1. Introduction
Global demand for electricity has been increasing and likely will continue to increase, especially as new and existing technologies are adopted globally. The problem is not the consumption of electricity per se, but rather the environmental impact associated with producing electricity, such as particulate, heavy metal, acid gas, and greenhouse gas emissions. For example, increased greenhouse gas emissions have changed the Earth’s natural atmospheric concentrations. With rising global temperatures and sea levels due to rising greenhouse gas concentrations in the atmosphere, there is a crucial need for eliminating greenhouse gas emissions from fossil-fueled combustion in order prevent future environmental and economic damage [
1]. In August of 2015, the U.S. Environmental Protection Agency (EPA) finalized the “Clean Power Plan”, which aims to cut carbon emissions thirty percent from what they were in 2005 by 2030 [
2]. This plan determines emission standards for pre-existing power plants [
3] and sets a limit of 636 kg (gross) of CO
2 emissions per MWh from new coal-based power plants, averaged over the lifetime of the power plant [
4].
In the United States, coal power plants are major generators of both electricity and greenhouse gas emissions, representing both 39% of total power produced and 76% of total CO
2 emissions from electricity generation nationwide [
5,
6]. Coal power plants are typically one of the two following processes: pulverized coal (PC) or integrated gasification combined cycle (IGCC) power plants. In the first type, the coal is pulverized and then combusted in a boiler to produce steam that is used to generate electricity in a Rankine-only cycle. In an IGCC power plant, however, the coal is first gasified to produce a H
2-and-CO-rich synthetic fuel gas, called syngas, and then this syngas is combusted in a gas turbine to generated electricity. Heat taken from both the syngas and the flue gas is also recovered to produce steam for additional electricity generation. This process is called a Brayton-Rankine (B-R) cycle or sometimes simply called a combined cycle (CC). IGCC power plants are typically associated with higher efficiencies than PC power plants, in part because the efficiency of combusting fuel in combined B-R cycles is higher than in Rankine-only cycles [
7]. When equipped with carbon capture and storage (CCS) processes, a PC power plant is expected to obtain a higher heating value (HHV) efficiency of around 26%, whereas an IGCC power plant is expected to obtain a HHV efficiency of 32% or higher with technologies currently in development [
8]. An advantage of removing CO
2 from syngas over flue gas is that the partial pressures of CO
2 in a pre-combustion fuel stream is roughly 200× higher than in a post-combustion flue gas stream. A related advantage of pre-combustion CO
2 removal is the greater physical contrast between CO
2 and H
2 molecules than CO
2 and N
2 molecules. The large partial pressure of CO
2 and the large contrast between CO
2 and H
2 allows one to use a physical solvent, such as Selexol™ [
9], which has good H
2S/CO
2 selectivity, very good CO
2/H
2 selectivity, and good kinetics for CO
2 absorption/release. Because of the potential for higher net electrical efficiency for a new build, we have chosen here to analyze IGCC-CCS rather than PC-CCS power plants. Though, we recognize that there are many more PC power plants in existence globally, and we recognize that there are good reasons to study CO
2 capture at both PC and IGCC power plants if we are to meaningfully reduce global CO
2 emissions.
Numerous previous authors have analyzed IGCC-CCS power plants from an energy, an exergy and/or an economic perspective. Some particularly useful reports on this subject are the U.S. DOE’s National Energy Technology Laboratory (NETL) Bituminous Baseline Reports, which give a detailed breakdown of the electrical efficiency and levelized cost of electricity of PC, PC-CCS, IGCC, and IGCC-CCS power plants using similar assumptions in all cases [
10]. In the latest NETL Bituminous Baseline Report with IGCC & IGCC-CCS configurations [
10], the GE-gasifier based IGCC power plant is reported to have an HHV electrical efficiency of 39.0% and the GE-gasifier based IGCC-CCS power plant is reported to have an HHV electrical efficiency of 32.6%. Another useful source for a detailed breakdown of the electricity generation and economics of PC, PC-CCS, IGCC, and IGCC-CCS power plants is the Integrated Environmental Control Model (IECM) by Dr. Edward Rubin’s research group at Carnegie Mellon University [
11]. In IECM v9.1, the GE-gasifier based IGCC power plant has an HHV electrical efficiency of 36.4% and the GE-gasifier based IGCC-CCS power plant has an HHV electrical efficiency of 30.8% [
12]. One general economic conclusion regarding IGCC-CCS power plants from both NETL reports and IECM is that the cost of the CO
2 capture and compression equipment is quite small compared with the other capital costs at the plant, such as the gasifier and the power cycle equipment. This means that the main economic effect of adding CO
2 capture and compression equipment to the IGCC power plant is the decrease in net overall power production.
