*x* = *loadGR*1, *loadGR*2, *loadGR*<sup>3</sup>

where the problem control variables (*x*) are the three ICE loads (*loadGR*1, *loadGR*2, *loadGR*3). The resulting optimization problem could be treated as a constrained, multivariable function, minimization problem.

The problem was solved considering the heat demand profile (Figure 2) and the ICE and boiler performance maps (Figure 3). The boiler efficiency trend describes the part-load behavior of a boiler conceived for the auxiliary purpose [18], thus characterized by high performance, even at very low loads. In particular, Figure 3 shows the ICE exhaust gas mass flow rate and temperature, the electrical and thermal power output versus load (data provided by the manufacturer). The thermal power was calculated, assuming cool exhaust gases down to 110 ◦C. It resulted that each ICE could be regulated between the 40 and the 100% of its load, covering a heat demand ranging between 3200 and 4600 kW; thus, the three ICEs together could cover a maximum heat demand equal to 13,800 kW.

**Figure 3.** Performance of internal combustion engines (ICEs) and back-up boilers at part-load conditions: (**a**) ICE exhaust gas conditions at part-load operation; (**b**) ICE electric and thermal power output at part-load operation; (**c**) ICE electric efficiency at part-load operation; (**d**) Boilers thermal efficiency at part-load operation.

Results of optimal load allocation procedure are reported in Figure 4, where the load of the three ICE units (Figure 4a), the CHP unit total electric power output (Figure 4b), and boilers load (Figure 4c) as a function of the heat demand are shown. Finally, in Figure 4d, the total natural gas consumption (due to ICE units and boilers) was plotted versus the heat demand. It could be observed that, in the range of power where the ICEs operated, only the least number of units necessary to cover the demand was activated. They were turned on at their minimum load, and then the load increased with the heat demand until reaching 100%. Beyond 13,800 kW, all the ICEs worked at full-load, but the generated power was not enough to cover the heat demand; therefore, also the back-up boilers were activated. The boilers were activated also at heat demand lower than 13,800 kW when the ICEs could not fulfill the heat demand due to the regulation limits expressed in (1).

The total amount of natural gas consumed to fulfill the DH yearly heat demand resulted equal to 138.36 GWh/year; it guaranteed to cover the heat demand and to deliver 54.22 GWh/year of electric power to the grid. The ICE and the boiler fuel consumption were evaluated on the basis of the efficiency maps reported in Figure 3c,d.

**Figure 4.** Combined heat and power (CHP) system output loads and fuel consumption as a function of the district heating (DH) heat demand: (**a**) load of the ICE units; (**b**) CHP unit total electric power output; (**c**) boilers load as a function of the heat demand; (**d**) total natural gas consumption.

#### **3. Integrating the Organic Rankine Cycle**

In order to increase the amount of electrical energy sold to the grid, the opportunity of integrating an ORC as a bottoming cycle of the ICEs was considered. Since ORC technology offers the possibility to generate electricity from low-grade heat sources, the residual heat of ICEs exhausts flue gases was here considered as a potential feeding source. Thus, the CHP plant optimal operation was compared with a new arrangement where ICEs were operated at full-load, and an ORC was integrated into the layout to recover ICEs' residual heat. Three different ICE-ORC layouts were proposed and analyzed. For each investigated arrangement, optimum ORC design (i.e., selection of organic fluid and key cycle parameters) was identified; based on ORC system off-design modeling, the amount of additional electrical energy producible during the year was then quantified.

#### *3.1. ORC Architecture*

A simple sub-critical and regenerative ORC architecture with an intermediate heat transfer fluid (IHTF) loop was assumed, as schematically shown in Figure 5. The key cycle components were: (i) the heat exchanger between ICE exhaust gases and heat carrier fluid, (ii) the evaporator (exchanging heat between the heat carrier fluid and organic fluid), (iii) the expander, (iv) the regenerative heat exchanger, and (v) the condenser and pumps. The use of an intermediate thermal circuit, which could act as buffer, was crucial for the application under consideration in order to: (i) avoid the direct contact between

exhaust gases and flammable organic fluid (such as hydrocarbon), (ii) smooth out the variation of temperature and mass flow rate of the heat source.

