathbf{G}\_{\text{DN}, \text{ind}} + \mathbf{G}\_{\text{D}ff, \text{av}} + \mathbf{G}\_{\text{D}ff, \text{ind}} - \mathbf{Q}\_{\text{gen}} - \mathbf{E}\_{\text{gen}} \\ &+ \mathbf{Q}\_{\text{cond}, \text{outd}} + \mathbf{Q}\_{\text{cond}, \text{ind}} + \mathbf{Q}\_{\text{cav}, \text{f}, \text{lub}} + \mathbf{Q}\_{\text{buff}, \text{loss}} \end{aligned} \tag{1}$$

where *mcavcp* refers to the cavity's thermal mass (in kJ/K); *G* to transmitted solar power (in W); *Q* to thermal flow (in W); *Egen* to electrical generation (in W); *DN* to direct irradiance; *Diff* to diffuse irradiance; *ind* and *outd* to the indoor and outdoor environments; *cond* to non-irradiation thermal transfer; *cav* to the cavity; and *flush* to the cavity flushing function (see Section 2.3.4).

#### 2.3.3. BITCoPT Area-Based Gap Transmittance

The direct irradiance transmitted through the cavity to a building zone was modeled by multiplying the transmitted irradiance reported by the BEM by *Tgap*, a function that represents the collector's area-based transmittance of direct irradiance (the "Gap function" block in Figure 4). When the solar vector is near to normal with the surface of an envelope region that incorporates a collector, insolation passes through the gaps between BITCoPT modules. As the solar vector moves away from envelope-normal and the collector modules track around pitch and yaw axes, the gaps decrease, falling to zero width at an excursion angle determined by the collector's geometry. *Tgap* was defined as

$$T\_{gap} = (1 - c\_{vert}) \left(\frac{\phi\_{full} - \phi}{\phi\_{full}}\right) (1 - c\_{lz}) \left(\frac{\lambda\_{full} - \lambda}{\lambda\_{full}}\right) \tag{2}$$

where *cvert* and *chz* are vertical and horizontal components of the fractions of envelope area filled by BITCoPT lens modules; *φ* and *λ* are the rotations of the BITCoPT modules around their pitch and yaw axes (in radians); and *φfull* and *λfull* (radians) are the respective threshold angles where the gaps between lenses decreased to nil when observed parallel to the solar vector. The floor for *Tgap* was 0, when modules are rotated beyond the pitch and yaw thresholds, and no direct irradiance was transmitted. *Tgap* was applied to only direct insolation since, as modeled, BITCoPT does not intercept diffuse insolation.

#### 2.3.4. Cavity Flushing Function

A cavity flushing function was implemented to simulate the removal of heat from the BITCoPT cavity, adjusting cavity temperature to complement the heating or cooling demands of the adjacent zone. The function simulated a fan moving air between the cavity and the environment (Figure 7). Flushing with outdoor air only was modeled, and the function remained decoupled from the building DOAS.

**Figure 7.** Modeled method of flushing excess heat from the cavity.

Flushing was implemented both in the precursor BEM (as a window:Airflow property in EnergyPlus, with a constant volumetric rate of 0.6 m3/s-m and assumption of no fan power) and in post-processing. In post-processing, the maximum allowable cavity temperature was *Tcav,lim,high* = 40 ◦C, but if zone cooling demand was significant, the target temperature was set to the outdoor (dry bulb) temperature). The post-processing flushing function was modeled as:

$$\begin{aligned} Q\_{\text{cav},fulls(i)[n]} &= K\_{fulls(i)} \ast \mathbb{C}\_{\text{cav}(i)} \ast \left( T\_{\text{cav}[n-1]} - \left( T\_{\text{cav},target} + T\_{\text{cav},offset} \right) \right) \\ &+ K\_{fulls,IND(i)} \left( I\_{DN,\text{cav}(i)[n]} \right) \end{aligned} \tag{3}$$

where *Qcav,flush* was the resulting thermal flow (in W); *Kflush* was the (non-dimensional) proportional gain tuned for each zone; *Ccav* was the cavity thermal mass (in kJ/K); *Tcav* was the cavity temperature; *Tcav,target* was set to either *Toutd,db[n*−*1]* or *Tind* depending on heating demand; and *Tcav,offset* = 2 ◦C was used to establish a dead band, preventing the system from operating if heat removal would be inconsequential. *Kflush,IDN* was a separate gain used to modify the flush rate according to zone's direct insolation at the current-time step. Fan power required to flush the cavity was determined assuming a constant pressure head (and therefore a constant power draw) multiplied by an hourly duty cycle. 100% of fan (electrical) power was designated to be taken up by the airflow (as increased temperature).