As discussed previously, there is a net power decrease of approximately 6 basis points when adding CO
2 capture and compression. Most of this net power reduction is due to the fairly high exergy in the CO
2 stream leaving the power plant in the CO
2 storage pipeline (15 MPa), whose pressure is significantly higher than both the partial pressure of the CO
2 in the environment (40 Pa) and the partial pressure of CO
2 in the flue gas stream from a coal power plant (15 kPa.) For example, our results later in this paper find that the exergy in the supercritical CO
2 leaving in the CO
2 pipeline is approximately 5% of the exergy in the inlet coal. The remaining power reduction when adding CCS is due to irreversibility within the subsystems that must be added to the power plant, such as the Water-Gas-Shift subsystem and CO
2 capture/compression subsystems. For example, Li et al. [
13] conducted an exergy analysis of an IGCC-CCS power plant and found that exergy destruction in the WGS reactor can be quite large when CO conversion rate above 90% are required.
Given this net power reduction, there have been attempts to improve IGCC-CCS power plants so that the net power from an IGCC-CCS power plant is closer to that of an IGCC power plant without CO
2 capture. For example, to improve both net power and overall economics, Franz et al. [
14] analyzed the use of porous ceramic membranes and a water-gas shift membrane reactor in comparison to a baseline model using the Selexol™ process. The IGCC baseline in Franz et al. [
14] has a net power of 47.4% without CCS. Their Selexol-only IGCC-CCS baseline had an efficiency of 37.1%. Their IGCC configuration with a H
2-selective ceramic membrane after the WGS subsystem had an efficiency of 37.1%. Finally, their IGCC configuration with a H
2-selective ceramic membrane within the WGS subsystem had an efficiency of 41.6%. While the efficiencies of all of their configurations are higher than those of equivalent NETL & IECM models, their relative conclusions are likely still valid because it appears that the reason for the slightly higher starting efficiencies than in NETL and IECM models of an IGCC power plant is due to slightly higher assumptions for the efficiency of heat exchangers, pumps and turbines in the Rankine cycle.
In a similar attempt to improve the efficiency of an IGCC-CCS power plant, Gray et al. [
15] presented research on the effect of advanced technologies on both the electrical efficiency and the overall economics of IGCC-CCS configurations. They found that, through a combination of various advanced technologies, the electrical efficiency of an IGCC-CCS power plant could be increased from approximately 31% (reference case) to approximately 40% and that the total plant cost (in $/kW) could be decreased by roughly 40% compared with their reference IGCC-CCS case. The bulk of the gain in efficiency and decrease in total capital cost was from using (a) a high temperature H
2-selective membrane and (b) an advanced turbine designed to operate on H
2 rather than CH
4.
Merkel et al. [
16] also studied the use of membranes in CCS. They stated that, while polymeric membranes have a lower hydrogen permeance and selectivity compared to palladium-based membranes, the advantages of polymeric membranes are that they are stable, inexpensive, and easily scaled-up. Krishnan et al. [
17] investigated high-temperature polybenzimidazole (PBI) membranes to be used in CCS. They stated that PBI membranes possess “excellent chemical resistance, a very high glass transition temperature (450 °C), good mechanical properties and excellent material processing ability” [
17]. Both sets of researchers listed above found that IGCC-CCS configurations with polymeric membranes increased the HHV electrical efficiency compared to their Selexol™ baseline IGCC-CCS cases. The general conclusion from all of the above papers is that H
2-selective membranes can increase the net electrical output and decrease the normalized capital costs (in $/kW.)