**Figure 5.** Schematic of the organic Rankine cycle (ORC) architecture.

#### *3.2. Proposed Series and Parallel Arrangements*

Three different layouts were investigated in this work, as schematically shown in Figure 6.

In the first case (Case A), the ORC was placed downstream the heat exchanger feeding the DH network, thus recovering the residual heat of exhausted gases. Therefore, the DH heat demand would determine the temperature of the gas feeding the ORC.

In the second architecture (Case B), the heat recovery heat exchanger position was reversed, with the ORC located upstream of the DH heat exchanger. Accordingly, in this second arrangement, the ORC would be fed by a constant temperature, equal to 400 ◦C (i.e., the temperature of ICE exhausted at full-load condition, see Figure 3a), while the total amount of heat input to the cycle would vary, according to the DH request.

In the third analyzed layout (Case C), a parallel arrangement was assumed: exhaust gas flow would be split between the two branches in order to primarily satisfy the DH load demand, while both branches featured the same inlet gas temperature value (equal to 400 ◦C). A collector was placed downstream components, in order to mix up the two different fluxes; the minimum requirement of 110 ◦C on temperature was maintained in section G3 of Figure 6.

Indeed, for all the above described integrated system architectures, the heat demand of DH was the primary requirement to meet. Consequently, the thermal power input to the ORC equaled the difference between the available thermal power in the ICE exhaust gas and heat delivered to the DH network ( . *Qth*,*DH*). For each investigated layout, Table 1 collects the assumed constant and variable parameters (see the section in Figure 6). The intermediate heat transfer fluid heat exchanger, which provided the thermal power input to the ORC, was indicated as IHTF HE for the sake of brevity.

**Figure 6.** Schematic of the proposed DH and intermediate heat transfer fluid (IHTF) heat exchangers arrangements: (**a**) case A; (**b**) case B; (**c**) case C.

**Table 1.** Constants and variable parameters for each proposed layout.


#### *3.3. Selected Organic Working Fluids*

Four different hydrocarbons (Cyclopentane, Benzene, Cyclohexane, and Toluene) were selected as potential working fluids, as demonstrated to be performing fluids for recovering heat at the temperature imposed by the ICEs exhausts [19,20]. Indeed, the selected fluids exhibited a critical temperature quite similar or slightly higher than the target evaporation temperature, in particular high enough to achieve a good thermal matching between fluids and exhaust gas, but not too high to lead to excessively low vapor densities and thus high system cost [21]. Refrigerants were excluded due to their low critical temperature, while siloxanes were excluded because their use would lead to excessively low saturation pressure values, assuming cooling fluid temperature close to ambient conditions [21]. The main properties of selected fluids are summarized in Tables 2 and 3, and the fluids saturation pressure values are plotted versus temperature for comparative purpose in Figure 7.

The IHTF considered was Therminol 62, one of the most popular high-temperature liquid phase heat transfer fluid. Therminol 62 was selected as it offers outstanding performance to 325 ◦C, including excellent thermal stability and low vapor pressure. These properties result in reliable, consistent performance of heat transfer systems over long periods of time (main properties are reported in Table 3) [22].

**Table 2.** Main properties of selected organic fluids [21].



**Table 3.** Main properties of selected IHTF [22].

**Figure 7.** Pressure-temperature saturation curves of selected hydrocarbons.

#### *3.4. ORC Thermodynamic Modeling and Design Assumptions*

The ORC system and the IHTF circuit were modeled and simulated by means of the commercial software *Thermoflex*™ [15]. The software, integrated with the FluidProp library for organic fluids thermodynamic properties evaluation, allowed simulating the energy system steady state, based on a lumped parameters approach.

First of all, the whole system layout was reproduced in the *Thermoflex*™ environment, combining the built-in library single component modules of heat exchangers, turbines, pumps, sources, and sinks. Boundary conditions and components design key parameter values used in the study (such as turbine isentropic efficiency, condensation pressure, pressure drops, and heat losses along the circuit) are listed in Tables 4–6, mainly selected in line with the state-of-the-art of conventional ORC technology [23].


