Comprehensive Multi-Criteria Comparison and Ranking of Natural Gas Liquefaction Process by Analytic Hierarchy Process (AHP)

1. Introduction

Energy is the most important element in the development of any society. Recently, natural gas has become more popular as an attractive energy source; however, its transfer to consumption locations is a challenging task. Liquefied natural gas (LNG) is easier to transfer and is more economical. LNG also constitutes the main reason of the development of natural gas liquefaction processes. Traditional LNG process included a propane pre-cooling step along with a mixed refrigerant step for gas liquefaction (C3MR). Today, technical advances and economic considerations have led to the emergence of new processes. The new processes follow several goals such as overcoming limitations (e.g., string size), process efficiency, reducing investment costs, and better performance.

Recently, natural gas liquefaction processes have attracted many researchers. Energy and exergy analyses method are used for five conventional liquefied natural gas processes (Vatani, Mehrpooya, & Palizdar, 2014b). Also, Exergy analysis of four small-scale liquefied natural gas processes was performed which showed that single mixed refrigerant (SMR) process had the best exergy efficiency (Remeljej & Hoadley, 2006). Additionally, Energy optimization in a liquefaction process by implementing genetic algorithm was carried out (Shirazi & Mowla, 2010). Exergy analysis of cascade refrigeration cycle used for natural gas liquefaction has also been reported to have a great potential for improvement (Kanoğlu, 2002). The analysis of PRICO liquefaction process including exergetic, exergoeconomic, and exergoenvironmental analysis have also been performed  (Morosuk, Tesch, Hiemann, Tsatsaronis, & Omar, 2015). The results of these studies showed the possible options for improving the LNG process. Moreover, advanced exergy analysis was performed on five natural gas liquefaction processes (Vatani, Mehrpooya, & Palizdar, 2014a). Conventional and advanced exergy analyses is studied on a cascade refrigeration system for LNG process (Tsatsaronis & Morosuk, 2010). Exergoeconomic analysis is used in single mixed refrigerant natural gas liquefaction processes and sensitivity of exergy destruction cost, and exergoeconomic factor to the operating variables of such processes (Mehrpooya & Ansarinasab, 2015).

 Selecting the best and the most suitable technology for gas liquefaction is a complex and very sensitive process which depends on many technical and economical design parameters. The technical parameters include power consumption, coefficient of performance, specific energy consumption, exergy efficiency, LNG production rate, refrigerant rate, and energy improvement potential. Economic issues include investment cost, performance cost, and lifecycle cost. To achieve an optimal LNG plant design, a comprehensive study including all relevant parameters is necessary and beneficial. Such a task is best performed by employing a multi-criteria decision-making method.

Analytic hierarchy process (AHP) method is one of the best and most accurate ranking and decision-making methods based on several indices (T. Saaty). It has been used for high energy related applications including wind observation location problem (Aras, Erdoğmuş, & Koç, 2004). A comprehensive decision-making analysis done with wind power integration projects based on improved fuzzy AHP and reported that the results attested to the feasibility of the method (Liu, Zhang, Liu, & Qian, 2012). A complete sustainability assessment process of coastal beach exploitation was presented by the analytic hierarchy process (AHP) (Tian, Bai, Sun, & Zhao, 2013). AHP model employed three dimensions of suitability, economic and social values, and ecosystem. Fuzzy AHP is used to select the best renewable energy alternatives in Indonesia (Tasri & Susilawati, 2014). Hydro power was reported as the best renewable energy source, followed by geothermal, solar, wind energy, and biomass. AHP method was used to perform a comparison between the different electricity power generation options in Jordan (Akash, Mamlook, & Mohsen, 1999). In addition to fossil fuel power plants nuclear, solar, wind, and hydro-power plants were also considered. The results showed that solar, wind, end hydro-power might be the best alternatives.

AHP method was also used to select the best renewable energy sources for sustainable development of electricity generation system in Malaysia (Ahmad & Tahar, 2014) where four major resources, hydropower, solar, wind, biomass were considered. AHP model employed four main criteria, technical, economic, social and environmental aspects, and twelve sub-criteria. Furthermore, AHP model prioritized those resources, revealing that solar was the most favorable resource followed by biomass. AHP method was utilized to select space heating systems for an industrial building (Chinese, Nardin, & Saro, 2011). The results revealed that qualitative attributes also significantly affected industrial heating system choices and the AHP was effective in handling these aspects. Additionally, this method is applied to selecting the best solar thermal collection technology for electricity generation in north-west India (Nixon, Dey, & Davies, 2010). These technologies were compared based on technical, economic and environmental criteria. In the same vein, researchers used AHP to evaluate space heating systems running on conventional and renewable energy sources in Jordan (Jaber, Jaber, Sawalha, & Mohsen, 2008). Moreover, the prioritization of the low-carbon energy sources in China by using an AHP method supports this argument (Ren & Sovacool, 2015). In addition, AHP method was used for the prioritization of energy conservation policy instruments (Kablan, 2004).