Our goal here is not to conduct an economic analysis with a cost comparison between different CO
2 capture technologies. Instead, our main focus here is conducting detailed exergy analyses of three cases in order to help build an understanding of where are the major sources of exergy destruction in the three cases. Our goal here is to present detailed, thorough, and transparent exergy analyses, which builds upon the previous work by authors who have published exergy analyses of IGCC and IGCC-CCS power plants [
8,
13,
18,
19,
20,
21,
22,
23,
24,
25]. As such, this paper presents comprehensive exergy analyses on three IGCC-CCS power plants: (a) a baseline model using Selexol™ for H
2S & CO
2 capture; (b) a modified version that incorporates a H
2-selective PBI-membrane operating at 250 °C; and (c) a modified version that incorporates a CO
2 selective membrane. The exergy analyses give crucial insight into understanding where useful work is being wasted and destroyed in each of the subsystems within an IGCC power plant, as well as insight into the exergy remaining in all of the streams exiting the power plant. The exergy analyses presented here are similar to the exergy analyses of IGCC processes conducted by Siefert et al. [
8,
19]. One difference compared with previous analyses is that we present a detailed exergy analysis of both the baseline and the modified cases. In previous work, we presented the exergy analyses of the modified cases without presenting a full exergy analysis of the baseline model [
8,
19]. In future work, we will be conducting a full techno-economic analysis of the three cases presented here [
26].
2. Exergy Methodology
One way to evaluate the efficiency of a process is to analyze the exergy destruction that occurs throughout the whole process. Exergy is defined as the maximum amount of useful work that can be obtained from a material by bringing it into equilibrium with its environment. This makes exergy a property of both the material and its surroundings. Here, the reference environment was taken to be the Earth’s atmosphere at standard temperature and pressure, 298 K and 1 atm, and with the chemical composition listed in
Table 1. The further away the material is from equilibrium with the reference environment, the more exergy is associated with the material. By this definition, exergy can only be positive, unless the material is in complete thermal, mechanical, and chemical equilibrium with the reference environment. Since exergy is equal to the theoretical maximum amount of useful work, tracking the exergy destruction is a good indication for which parts or processes within the power plants have the greatest potential to be improved so as to increase net power production. As seen in
Table 1, we have assumed for simplicity that the activities of the liquid and solid species are equal to 1, even though these species do form a mixture in the environment and hence have activities less than 1. We made this assumption for liquid water because different power plants will have reference water environments with different levels of salinity; and we made this assumption for solids because the largest error in the calculation of the exergy in the coal ash and coal slag is the uncertainty in the exact crystal state of the solids entering and exiting the gasifier. This error is much larger than the error in assuming that the activity of each solid species is equal to one.
2.1. Exergy Destruction within Each Subsystem
A high-level exergy analysis of an IGCC-CCS power plant was conducted by Kunze et al. [
18], in which the overall plant was separated into 3 different subsystems (Gasifier Island, Syngas Treatment, and Combined Cycle.) Here, we break the IGCC-CCS power plant into a total of 10 (or 11) subsystems within the power plant. These 10 (or 11) subsystems can be seen in
Figure 1,
Figure 2 and
Figure 3, each of which will be discussed in more detail in
Section 3.1,
Section 3.2 and
Section 3.3, respectively. All subsystems were assumed to be operating at steady-state conditions. We determined the exergy destruction within each subsystem by taking into account the inlet and outlet streams, the heat entering and leaving, as well as the work being consumed and produced. Since Aspen Plus does not directly calculate the thermal energy sent to the environment, this term was calculated using the First Law of Thermodynamics, as shown in Equation (1).
Equation (1): First Law of Thermodynamics for a steady-state control volume solving for thermal energy transfer to the environment.
where
is the thermal energy sent to the environment,
stands for the molar flowrate of streams,
stands for the molar enthalpy of streams,
is the electrical work leaving the system, and
stands for the thermal energy transfer out of (+) or into (−) the subsystem from another subsystem. Diagrams of all of the subsystems for the three models analyzed can be found in the
Supplemental Materials, along with a summary of the nomenclature used in this text. The irreversible production of entropy was then calculated using the Second Law of Thermodynamics, taking into account that thermal energy leaves a subsystem at varying temperatures, as shown in Equation (2).
Equation (2): Second Law of Thermodynamics for a steady-state control volume with thermal energy leaving at a range of temperatures.
where
represents the irreversible entropy production,
stands for entropy and
T stands for the temperature of the heat. As shown in the last two terms in Equation (2), heat was separated into two categories: (1) heat leaving to the environment and (2) heat being transferred between two subsystems. We did this in order to differentiate between heat used elsewhere in the plant and heat exiting to the environment, the latter of which was assumed to have zero exergy because it will eventually come into equilibrium with the environment. Heat being transferred between subsystems, however, has a certain amount of exergy associated with it. Equation (3) describes how to calculate the exergy associated with heat transferred between sub-systems.