**Table 4.** ORC layouts dependent assumptions.


**Table 5.** ORC general assumptions.

**Table 6.** ORC condensation pressure assumptions.


In order to perform a realistic evaluation of the ORC performance, both thermodynamic design and off-design analyses were performed. Off-design performance of the plants were obtained passing through three main steps: (i) preliminary thermodynamic heat and mass balances; (ii) engineering design phase, in which design parameters of the power plant are defined (i.e., size of components, geometric details, etc.); (iii) thus, off-design analysis where, being defined the design characteristics and fixed the geometry of the components, depending on selected control logic, it is possible to predict both components and overall system behavior under different operating conditions.

The main equations used to calculate streams' design parameters included energy balances at the single components. A more exhaustive description of the Thermoflex modeling approach can be found in [23]. Concerning the off-design modeling approach, the heat exchangers off-design behavior was calculated according to the method of thermal resistance scaling [15]. Being determined as the design-point overall thermal resistance of the heat exchanger, the two fluid-side resistances were scaled at off-design, using the single-parameter scaling method [15]. Normalized heat loss in each heat exchanger was considered, and it was entered as a percentage, relative to the heat transferred out of the higher temperature fluid. Flow resistance coefficients, initialized in the design-point stage, were used to model the pressure drops across each side of the heat exchanger in off-design. Regarding the ORC expander, the sliding pressure part-load control was assumed to model its behavior at off-design conditions. Thus, whenever possible (based on topper off-design performance), the organic fluid temperature at the expander inlet was kept constant at its rated design value, while mass flow and pressure varied proportionally assuming chocking conditions (constant mass flow function) at the turbine inlet. Expander isentropic efficiency at off-design was corrected, starting from the design point value, based on flow function.

Table 4 shows the selected design values for the inlet/outlet temperature and mass flow of the gas at the IHTF HE, for the different proposed layouts; these values corresponded to a DH demand equal to 3 MW, the mean value of the thermal demand in the period of minimum request (May to September). The IHTF circuit design temperature at the HE outlet (Tdes) was set equal to 300 ◦C in Case B and Case C layouts, while three possible values (Case A1: 200 ◦C, Case A2: 250 ◦C, and Case A3: 300 ◦C) were considered in Case A layout, where the IHTF HE was downstream the DH HE.

The condensation pressure depends on the working fluid, and it is bound by the cold source temperature. In this study, the cooling medium temperature was assumed equal to 25 ◦C as representative of a medium climate condition [24]. Considering a reasonable temperature difference between organic fluid saturation temperature and cooling medium temperature inlet equal to 15 ◦C, the organic fluid condensation temperature was set equal to 40 ◦C for all considered fluids, resulting in the condensation pressure values in Table 6 (see also Figure 7). In order to guarantee a complete fluid condensation inside the condenser, a subcooling equal to 5 ◦C was assumed.

#### **4. Results and Discussion**

#### *4.1. Optimization of the Cycle Parameter and Working Fluid Selection*

A preliminary thermodynamic sensitivity analysis was performed on the considered working fluids, in order to optimize the cycle evaporation pressure and then determine the fluid that maximizes the net electric power production, for each analyzed arrangement.

The influence of organic fluid and key cycle operating parameters are presented in Figure 8, with reference to Case A layout. Since the exhaust gas entering the IHTF HE was variable in case A (see Table 1), a comparative analysis of the net power output is presented in Figure 8 for each analyzed *Tdes* value and for each organic fluid. Results showed that Cyclopentane achieved the best performance compared to other hydrocarbons, for all the considered *Tdes* values. However, it must be pointed out that ORC performance was reduced by decreasing the IHTF temperature. Optimum evaporation pressure values could be identified equal to 40, 16, and 10 bar, respectively, at 300, 250, and 200 ◦C. As expected, the highest ORC performance was achieved for the highest temperature (i.e., 1490 kW of net power output for Cyclopentane with a pressure equal to 40 bar and *Tdes* equal to 300 ◦C). If the design IHTF temperature was reduced to 200 ◦C, the maximum ORC power output was reduced by nearly 20%.

