Combination Of Genetic Algorithm Research Articles

Parametric analysis of wastewater electrolysis for green hydrogen production: A combined RSM, genetic algorithm, and particle swarm optimization approach

The production of green hydrogen from wastewater presents a significant opportunity to address environmental challenges associated with wastewater treatment and fulfill the increasing demand for renewable and sustainable energy sources. This study aimed to determine the optimal operating parameters for maximizing hydrogen production through wastewater electrolysis. To design the experimental runs, a Box-Behnken design (BBD) with an L17 array was implemented, while Response Surface Methodology (RSM) in conjunction with Genetic Algorithms (GA) and Particle Swarm Optimization (PSO) was employed to develop predictive models and optimize the operating parameters. The operating parameters considered were the catalyst amount (g), electrode voltage (V), and electrolysis time (min), and the response variable was the oxyhydrogen (HHO) gas generation rate (L/min). The comparative analysis shows that the optimized parameters obtained through RSM surpassed those obtained through GA and PSO, with a catalyst amount of 99.42 g, an electrode voltage of 22.9 V, and an electrolysis time of 23.17 min. Consequently, the HHO generation rate reached a maximum value of 12.42 L/min. Furthermore, the experimental validation indicated a close agreement between the model's predicted results and the actual experimental outcomes. This study shows the effectiveness of combining RSM, GA, and PSO for accurate prediction and optimization of operating parameters in wastewater electrolysis for green hydrogen production. The identified optimal conditions contribute to enhanced efficiency and increased hydrogen yield, thereby promoting the advancement of sustainable energy systems.

Research on Morphological Innovation Design of Industrial Products Based on Combinatorial Genetic Algorithm

Abstract This paper applies the combinatorial genetic algorithm to the design of industrial product forms to achieve the effect of industrial product form design innovation. The standard genetic algorithm is applied to industrial product design, and the optimal order of genetic optimization is constructed through continuous population genetics, combined with the minimum value of the objective function to construct the constraints of the innovative design of industrial products. Using the method of combinatorial optimization to integrate different elements in industrial product design, a coding method based on rows and columns is proposed for the set coverage problem of design elements. Analyzing the user’s preference for industrial product form design allows us to explore the demand direction of industrial product innovation design and formulate a suitable innovation design scheme. The results show that the consensus satisfaction of hair dryer form design in genetic optimization 2nd and 9th are higher than 0.8. When users are stimulated by like and dislike sample pictures, there is no significant difference in the average response time, p=0.337>0.05, and the difference in the average response time of the users is more significant p=0.020<0.05 when they are confronted with the preferred design shape.

A Rigid Image Registration by Combined Local Features and Genetic Algorithms

Abstract Image registration is an essential pre-processing step required for many image processing applications such as medical imaging and computer vision. The aim is to geometrically align two or more images of the same scene by establishing a mapping that relies on each point from one image to its corresponding point of another image. Scale invariant feature transform (SIFT) and speeded up robust features (SURF) are well-liked local features descriptors that have been extensively utilised for feature-based image registration due to their inherent properties such as invariance, changes in illumination, and noise. Moreover, the task of registration can be viewed as an optimization problem that can be solved by applying genetic algorithms (GAs). This paper presents an efficient feature image registration method based on combined local features and GAs. Firstly, the procedure consists of extracting the local features from the images by combining SIFT and SURF algorithms and matching them to refine the feature set data. Therefore, an adaptive GA based on fitness sharing and elitism techniques is employed to find the optimal rigid transformation parameters that best align the feature points by minimizing a distance metric. The suggested method is applied for registering medical images and the obtained results are significant compared to other feature-based approaches with reasonable computation time.

A combined genetic algorithm and active learning approach to build and test surrogate models in Process Systems Engineering

In Process Systems Engineering, computationally-demanding models are frequent and plentiful. Handling such complexity in an optimization framework in a fast and reliable way is essential, not only for generating meaningful solutions but also for providing decision support. Indeed, optimization results need to be obtained efficiently without compromising accuracy or solution quality. Surrogate models are a cheap way of replacing complex ones, while still capturing the intrinsic features that make them unique and valuable. In this work, a methodology to build surrogate models is developed. It combines a genetic algorithm with an active learning method, harvesting the benefits of both approaches — on the one hand leveraging nature-inspired optimization procedures that explore the optimization space area, and conversely a ‘smart’ approach to adding meaningful sample points to the training stage. The methodology, coined GA-AL, is tested and validated for chemical absorption of CO2 in a biogas mixture inserted in a utility superstructure framework. Nine surrogate modes are tested, with Artificial Neural Networks, Random Forest and Kringing outperforming other approaches, assessed via four performance criteria.