In this article, the AHP method was employed to inclusively compare and prioritize five popular natural gas liquefaction processes (MFC-Linde, DMR-APCI, C3MR- Linde, SMR-APCI and SMR-Linde) considering eight technical and economic criteria. In this article, the variations of model resulted to change in the impact weight of each criterion and their effect on the aggregate priority of the alternative LNG processes were also assessed.

2. Process Description

Linde Company introduced a simple process for natural gas liquefaction with one refrigeration cycle namely Single Mixed Refrigerant processes (SMR) (Foeg, Bach, Stockman, Heiersted, & Fredheim, 1998). Capital costs of this process are low due to few number of components. Figure 1 shows the Schematic of SMR process by Linde Company. The refrigerant used in this process was a mixture of methane, ethane, propane, butane and nitrogen. This process consisted of three compressor and four heat exchanger as main equipment.

The Air Products and Chemicals Inc. (APCI), presented another Single Mixed Refrigerant (SMR) process (Roberts, Agrawal, & Daugherty, 2002) with low equipment. Regarding to energy consumption viewpoint, SMR-APCI was better than SMR-Linde. Figure 2 shows the Schematic of SMR process by APCI Company. This process had only two heat exchangers with low capital cost.

Figure 1. Schematic of SMR-Linde Process [6]

Figure 2. Schematic of SMR-APCI Process [6]

Linde Company in another patent (Foeg, et al., 1998) presented a process for natural gas liquefaction with two refrigeration cycle namely propane pre-cooled mixed refrigerant (C3MR) process. This process for pre-cooling uses pure propane but for liquefaction and sub-cooling uses mixed refrigerant as refrigerant. Schematic of C3MR process by Linde AG is shown in Figure 3. Unlike complexity this process, it was economical due to high efficiency.

The Double Mixed Refrigerant (DMR) is a process which in pre-cooling cycle uses mixed refrigerant unlike C3MR process that uses pure propane as refrigerant in pre-cooling cycle. APCI introduced a double mixed refrigerant process with a high efficiency (Roberts & Agrawal, 2001), as shown in Figure 4. Two multi-stream heat exchangers (E-1 and E-2) were used for pre-cooling the natural gas in the first mixed refrigerant cycle, and two others heat exchangers (E-3 and E-4) were used for sub-cooling and liquefaction, respectively.

In another patent (Foeg, et al., 1998) a new high capacity LNG process called Mixed Fluid Cascade (MFC) which had three refrigeration cycles was presented by Linde AG and Stat oil. Because of three different mixed refrigerants used in each cycle, the energy efficiency of this process was high, which resulted in an increase of fixed cost and a decrease in operating costs, respectively.

Figure 5 shows the Schematic of MFC process by Linde Company.

Figure 3. Schematic of C3MR-Linde Process [6]

Figure 4. Schematic of DMR-APCI Process [6]

Figure 5. Schematic of MFC-Linde Process [6]

3. Processes Simulation

The first step in the analysis of these processes is modeling and simulation. In this article, the processes were simulated by Aspen HYSYS software ("Hyprotech HYSYS v3.2 user guide," 2003). PRSV equation of state was possible to simulate a gas process (Vatani, et al., 2014a) due to in this study PRSV was used for simulation in HYSYS. By simulation, different flow properties such as pressure, temperature, and flow rates were specified which were later required for energy and exergy analysis. The summary of the simulations results for selected streams of liquefaction processes are shown in Tables 1-5.

Table 1. Operating Conditions for SMR - Linde Process Streams [6]

Table 2. Operating Conditions for SMR-APCI Process Streams [6]

Table 3. Operating Conditions for C3MR-Linde Process Streams [6]

Table 4. Operating Conditions for DMR-APCI Process Streams [6]

Table 5. Operating Conditions for MFC-Linde Process Streams [6]

4. Energy Analysis

Specific energy consumption (SEC), coefficient of performance (COP), and power consumption (PC) were the criteria for the LNG process ranking which were obtained by the energy analysis. Specific energy consumption was defined as the ratio of the energy used in the process in kWh to LNG produced in kg; coefficient of performance was the ratio of total heat removed from the gas to total work of the cycle and the power consumed was the power required by the process. These values, which have been obtained for different LNG processes of interest from the simulation results, are given in Table 6.