**Figure 8.** Sensitivity analysis results on organic fluid and IHTF design temperature in Case A layout and different IHFT temperature cases.

Figure 9 shows the results for Case B and Case C: a fixed IHTF design temperature value equal to 300 ◦C was considered. Indeed, these proposed layouts featured the maximum temperature of exhaust gas entering the IHTF HE (TG1 and TG1", respectively, see Table 1). Cyclopentane still provided the best performance in Case B and Case C layouts. The maximum power output was achieved with evaporation pressure equal to 40 bar for both layout arrangements. The highest ORC power output value was obtained in Case B, in which the IHTF HE was fed with the exhaust gas at its highest temperature and mass flow if compared with the other cases.

**Figure 9.** Sensitivity analysis results on organic fluid and IHTF design temperature: Case B and Case C layout.

#### *4.2. O*ff*-Design Analysis*

Based on sensitivity analysis results and design assumptions of Tables 4–6, off-design evaluation of proposed layouts was carried out, considering Cyclopentane as working fluid in all the cases. The evaporation pressure in design conditions was set equal to the optimum values, based on the design thermodynamic results described in paragraph 4.1. The input variables for the three analyzed layouts (see Table 1) were varied in order to simulate the ORC performance under variable DH heat demand.

Figure 10 shows the ORC net power output as a function of different considered input variables, namely the IHTF HE inlet/outlet gas temperature and inlet mass flow, for the different layout arrangements. The input variables ranged between the 110% design load condition and the minimum possible operating value. In detail, Figure 10a shows the performance behavior of Case A layout as a function of the temperature of the exhaust gas feeding the IHTF HE (*TG*2, see Figure 6), for the different IHTF design temperature. The figure indicated that sizing the ORC on higher IHTF design temperature improved the power production in design conditions, but affected the performance in off-design and limited the operational range of the ORC. Indeed, in Case A1, the ORC could exploit exhaust gas temperature values down to 310 ◦C; for the Case A2, down to 260 ◦C; for the Case A3, down to 240 ◦C. Figure 10b shows the performance of Case B layout as a function of the gas outlet temperature (*TG*2, see Figure 6), while, in Figure 10c, ORC power output was plotted versus the gas mass flow rate ( . *mG*<sup>1</sup> , see Figure 6). In Case B, the exhaust gas outlet temperature ranged between 190 ◦C and 320 ◦C. In Case C, the ORC could operate with gas mass flow ranging between 12 and 36 kg/s.

Actually, an ORC could be regulated from full-load down to minimum 30% continuously with fast response, while it was able to increase its power output, up to 110% of its rated power, for a limited amount of hour. Thus, in this analysis, it was assumed that the ORC could work during the year even in the moments when the heat demand was lower than 3 MW, when the IHTF HE was fed with a thermal power higher than the design one until the ORC load did not exceed the 110%.

**Figure 10.** ORC net electric power at part-load conditions: (**a**) effect of gas temperature at IHTF HE inlet, for layout A and different *Tdes* values; (**b**) effect of gas temperature at IHTF HE outlet, for layout B; (**c**) effect of gas mass flow.

#### *4.3. Energy Evaluation*

According to the DH yearly heat demand profile of Figure 2 and to the ORC performance trends of Figure 10, main comparative energy results were quantified and summarized in Table 7, showing the yearly fuel consumption and ICE generated energy, the ORC design and peak power output, the ORC average power during operating hours, the ORC yearly generated electrical energy, the ORC + ICEs yearly generated electrical energy, and additional generated energy, for each analyzed case study. The number of operating hours and generated electric power output for each investigated layout are also shown and compared in Figure 11, in the form of power outputs monotonic profiles. The ORC performance referred to the net ORC electric power output obtained as gross electric power output value minus IHTF and ORC pumps consumptions (corresponding to about the 8% of the ORC gross electric power output), minus air condenser fan consumptions (corresponding to another 8% of the ORC gross electric power output alone).