These flushing controls maintained the cavity temperature close to the chosen target, minimizing unwanted non-insolation thermal transfer from the envelope cavity to the building's interior, and contributing energy to under-heated zone conditions. Gains were tuned to optimize net generation at the building meter.

#### 2.3.5. Thermally Driven Chillers Model

Adsorption chillers were modeled by fitting a solution surface to a manufacturer's COP data [59] relative to temperatures of buffers, chilled loop (*Tchilled*), and rejection loop (*Trej*). Upper and lower limits on *COPchiller* were implemented to keep it between 0 and 0.56. The chillers were controlled to activate when the temperature of their paired buffer exceeded a threshold (*Tbuffer* > 60 ◦C). Chillers drew energy according to the excess in the buffer over the threshold, with no capacity limit and no dependence of COP on the fractional capacity, representing multiple chillers ganged in parallel.

#### 2.3.6. Auxiliary Heat Pumps and Heat Rejection Model

Water–water heat pumps were incorporated in the thermal control strategy to maintain the temperatures of the chilled and rejection-range distribution loops (see Figure 4, "Distribution Loops" block). If the chilled loop became too warm, energy was pumped from it to the rejection-range loop. Pumps are modeled steady state, at a COP varying with temperature lift, with functions sourced from the literature [41], capped at 16.0, with an 0.5 exergetic efficiency.

To prevent the rejection-range loop from getting too warm, heat removal from the loop to the outdoor environment was simulated by modeling a dry cooler, or fan-assisted water-to-air heat exchanger. The heat removal function, *Qrej,env* (in W) was:

$$Q\_{rej,env[n]} = K\_{rej} \* C\_{loop,rej.} \* \left( T\_{rej[n-1]} - T\_{rej,target} \right) \tag{4}$$

where *Krej* was the (non-dimensional) proportional gain; *Cloop,rej* was the thermal mass of the rejection loop (in kJ/K); *Trej* is the rejection distribution loop temperature; and *Trej,target* was the temperature set point. Like with the cavity flush fans, the fan power required to remove heat was determined with a constant power draw and an hourly duty cycle.

#### 2.3.7. Loop Temperature Controls

The temperature of the chilled loop (*Tchilled*) was maintained by the activity of the adsorption chillers and auxiliary heat pump (when necessary), removing the energy gained from space cooling. If *Tchilled* was ever driven too low (*Tchilled* > 11.4 ◦C), both the chillers and heat pumps were deactivated in the next time step. If the chilled loop temperature rose too high (*Tchilled* > 17.0 ◦C, representing over-loaded adsorption capacity) the auxiliary heat pump activated in the next time step.

The rejection-range loop temperature (*Trej*) was controlled by the activity of the auxiliary boiler, the auxiliary heat pump, and the heat rejection system. A target temperature *Trej,target* was set, either to *Tout,db* (if heating demands were expected to be low) or to *Trej* = 27 ◦C. If *Trej* dropped below *Trej,target*, the auxiliary heat pump activated at the next time step. If the heat pump was unavailable (due to a low chilled loop temperature), the boiler activated. If *Trej* was too high, the heat rejection system was called, proportional to the difference between *Trej* and *Trej,target*. An additional check of the current heat rejection COP (*COPrej* > 3.0) prevented operation when the temperature difference was small, to favor night flushing of excess energy in the rejection loop. These controls together maintained loop temperatures within useful bounds, providing cooling and heating to the zones, and heat rejection capacity for the adsorption chillers.