5. Exergy Analysis

Exergy analysis was used in cryogenics industry for improving the efficiency of process cycles by recognizing the effect of the efficiency of equipment on the general process. The equipment or cycles whose improvement is more beneficial to the process are specified. By adding cost, reliability, and environmental requirements data to this technique, a basic method is obtained for selecting and improving LNG plants. Conventional and advanced exergy analysis indices include exergy efficiency (EE) obtained from ordinary exergy analysis and energy improvement potential (EIP) obtained from advanced exergy analysis.

The exergy destruction rate (Bejan & Tsatsaronis, 1996):

Where , and  represent the fuel exergy, product exergy and exergy destruction rates, respectively.

The exergy efficiency is defined as (Bejan & Tsatsaronis, 1996):

Advanced exergy analysis was performed based on the results of exergy analysis. The main idea of this analysis was to categorize the irreversibility or exergy destruction of the process components. Based on the removing ability, the exergy destruction was divided to two other parts:

• Avoidable exergy destruction
• Unavoidable exergy destruction

The unavoidable part of exergy destruction of the component presents a part which cannot be eliminated, even if the best available technologies are used. While avoidable part can be eliminated through technical improvements of the process equipment. Energy improvement potential of each process is defined as ratio of total avoidable exergy destruction to total exergy destruction of process (Vatani, et al., 2014a):

(3)

The higher the EIP value, there is more potential for energy improvement of the process. Exergy efficiency values and potential improvement percentages in LNG processes are given in Table 6.

Competency should be completely evaluated in terms of lifecycle and heat efficiency. Type and amount of refrigerant used in a process are important indices of liquefaction cycles. If the refrigerant is provided from products of LNG plant, lifecycle should be taken into account in the calculation of total efficiency and evaluation of final cost. The investment made in the liquefaction plants should not violate cost effectiveness of the process: The number of equipment (NOE), as the major capital cost items, utilized in the process should be as low as possible. Another important index of liquefaction cycles is LNG production rate (LPR). LNG production rate, number of equipment and refrigerant rate (RR) of the LNG processes are given in Table 6.

6. AHP Method

One of the most wide spread used methods in multi-criteria decision making models is the analytical hierarchy process (AHP), introduced in 1970 by Saaty. AHP uses a hierarchical structure to represent a decision making problem, the first step is to build a graphical representation of the problem in which the goal, criteria and alternatives are indicated. Level one in the hierarchy indicates the goal, while the criteria and factors affecting the decision goal are set in the intermediate levels and the last level is the decision alternatives. As shown in Figure 6, the goal of interest, i.e. prioritization of the LNG processes, is located at the first layer, and evaluation criteria are located in the next layers, and the last level contains the LNG processes as the decision alternatives. Due to application of different computational methods in the second layer, the data in the second level do not have a uniform scale while values with the same scale is needed to make the comparison between the data. For this reason, the criteria were normalized to a common scale within the interval [0, 1] using the following relation:

Where r_ij is normalized value and f_ij is the value of the ith criterion function for alternative jth. The AHP normalized decision matrix is shown in Table 7.

Table 6. Criteria for Natural Gas Liquefaction Processes Selection (f_ij)

Figure 6. AHP Decision Hierarchy

Table 7. AHP Normalized Decision Matrix (r_ij)

The implementation of the AHP method, involves the following steps (T. Saaty):

1- Pair comparison of decision elements and allocation of numeric values which indicates priority or importance between the two elements.

where  is the priority of the ith coefficient with respect to jth coefficient.

2- Elements of the pair comparison matrix A is then normalized using the following relation:

Then, the normalized pair comparison matrix  is obtained as:

3- Numbers in each row in the matrix  are summed up:

Then, the weight vector  is obtained from the following relation:

Where,

4- The maximum value of is obtained from the following equation:

5- The consistency rate (CR) is obtained as the ratio of consistency index (CI) to random index (RI), RI figures for different values of m as suggested by (T. L. Saaty, 2000), are shown in Table 8. For obtaining RI parameter, square matrices (n*n) with random entries but the properties of pairwise comparison matrices is formed then by calculating the average of the eigenvalues of mentioned matrices by computer RI parameter is obtained.

Where,

If CR ˂ 0.1, the pair comparison matrix has an acceptable consistency, but if CR ≥ 0.1, the pair comparison matrix is inconsistent and the comparisons must be revised.