In the new arrangement, the ICEs worked the whole year at full-load, producing more energy than in the original set-up, in which they often operate at part-load conditions following the heat demand. In particular, in the new arrangement, the ICEs produced 170.89 GWh/years of additional electric energy, consuming 344.59 GWh/years of fuel more than in the original set-up. The exhaust gas was recovered, furthermore, to feed the IHTF HE, and an additional amount of energy was produced by means of the ORC. Therefore, the total electric energy produced by the new arrangement depended on the considered ORC layout performance.


**Table 7.** Main energy results for the analyzed configurations.

**Figure 11.** Power output monotonic profiles during the year for each investigated layout.

Case B showed the highest ORC power peak value, but not the highest amount of producible electric energy, due to the limited number of hours in which it operates during the year (about 3126). Thus, in order to evaluate the performance of the ORC on the year basis, it was fundamental to analyze its behavior also at part-load conditions and to take into account its regulation limits. In this case study, the number of operating hours turned out to be fundamental: the layout that allowed to produce the maximum amount of ORC electric energy per year, 9.77 GWh/year, was the Case C, case where the ORC could operate for the greatest number of hours, equal to 7050 h/year. More detailed results concerning the ORC best case (Case C) are reported in Appendix A for completeness. In particular, the monthly profile of heat demand, electric power generated by the old arrangement and the new one, and ORC net electric power output are grouped in Figure A1. Daily profile of heat demand and ORC net electric power output are presented and compared in Figure A2, for two days (February 15th and July 15th), chosen as representative of winter and summer operations of the DHN.

In order to measure the benefits of the ICE-ORC CHP production, a comparison was carried out between two scenarios, in terms of primary energy consumption: in the first scenario, the old CHP system was considered, while, in the second scenario, the new CHP was considered, where an additional electric energy production (Δ*EEL*) occurred, and heat demand remained the same (see Table 7).

A relative global primary energy saving (*GPES*) could be calculated as follows:

$$GPES = \frac{F\_{old} - F\_{new}}{F\_{old}} \tag{2}$$

where *Fold* is the primary energy consumption in the old scenario, and *Fnew* is the primary energy consumption in the second one, respectively, calculated according to Equation (3) and Equation (4).

$$F\_{\text{new}} = F\_{\text{ICEnew}} + F\_{\text{BOILERnew}} \tag{3}$$

$$F\_{\text{old}} = F\_{\text{ICEold}} + F\_{\text{BOILERold}} + \frac{\Delta E\_{\text{EL}}}{\eta\_{\text{EL.GENMIX'}}\eta} \tag{4}$$

In particular, the primary energy consumption could be split into three contributions: a contribution due to the ICEs operation, *FICE*, a contribution due to the boiler's operation, *FBOILER*, and a third term due to, Δ*EEL*. This term represented the fuel saved by producing Δ*EEL* with the new arrangement in place of external electric energy production, and it was calculated considering the European electric generating mix. This latest contribution was estimated as the ratio between the additional generated electric energy from the engines and the ORC, Δ*EEL*, and the average electric efficiency of the European electric generating mix, η*EL GEN MIX*, and a coefficient, *p*, which took into account the grid losses depending on the feed-in voltage connection to the grid; in particular, *p* was intended to promote the feed-in of electricity at lower voltage.

In this analysis, η*EL GEN MIX* was assumed equal to 40%, which corresponded to the EU28 average electric efficiency of the electric generating mix in 2017 [25]. *p* was assumed equal to 0.985, which corresponded to the correction factor related to the avoided grid losses, for a CHP system connected to the grid at a voltage equal to 150 kV, as provided by the Directive on primary energy-saving calculation [26].