#### 2.3.8. Utility Cost Metrics

Costs for electrical use, capacity, and demand and gas use charges were initially computed within the BEM. Simulation outputs for the baseline configurations were used and scaled on an hourly basis for the Q-MEF configurations. Electricity costs were determined with monthly peak demand charges (USD 17.00/kW for the winter and USD 38.15/kW for the summer) and hourly energy charges (USD 0.125/kWh on-peak and USD 0.105/kWh off-peak), with surplus electricity generation net-metered at 100% of the current rate. The cost for natural gas was USD 1.30 per therm.

#### **3. Results**

Simulating a medium-scale commercial office building according to the Q-MEF framework demonstrated implications for lighting demands, cooling and heating loads, peak demands, net energy use, operational costs, and design considerations.

#### *3.1. Daylighting Impact on Lighting Energy Use*

Modeled lighting energy use in the high-glazing configurations improved between 13% and 17% over BL40 (Table 4). Daylighting controls were active in all configurations.


**Table 4.** Lighting loads under daylighting controls.

Because daylighting controls were modeled in the precursor BEMs, the controls did not respond to the reduction of direct irradiance due to BITCoPT. Prior daylighting and glare analyses of BITCoPT [58] determined sufficient illuminance through the same depth as the perimeter zones modeled in this study (4.57 m), and as other equipment (such as blinds) was not modeled to control for over-lit moments, it was determined that the daylighting behavior would translate from the precursor BEMs to the Q-MEF configurations.

Daylighting potentially has a great effect on the energy use profile of a building, but because it is highly contingent on occupant behavior [60] representation in the simulation method employed here is difficult. It is possible that the daylighting energy benefits of BITCoPT are over-predicted, as cloudy moments would cause under-lighting. However, experimentation has suggested that lighting through the system increases with partial clouds, as there is more diffuse light to transmit [3]. It is also possible that benefits are underpredicted due to glare, as, during brighter moments in the baseline buildings, occupants who experience excessive brightness or glare might deploy blinds and electric lighting. More differentiation between the configurations might be noted if active technologies (blinds, BITCoPT) and occupant behavior were modeled dynamically with the sensors and dimmers.

#### *3.2. Heating and Cooling Loads*

The incorporation of Q-MEF systems resulted in various responses in heating and cooling loads across the climate types, demonstrating complex interdependencies between envelope loads and building demands (Table 5). Heating loads comprise the sum of modeled baseboard heating and DOAS preheating, while the cooling loads consist only of the modeled chilled beam responses.

**Table 5.** Yearly facility heating and cooling loads of twelve configurations (loads to be addressed by building thermal control systems).


Due to the preponderance of cooling demands in the moderate MTV climate, the heating increases were low (though not negligible) relative to total demands. The NYC climate showed sensitivity to glazing area and type, as the BL95 and QMEF cases showed marked changes in the total loads. Installation of BITCoPT (the Q-MEF case vs. the BL95 case) resulted in reductions in cooling loads in all climates, although the Q-MEF + Tilt case did not decrease loads further.

The difference in loads occurred mainly in the south, east, and west perimeter zones, where decreased heat gain during direct solar conditions due to BITCoPT caused more frequent net-heating loads. In effect, the Q-MEF configurations resulted in more "skindominated" behavior of the building, where heating and cooling demands correlate to the difference between indoor and outdoor temperatures, as opposed to "core-dominated" behavior, where internal energy gains cause persistent cooling demands throughout the year (shown in the evolving Heating–Cooling ratio, in Table 5).

#### *3.3. On-Site Thermal Collection and Application*

In addition to driving chilling processes, thermal energy collected by BITCoPT was applied to building demands for zone heating and SHW preheating (Table 6). The final column in Table 6 is a sum of the zone heating loads, SHW preheating, collected thermal energy (negative sign), and DOAS heating (not described in a separate column).