7. Results and Discussion

7.1. LNG Processes Prioritization

The results of AHP method employed on five alternative natural gas liquefaction processes (MFC-Linde, DMR-APCI, C3MR-Linde, SMR-APCI and SMR-Linde) were prioritized according to eight criteria, namely power consumption (PC), coefficient of performance (COP), specific energy consumption (SEC), exergy efficiency (EE), LNG production rate (LPR), refrigerant rate (RR), number of equipment (NOE) used in the process, and energy improvement potential (EIP) (Tables 9 to 16).

Regarding COP criterion, MFC process, with a priority factor equal to 0.243, had higher priority over other processes with DMR process in the second rank with a priority factor of 0.208 and SMR-Linde process in the last rank with a priority factor equal to 0.171. This shows that the MFC process had the highest performance among the processes investigated. AHP results for the COP criterion of natural gas liquefaction processes are presented in Table 9.

Table 8. RI Numbers for Different Values of m

Regarding PC criteria, DMR process had more favorable condition and stayed in the first rank with a priority factor of 0.267, while in the second rank was MFC process with a priority factor of 0.209, and in the last rank was SMR-Linde process due to its higher power demand compared to other processes. Therefore, in the places with limited power access, DMR process was the favorite process. AHP results for the PC criterion of natural gas liquefaction processes are presented in Table 10.

Considering EE criteria, MFC process took the first place with a priority factor of 0.220, while in the second and fifth ranks are C3MR and SMR-Linde processes with priority factors of 0.203 and 0.170, respectively. AHP results for the EE criterion of natural gas liquefaction processes are presented in Table 11.

Regarding criteria NOE, SMR-APCI process was in the first rank due to its fewer number of equipment while C3MR process was in the last rank due to its highly complex process with larger number of equipment. AHP results for the NOE criterion of natural gas liquefaction processes are presented in Table 12.

Table 9. AHP Results for the COP Criterion of Natural Gas Liquefaction Processes

Table 10. AHP Results for the PC Criterion of Natural Gas Liquefaction Processes

Table 11. AHP Results for the EE Criterion of Natural Gas Liquefaction Processes

Table 12. AHP Results for the NOE Criterion of Natural Gas Liquefaction Processes

According to RR criterion, DMR process has the highest rank because it used fewer refrigerant rates compared to other processes, while MFC process was in the last rank due to its great refrigerant rate. AHP results for the RR criterion of natural gas liquefaction processes are presented in Table 13.

MFC Process produces high LNG production rate, and therefore, its specific energy consumption (SEC) was lower than other processes and had more favorable condition, while SMR-Linde process was in the last rank in terms of SEC criterion. AHP results for the RR criterion of natural gas liquefaction processes are presented in Table 14.

Considering EIP criteria, MFC process took the first place with a priority factor of 0.231, while in the second and fifth ranks were C3MR and DMR processes with priority factors of 0.224 and 0.173, respectively. AHP results for the EIP criterion of natural gas liquefaction processes are presented in Table 15.

Table 13. AHP Results for the RR Criterion of Natural Gas Liquefaction Processes

Table 14. AHP Results for the SEC Criterion of Natural Gas Liquefaction Processes

Table 15. AHP Results for the EIP Criterion of Natural Gas Liquefaction Processes

Considering LPR criteria, MFC process took the first place with a priority factor of 0.213, while in the fifth ranks was DMR process with priority factor of 0.154. AHP results for the LPR criterion of natural gas liquefaction processes are presented in Table 16.

The result of AHP for prioritization of the natural gas liquefaction processes is shown in Table 17. As shown, when all the criteria were simultaneously taken into consideration, DMR process had a relatively higher priority over the other processes and ranked the first with a priority equal to 0.231; while MFC, C3MR, SMR-APCI and SMR-Linde processes were ranked in the next places, respectively.

Table 16. AHP Results for the LPR Criterion of Natural Gas Liquefaction Processes

Table 17. AHP Results for Prioritization of the Natural Gas Liquefaction Processes

7.2. Criterion Impact Weight Alterations Analysis

To this point, it was assumed that all criteria had equal impact or significance on the LNG plants different processes overall performance. However, there were many instances in which one of this criterion had a greater impact on the LNG processes, because of technical, geographical, energy source and other limitations on the site. Therefore, in this section, the changes in the importance of each criterion on the LNG processes ranking, which was labeled as impact weight, was investigated. Also it should be noted that this a different weight to what was used in the previous section as the process priority factors to prioritize different LNG processes.

As shown in Figure 7, axes X and Y show the criterion’s impact weight and alternative’s LNG processes priority factors, respectively. For example when that weight of COP was zero, (this means that the COP criterion had removed and the number of criteria has got to 7), weight of other criteria were the same and equal to (1/7=0.143). Also, when that weight of COP was one, (This means that the ranking was done only on the basis of COP criterion and the other criteria had removed), weight of other criteria were the same and equal to zero. The vertical dashed line on X axis indicated the location of the impact weight in the previous section analysis, in which all criteria impact weights were the same and equal to (1/8=0.125).