Equation (3): Exergy in thermal energy transferred between subsystems.
where
is the exergy associated with the heat transfer,
is the temperature of the inlet of a stream that gives/receives thermal energy to/from another subsystem, and
is the temperature of the outlet of such a stream. It should be noted that Equation (3) as well as the last term in Equation (2) are only valid when the specific heat capacity of the fluid does not significantly vary along the length of the heat exchanger.
When heat is transferred between two subsystems, the Steam subsystem was always involved. The exergy destruction associated with heat transfer was calculated two different ways: (a) always being assigned to the Steam subsystem and (b) being assigned to the respective subsystem transferring thermal energy to/from the Steam subsystem. In case (b), and are the temperatures of the inlet and outlet of the streams on the Steam side of the HX. Whereas in case (a), and are the temperatures of the inlet and outlet of the streams on the non-Steam side of the HX.
Once the irreversible entropy production within a subsystem is calculated using Equation (2), the exergy destruction within the subsystem can be calculated using the Gouy-Stodola Theorem, which is shown in Equation (4).
Equation (4): Guoy-Stodola Theorem. This theorem states that the rate of exergy destroyed within a subsystem, , is equivalent to the product of the temperature of the environment and the entropy generation rate due to irreversible processes, , within the subsystem.
2.2. Exergy Entering and Exiting Subsystems
An exergy analysis is more than just a second law analysis. Using only Equations (2) and (4), one cannot conduct a full exergy analysis because these equations do not calculate the exergy of streams entering/exiting the entire power plant. To calculate the exergy of a stream, an Excel workbook was used to carry through a series of calculations. Equation (5) is the equation used to calculate the molar exergy of a material stream entering/exiting a subsystem, assuming the gravitational potential and kinetic exergy are negligible.
Equation (5): Molar exergy of a material stream entering/exiting a subsystem. Here, and represent the molar enthalpy and molar entropy, respectively, of the stream at the temperature and pressure at which they enter/exit a subsystem while the and represent the molar enthalpy and molar entropy of the stream if the elements in the stream were at the same temperature and mechanical pressure of the reference environment. is the molar chemical exergy, which is related to the Gibbs free energy of a species when brought into full chemical equilibrium with the chemically-stable species in the environment, i.e., N2, O2, H2O, CO2, CaCO3, CaSO4·2H2O, SiO2, Al2O3, and Fe2O3.
As shown in Equation (5), the exergy of a stream has both a thermo-mechanical and a chemical exergy component. The specific chemical exergy at STP of each species used in the analysis is shown in
Table 2, which is similar to the values derived by Szargut, Morris and Stewart [
27], but derived independently using standard values of the Gibbs free energy of formation and the composition of the environment as listed in
Table 1. It should be noted that most of these values for chemical exergy can also be found in Table A-26 of Moran and Shapiro’s 6th edition of Fundamentals of Engineering Thermodynamics [
28]. When a mixture is being analyzed, the chemical exergy can be calculated through a weighted average of the chemical exergy of the individual components based on their respective mole fractions. This is due to the fact that gases in the Earth’s atmosphere can be treated as an ideal gas mixture. For simplicity and because the actual crystal state of these species in the coal ash/slag are unknown, we assume that the following species have zero chemical exergy: SiO
2, Al
2O
3, CaO, MgO, Na
2O, K
2O, and Fe
2O
3. As seen in
Table 2, we did assume that FeO and Fe have chemical exergy, as calculated from the Gibbs free energy of the reactions that create Fe
2O
3 from FeO and Fe using oxygen at the partial pressure of the environment. While the crystal state of the SiO
2, Al
2O
3, CaO, MgO, Na
2O, and K
2O in the coal ash does change between the inlet and outlet of the gasifier [
29], the effect of this change on exergy calculations is small and was ignored because, unlike Fe
2O
3, no redox reaction occurs for these species.
In addition to using the Gouy-Stodola Theorem to calculate the exergy destruction within a subsystem, one can directly calculate the exergy destruction within a subsystem by calculating the difference between the exergy that enters a subsystem and the exergy that leaves a subsystem. This calculation is shown in Equation (6).