**Table 6.** Summary of thermal collection and heating applications.

Via the rejection loop, collected thermal energy was applied to zone heating loads and SHW preheating (although with low demand in office-dominated buildings, the latter factor was small). Boiler (and fuel) usage was observed to increase in response to heating demands, despite net thermal collection in some cases exceeding net heating demands. This disparity indicates non-optimal behaviors in the thermal storage mechanisms, including mismatches between the times of collection and demand. This mismatch occurred largely over the annual cycle (Figure 8), indicating the usefulness of ground-source thermal storage, a function which was not implemented in the current Q-MEF model.

**Figure 8.** *Cont*.

**Figure 8.** Thermal energy transfers, year-long cumulative summary.

#### *3.4. Solar Cooling with Adsorption Chilling*

Solar thermal energy, collected by BITCoPT, was applied to drive adsorption chillers (Table 7). This cooling was additive to any passive reductions in direct solar gains from the system, with the added benefit of dispatchability since the capacity was stored in the thermally massive, chilled distribution loop.

Cooling loads in the Q-MEF configurations were lower than in the BL95 configuration. The work done by adsorption chillers further reduced the cooling required from the auxiliary heat pump. The magnitude of collected thermal energy (Table 6) relative to loads suggests that more systems (such as night-flush controls and ground-source heat exchange) would be useful to perform more controlled storage and release of heat over both diurnal and annual cycles.

Cooling power produced from solar energy was not sufficient to provide 100% of modeled demands. Chilling processes capable of higher COPs might close that gap. This requires higher operating temperatures, and therefore higher solar concentration ratios, to boost exergy collection and offset losses from the thermal collection stage. A higher rejection temperature would also then be allowable, reducing the gap between that target and the target for zone heating (the two services being provided by the common loop). This tradeoff was not explored here, but it is noted that the required insulation of the hydronic system is complex, and at higher cell operating temperatures (roughly 100 ◦C and above) radiation from the cell becomes a significant thermal loss factor.


**Table 7.** Solar cooling systems summary, showing reduced cooling loads in Q-MEF configurations with adsorption chillers offsetting fraction of remainder.

#### *3.5. Energy Use Profile Comparison*

Broadly, Q-MEF simulations improved over the baselines for all observed metrics net electrical demand, net energy demand, and peak electrical demand (defined as the maximum observed electrical demands between noon and 5 pm during the summer season). There were conditional exceptions: BL40 demonstrated the lowest peak demand of the NYC models, and the demand reductions of QMEF over BL40 were trivial. In both these cases, reductions in electrical EUI were still significant. Results are summarized in Table 8.

**Table 8.** Summary: net energy use intensities, by demand type (kWh-Q/m2-yr, kWh-E/m2-yr and kWh/m2-yr), and peak electrical demand from grid (kW-E).


The Q-MEF + Tilt configuration demonstrated the highest generation and lowest total site EUI across climates. It is notable that electricity EUI reaches net zero in the Phoenix and Mountain View climates, although there remains significant consumption of gas for heating—higher than that of the baselines.

Maximum on-peak electrical draw (during summer-season afternoons) decreased from the BL95 to Q-MEF configurations by 3% (NYC), 32% (PHX), and 19% (MTV), while in the Phoenix climate, there was a decrease of 17% relative to BL40 as well. Peak demand reductions for the Q-MEF + Tilt configuration relative to BL95 were 6% (NYC), 28% (PHX), and 20% (MTV) (Table 9).


**Table 9.** Percent changes from baseline to Q-MEF configurations in electricity EUI, total EUI (electricity + gas), and peak electricity demand.

Although it is significant that Q-MEF + Tilt demonstrated net zero electrical use, benefits were incremental (or negative) over the straight Q-MEF configuration, suggesting further analysis comparing the marginal utility of the energy benefits with the marginal costs of increasing the complexity of the building's design. It's notable that in the Phoenix condition, the Q-MEF + Tilt configuration showed higher peak demand than the straight Q-MEF configuration, differing from the other two climates. This is due to an attenuation of the cavity flush function from the combination of elevated outdoor temperatures and the increased available insolation.

#### *3.6. Utility Cost Analysis*

Annual energy costs were calculated for all configurations according to rates for electricity demand (kW), supply (kWh) and natural gas supply (therms or kWh) (Figure 9, Figure 10, Table 10). Q-MEF configurations reduced total use costs, while demand costs were generally lower than costs in BL95, and similar to costs in BL40.