Responses of the LNG processes to the variation in impact weight of criterion COP are shown Figure 7. As shown, by a 20% increase and decrease in the impact weight of criterion COP, the order of prioritization did not change; however, by a 30% increase or more in the impact weight of criterion COP, DMR and C3MR were respectively replaced by alternatives MFC Linde and SMR-APCI processes. MFC-Linde process showed the highest sensitivity, while DMR-APCI process had the lowest sensitivity to variation in the impact weight of criterion COP, also no increases or decreases was seen in the ranking of SMR-Linde process.

The rankings alterations of the alternatives process by the variation in impact weight of criterion PC is shown in Figure 8. As shown in the figure, by increasing the impact weight of criterion PC, no change in the prioritization of alternatives was observed; however, a 30% decrease in the impact weight of criterion PC, the rankings of the alternatives DMR and C3MR were respectively replaced by alternatives MFC and SMR-APCI processes. DMR process had the highest sensitivity, while MFC and C3MR processes has the lowest sensitivity to the variation in the impact weight of criterion PC, also, no increases or decreases was seen in the ranking of the SMR-Linde process.

Figure 7. Variations in Performance Score of LNG Pocesses with Respect to Weight of COP

Figure 8. Variations in Performance Score of LNG Processes with Respect to Weight of PC

The rankings alterations of the alternatives process by the variation in impact weight of criterion EE is shown in Figure 9. As shown in the figure, by a 20% increase or decrease in the impact weight of criterion EE, no change in the prioritization of alternative processes was observed; but by a 30% increase in the impact weight of criterion EE, alternative process DMR was ranked after alternative processes MFC and C3MR, also, no increases or decreases was seen in the ranking of the SMR-Linde process.

The rankings alterations of the alternatives process by the variation in impact weight of criterion EIP is shown in Figure 10. As shown in the figure, by a 30% decrease in the impact weight of criterion EIP, alternative process SMR-APCI would have a higher rank than alternative process C3MR. Alternative process DMR had higher sensitivity to the criterion EIP and when the weight of criterion EIP is 0.15, 0.23, 0.6, and 0.96, the rank of this alternative was replaced by alternatives MFC, C3MR, SMR-Linde and SMR-APCI processes, respectively.

Figure 9. Variations in Performance Score of LNG Processes with Respect to Weight of EE

Figure 10. Variations in Performance Score of LNG Processes with Respect to Weight of EIP

The rankings alterations of the alternatives process by the variation in impact weight of criterion LPR is shown in Figure 11. As shown in the figure, by decreasing the impact weight of criterion LPR, no change in the prioritization of alternatives was observed. Alternative process DMR has higher sensitivity to criterion LPR and when the impact weight of criterion LPR was 0.14, 0.3, 0.34, and 0.42, this alternative was replaced by alternatives MFC, C3MR, SMR- APCI and SMR- Linde processes, respectively.

The rankings alterations of the alternatives process by the variation in impact weight of criterion NOE is shown in Figure 12. As shown in the figure, by a 20% decrease in the impact weight of criterion NOE, alternative process DMR exchange its rank with alternative process MFC, and by a 20% increase in the impact weight of criterion NOE, alternative process C3MR ranking was replaced by alternative process SMR-APCI.

Figure 11. Variations in Performance Score of LNG Processes with Respect to Weight of LPR

Figure 12. Variations in Performance Score of LNG Processes with Respect to Weight of NOE

The rankings alterations of the alternatives process by the variation in impact weight of criterion RR is shown in Figure 13. As shown in the figure, by a 20% decrease in the impact weight of criterion RR, alternative process MFC was in the first rank, while alternative DMR was in the second rank. Moreover, alternatives processes DMR and MFC had the highest sensitivity to this criterion.

The rankings alterations of the alternatives process by the variation in impact weight of criterion SEC is shown in Figure 14. As shown in the figure, decreasing the impact weight of criterion SEC to 0.06 causes a change in the ranks of alternatives processes C3MR and SMR-APCI, and when the weight of criterion SEC is 0.21 and 0.79, alternative process DMR was replaced by alternatives MFC and C3MR processes, respectively, also, no increases or decreases was seen in the ranking of the SMR-Linde process

Figure 13. Variations in Performance Score of LNG Processes with Respect to Weight of RR

Figure 14. Variations in Performance Score of LNG Processes with Respect to Weight of SEC