Results concerning the *GPES* are also reported in Table 7. The *GPES* analysis highlighted that the ICE-ORC arrangement introduced a positive primary energy saving. The *GPES*, however, differed for the different analyzed configuration; in particular, the highest value of the *GPES* was obtained for the Case C (19.1%), while the lowest value was obtained for the Case B (17.4%), for the aforementioned reasons. The inclusion of the ORC entailed positive values of *GPES* mainly because it allowed producing additional electric energy with remarkably high-efficiency values close to 52%, higher than the considered average European electric generating mix efficiency and in line with the current technology of mid-size combined cycles.

#### *4.4. Techno-Economic Feasibility of Analyzed ICE-ORC Layouts*

The ICE-ORC solution was compared to the original arrangement in terms of economic performance, by means of the differential net present value index. The differential net present value is defined in Equation (5), where Δ*Ci* is the differential cost, Δ*Ri* is the differential revenue at the i-th year, *Ma* is a factor assumed equal to 0.9 [27] that accounts for the ORC yearly maintenance costs, *q* is the discount rate assumed equal to 6%, and *IORC* is the investment on the ORC system. The differential cost was evaluated as the product between the natural gas price and the differential primary energy consumption (Equation (6)). Different gas price values were considered in this analysis: a medium value assumed equal to 33 EUR/MWh, a high value equal to 44 EUR/MWh, and a low value equal to 22 EUR/MWh. These values corresponded, respectively, to the average, the highest, and the lowest natural gas price encountered in the European countries at the beginning of 2019 [28]). The differential revenue was estimated according to Equation (7) as the product between the electric energy sell price and the differential generated energy. The investment cost of ORC was quantified according to the trend line in Figure 12, where specific investment cost was plotted as a function of the ORC size, based on product data of an ORC market leader manufacturer [23].

Equation (8) could be manipulated in order to evaluate the electric energy price that guarantees the return on investment in a given time period, i.e., the payback period (*PB*), using Equation (5).

$$
\Delta NPV = \sum\_{i=1}^{n} \frac{\Delta R\_i \cdot M\_d - \Delta C\_i}{\left(1 + q\right)^i} - I\_{ORC} \tag{5}
$$

$$
\Delta \mathbf{C}\_i = \Delta \mathbf{F} \cdot \mathbf{C}\_{FIL\to} \tag{6}
$$

$$
\Delta R\_i = \Delta E\_{EL} \cdot \mathbb{C}\_{EL} \tag{7}
$$

$$\mathcal{C}\_{EL} = \left(\frac{I\_{ORC}}{\sum\_{i=1}^{PB} \frac{1}{(1+q)^i} \cdot PB} + \Delta \mathcal{C}\_i\right) \cdot \frac{1}{\Delta E\_{EL} \cdot \mathcal{M}\_a} \tag{8}$$

**Figure 12.** ORC specific investment cost as a function of plant size [23].

Figure 13 displays the results of the economic assessment in terms of electric energy sell price to return on investment in a given payback period. In particular, Figure 13a shows a comparison between the different ICE + ORC layouts, by considering the same natural gas price (the medium value), while Figure 13b shows the influence of the natural gas price. As expected, the payback period decreased when the electric energy price increased; indeed, higher electric energy prices led to higher revenues and, thus, to lower payback periods.

**Figure 13.** Electric energy sell price to return on the investment in a given payback period: (**a**) Comparison between the different ICE + ORC configurations; (**b**) Influence of the natural gas price value.

The trends were very similar for the different layouts, only Case B deviated slightly from the others. From the economic point of view, the Case B was the worst one because it presented the highest investment (due to the highest size) and the lowest yearly earning (due to the lowest yearly energy production). In general, the ORC solution proved to be a good investment since it allowed returning on the investment in barely 5 years, by selling the electric energy at a price of about 70 EUR/MWh, considering a medium natural gas price (33 EUR/MWh). When the higher natural gas price was considered, instead, the electric energy sell price could be lower in order to return on the investment in the same payback period; when lower natural gas price was considered, on the contrary, the electric energy sell price must increase. Especially for the Case C, in order to return on the investment in 5 years, the electric energy should be sold at a price of 95 EUR/MWh, considering the high gas price, or at a price of even 50 EUR/MWh, considering the low gas price.