**Figure 9.** Monthly utility costs, summed (electrical energy, electrical demand, gas).

**Figure 10.** Monthly utility costs breakdown (Mountain View shown) showing net-meter benefits in the summer for Q-MEF + Tilt configuration.



Observed in Table 8, Q-MEF configurations showed increased natural gas consumption, but due to lower electricity use, the total EUI was reduced in all Q-MEF cases, relative to BL40 and BL95. The analysis of costs demonstrates that (due to the low cost of natural gas compared to electricity) Q-MEF configurations showed annual reductions (Table 10).

Relative to the highly glazed baseline (BL95), peak electrical draw was reduced for each Q-MEF configuration, resulting in lower demand charges. Compared to BL40, the demand charge for Phoenix was reduced, but New York City and Mountain View had similar demand charges. On-site electrical storage was not included in this study, though if employed, it would be expected to further reduce peak demand in some months.

It is a natural goal of techno-economic analysis to reveal the cost/benefit impacts of the proposition. To determine the financial costs and benefits of Q-MEF, which is intended to address a broad range of architectural criteria, the savings in utility costs would be weighed along with the expected changes in lease rates due to changes in the thermal comfort and desirability of occupied spaces, and the installation and maintenance costs for Q-MEF systems. These costs would be considered relative to the costs of the baseline configuration's mechanical systems, or other common systems such as Variable Air Volume HVAC, which incurs additional effects on a building project's value, such as reduced inhabitable room height due to the depth of duct work. The overall cost analysis of Q-MEF is highly contextual, due to the interaction of these factors and additional localized factors, such as capitalization rates expected on monies obtained to finance a project, which is an in-part function of a perceived risk. A detailed cost analysis is therefore usefully done at the scale of individual projects, or for broader applicability, by parameterizing these factors and undertaking the resulting matrix of sensitivity analyses. That breadth of analysis is outside the bounds of this study, which, for the sake of its own broader applicability, considered generalized circumstances. The present results, however, are a precursor to such a technoeconomic analysis, which merits further investigation.

#### **4. Discussion**

The modeled application of the specific systems described in this study according to the Q-MEF strategy resulted, in simulation, in significant reductions in energy use in the three modeled climates. On-site net-zero electricity was demonstrated in two of the climates—a significant result for buildings in the modeled size class. The results further suggest that additional modifications might show additional energy-use benefits: the baseline configurations were designed according to current minimum efficiency codes but did not incorporate the full gamut of currently available high-performance building strategies, such as DOAS enthalpy recovery (important in high-humidity climates such as New York) or ground-source heat exchange (particularly useful in climates with steady cycles such as Phoenix). Pursuing the current industrial interest in full electrification might also reveal further reductions, but as the modeling process dictates systems similarity between proposed and baseline models (which employ gas heating), these impacts were not represented. The application of these and other strategies would decrease the baselines' EUI, and therefore the EUI of the Q-MEF implementations as well.

Not all high-performance design strategies are synergistic with Q-MEF. External shading devices, for example, are a passive-design strategy to reduce fenestration gains. Q-MEF attempts to internalize these gains, which increases cooling loads, particularly over the BL40 baseline, but overall reduces both site energy use (all climates) and peak demands (in Phoenix, Table 8). Although counter-intuitive from the passive design perspective, these results suggest the benefits of engaging the solar resource.

A primary benefit to coolth storage systems is peak demand reduction, which commercial systems bank on for their value proposition [16]. The demand reductions demonstrated in this study, though significant, may not fully realize the benefits of this storage. The gap may be due to the simplified controls in the Q-MEF model, which were chosen in part for compatibility with the post-processing modeling method. Optimizing controls has been shown to benefit thermal storage applications [61] and might improve the utilization of the modeled storage.