Equation (6): Exergy destruction within a steady-state subsystem.
where
is calculated using Equation (3) and
is calculated using Equation (5).
The two routes for calculating exergy destruction, Equations (4) and (6), were used as checks to see if the values determined were the same. As can be checked for all subsystems (see
Supplemental Materials), the normalized exergy destruction as calculated using Equation (4) was within ±0.2% of the value calculated using Equation (6), which leads to high confidence in the values presented for the exergy destruction within the subsystems.
2.3. Exergy of Solid Mixtures
The only solid mixtures analyzed in this process are: (1) the solid coal entering, which is the primary source of exergy; and (2) the slag leaving the Gasifier subsystem. As in Field and Brasington [
30], the coal represented in our models is an Illinois#6 bituminous coal with a dry basis HHV of 30.5 MJ·kg
−1, which converts into a “with moisture” HHV of 27.1 MJ·kg
−1. Important values for these solid species streams are listed in
Table 3. The enthalpy, entropy and exergy of the inlet coal were estimated by simulating the coal in AspenPlus as a combination of C
12H
10O, C
4H
5N, C, S, liquid water and ash species, using the elemental analysis provided by Field and Brasington [
30]. The exergy of the inlet coal was estimated to be 28.1 MJ·kg
−1, which implies a total exergy in the inlet coal of 1747 MW. Using the HHV values listed above, one can convert the exergy efficiencies presented later in this report into HHV electrical efficiencies by multiplying by 28.1/27.1 (103%). Similar to the inlet coal, the enthalpy, entropy and exergy of the exiting slag were estimated by simulating the slag in AspenPlus as a combination of C
12H
10O, C, and ash species, leading to an exergy per mass of 5.0 MJ·kg
−1 and a total exergy remaining in the coal slag of 33 MW.
4. Membrane Model
A counter-current hollow fiber membrane module was utilized in the H2-selective and CO2-selective membrane units. The feed gas entered the shell side of the membrane module and permeated to the fibers’ bore with different permeance for different species. The remaining part of the feed gas exits the shell side (at the other end) as the retentate stream. The sweep gas (if utilized) entered the fiber bore side at the opposite end of the feed. The permeate stream exits the fiber bore side at the same end as the feed no matter whether the sweep gas was used. The hollow fiber membrane was modeled assuming an asymmetric architecture, where a thin selective layer is coated on a porous support two to three orders of magnitudes thicker. The selective layer (or skin) of the membrane faces the shell side in the module.
A one-dimensional partial-different-equation (PDE) model was developed for the membrane module in Aspen Custom Modeler (ACM). The major assumptions of the model are based on the research work of Pan [
37]. In this model, it was assumed that all fibers are identical, perfect straight, cylindrical hollow tubes and gas passes through the fiber bore side and the shell side as a plug flow. Hence, the gas flowrate and concentration were assumed to be uniformly distributed in the radial direction in both shell and fiber bore side. The membrane module was assumed to be operated under isothermal conditions so that energy balance was not considered in the model. Mass balances for all species were established along the axial direction in both shell and fiber bore side. Pressures along the length of each side of the fiber were calculated by the Hagen-Poiseuille equation. Membrane permeance and selectivity were assumed to be constant during the entire membrane life time. The detailed description of the membrane module model can be found in the work of Morinelly and Miller [
38].
In order to apply the ACM membrane model to the AspenPlus process model, the ACM model was first exported as a .MSI. Then the .msi file was installed on the computer and the included ACM model was imported into AspenPlus as a model library. This way, AspenPlus can directly implement the membrane model without invoking ACM. In order to run the membrane model, several parameters need to be provided as model inputs, including the permeability of H
2 (or CO
2), the selectivity of H
2 (or CO
2) to other species, and the pressure of permeate (if no sweep is used). Additional optional inputs include the inner and outer diameter of the hollow fibers, the length of the fiber, and the volume fraction taken up by the fibers in the module, whose default values were already specified in the model. The model then returns the total number of tubes, the membrane area, flowrates, pressures, and compositions of the permeate and the retentate streams. The .MSI files for the H
2-selective membrane module (with sweep) and CO
2-selective membrane module (without sweep) are also available as
supplemental materials, along with the Aspen .bkp files for the three cases modeled here.