A general shift from cooling loads to heating loads was observed with Q-MEF vs. baseline configurations (Table 5), which is in keeping with the Q-MEF concept of utilizing available solar resources, rather than mitigating and compensating for them. In the Mountain View climate, the cooling-dominated baselines were flipped to more heating operation. This flexibility around the balanced point suggests that introducing ground-source heat exchange, which benefits from that neutrality, might complement the Q-MEF strategy. Ground-sourcing would be worth investigating in the New York climate as well, which shifted from balanced to heating-dominated, as such heat pump-enabled strategies are intrinsically more effective at heating than cooling. Ground-source systems likely could not "keep up" with the required quantity of heat rejection in the more extreme Phoenix climate, suggesting other strategies might be useful, such as radiative night-sky sinking, which has recently advanced through material investigations [62].

In this study, the thermal collection efficiency of BITCoPT averaged lower than in precedent studies. This stems from conflicting demands on the envelope cavity: sensitivity analysis performed through multiple simulation runs showed that allowing elevated cavity temperatures reduced transport losses in BITCoPT, but increased perimeter zone cooling demands and overall net energy use. This trade-off reinforces how a building with Q-MEF is a coupled system of components that experience unique forcing functions. Optimizing overall objectives in such a system requires subordinating the peak performance of specific sub-systems.

#### **5. Conclusions**

To test the integration of technologies through an architectural, whole-system approach to design, the production of benefits from on-site solar resources was explored through simulations with a parametric group of building energy models. The models were assembled from precursors: a set of building energy models, a model of an active envelope technology, and representations of adsorption chillers, hydronic thermal distribution and storage elements, water-to-water heat pumps, and ancillary systems. The overall strategy was described as a quality-matched energy flow (Q-MEF) network. Simulations of a 5000 m2, three-floor office building demonstrated reductions in electrical use over 70% from the baseline in a humid-continental climate (New York City), and on-site net zero electricity in arid subtropical (Phoenix) and semi-arid Mediterranean (Mountain View) climates. Peak purchased electrical demands decreased up to 6%, 32%, and 20% respectively. The magnitude of these results suggests the usefulness of deeply integrating the multifunctional envelope technology with the balance of a building's systems that process and distribute collected thermal energy. Demand for (purchased) electricity remained significant, indicating the potential for future investigation of on-site electrical storage, in addition to the modeled thermal storage. Cost outcomes were reported, with reductions (in summed energy and demand charges) of 35% in New York City, 64% in Phoenix, and 66% in Mountain View, relative to the highly glazed baseline model configurations. A shift was noted, from demand for cooling in the baseline configurations to demand for heating in the Q-MEF configurations. This shift indicates potential benefits from additional thermal storage technologies beyond those modeled in this study (such as ground-source heat exchange), and, overall, the possibility of energy-positive performance for this common class of buildings, in a range of climates.

**Author Contributions:** Conceptualization, N.N., J.O.V., M.J. and A.D.; Data curation, N.N., J.S.S., M.A.E. and K.P.; Formal analysis, N.N., J.S.S. and M.J.; Funding acquisition, N.N., J.O.V., M.J. and A.D.; Investigation, N.N., J.S.S., M.A.E. and K.P.; Methodology, N.N., K.P., J.O.V., M.J. and A.D.; Project administration, A.D.; Resources, N.N. and A.D.; Software, N.N., J.S.S., M.A.E. and K.P.; Supervision, A.D.; Validation, N.N., J.S.S., M.A.E. and K.P.; Visualization, N.N., M.A.E. and A.D.; Writing—original draft, N.N., M.A.E. and A.D.; Writing—review and editing, N.N., J.O.V., M.J. and A.D. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the U.S. Department of Energy, grant numbers DE-FG36- 06GO86070 and 09EE0002285, the New York State Energy Research and Development Authority (NYSERDA), grant numbers A50417 and J50367, and the New York State Office of Science, Technology and Academic Research.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data avaialable upon request.

**Acknowledgments:** These findings are an expansion of work presented at the 16th Conference on Sustainable Development of Energy, Water, and Environment Systems (SDEWES) held (on-site and virtually) in Dubrovnik, Croatia, 10–15 October 2021. Material support was provided by SOM, LLC.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **Nomenclature**



#### **References**