Swarm Intelligence Response Methods Based on Urban Crime Event Prediction

Cities attract a large number of inhabitants due to their more advanced industrial and commercial sectors and more abundant and convenient living conditions. According to statistics, more than half of the world’s population resides in urban areas, contributing to the prosperity of cities. However, it also brings more crime risks to the city. Crime prediction based on spatiotemporal data, along with the implementation of multiple unmanned drone patrols and responses, can effectively reduce a city’s crime rate. This paper utilizes machine learning and data mining techniques, predicts crime incidents in small geographic areas with short timeframes, and proposes a random forest algorithm based on oversampling, which outperforms other prediction algorithms in terms of performance. The research results indicate that the random forest algorithm based on oversampling can effectively predict crimes with an accuracy rate of up to 95%, and an AUC value close to 0.99. Based on the crime prediction results, this paper proposes a multi-drone patrol response strategy to patrol and respond to predicted high-crime areas, which is based on target clustering and combined genetic algorithms. This strategy may help with the pre-warning patrol planning within an hourly range. This paper aims to combine crime event predictions with crowd-sourced cruise responses to proactively identify potential crimes, providing an effective solution to reduce urban crime rates.

Hybrid genetic algorithm to minimize scheduling cost with unequal and job dependent earliness tardiness cost

This article presents two combinatorial genetic algorithms (GA), unequal earliness tardiness-GA (UET-GA) and job-dependent earliness tardiness-GA (JDET-GA) for the single-machine scheduling problem to minimize earliness tardiness (ET) cost. The sequence of jobs produced in basic UET and JDET as a chromosome is added to the random population of GA. The best sequence from each epoch is also injected as a population member in the subsequent epoch. The proposed improvement seeks to achieve convergence in less time to search for an optimal solution. Although the GA has been implemented very successfully on many different types of optimization problems, it has been learnt that the algorithm has a search ability difficulty that makes computations NP-hard for types of optimization problems, such as permutation-based optimization problems. The use of a plain random population initialization results in this flaw. To reinforce the random population initialization, the proposed enhancement is utilized to obtain convergence and find a promising solution. The cost is further significantly lowered offering the due date as a decision variable with JDET-GA. Multiple tests were run on well-known single-machine benchmark examples to demonstrate the efficacy of the proposed methodology, and the results are displayed by comparing them with the fundamental UET and JDET approaches with a notable improvement in cost reduction.

Interaction behavior study of HCl on (ZnS)n (n = 1–12) clusters and HCl effect on Hg0 adsorption

AbstractThe interaction behavior of HCl and (ZnS)n (n = 1–12) clusters and HCl effect on Hg0 adsorbed by (ZnS)n have been studied theoretically. The combined genetic algorithm and density functional theory (GA‐DFT) method has been used to obtain the structures of (ZnS)nHCl and (ZnS)nHgHCl (n = 1–12) clusters. The structural properties of (ZnS)nHCl and (ZnS)nHgHCl have been analyzed. The adsorption energies and interaction energies have been calculated. Bond length and bond order analysis has revealed that SH and ZnCl bonds form after HCl adsorbed on (ZnS)n clusters, while Hg0 can only weakly bind with (ZnS)nHCl clusters. According to thermodynamic adsorption analysis, the formation of (ZnS)nHCl clusters from (ZnS)n and HCl are spontaneous because of their negative Gibbs free energy changes. The formation of (ZnS)nHgHCl from (ZnS)nHCl and Hg are nonspontaneous for n = 1–4 and 9, and the Gibbs free energy changes have small negative values for other sizes. Electron localization function and noncovalent interaction (NCI) analysis of (ZnS)10HgHCl manifest that Hg and its nearest Zn form zinc amalgam. Projected density of state study has been performed to obtain the interaction nature of HCl and (ZnS)n clusters and Hg0 adsorption on (ZnS)nHCl clusters. Based on our study, HCl is chemical adsorbed by (ZnS)n clusters except (ZnS)4 cluster. After (ZnS)n adsorbs HCl, Hg0 can physically adsorb on (ZnS)nHCl clusters. The strength of Hg0 on (ZnS)nHCl is comparable to that of Hg0 on (ZnS)n, indicating that HCl can hardly affect the adsorption of Hg0 on ZnS clusters.

Conceptual design and optimisation of a novel hybrid device for capturing offshore wind and wave energy

AbstractThe access to the offshore wind resource in the deep sea requires the development of innovative solutions which reduce the cost of energy. Novel technologies propose the hybrid combination of wind and wave energy to improve the synergy between these technologies sharing costs, such as mooring and electrical connexion. This work proposes a novel hybrid wind and wave energy system integrating a floating offshore wind turbine with three-point absorbers wave energy converters (WECs). The WECs are an integral part of the floating structure and contribute significantly to the hydrostatic and dynamic stability of the system. Their geometry is optimised considering a cylindrical, semi-cylindrical and spherical shape for the Pantelleria case study. The cylindrical shape with the largest radius and the lowest height is the optimal solution in terms of reducing structural costs and maximising the performance of the WECs. The in-house hydrostatic stability tool and the time domain model MOST are used to optimise the WECs, with a combined meta-heuristic genetic algorithm with the Kriging surrogate model and a local Nelder–Mead optimization in the final simulations. The power of the WECs is estimated with both linear and variable motor flow hydraulic PTOs to obtain a more realistic electrical power generation. Generally, the hybrid device proved to be more competitive than the floating wind turbine alone, with a LCOE reduction up to 11%. Performance of the hybrid device can be further improved when more energetic sites are considered, as the energy generated by the WECs is higher.

Prediction of extranodal extension in head and neck squamous cell carcinoma by CT images using an evolutionary learning model

BackgroundExtranodal extension (ENE) in head and neck squamous cell carcinoma (HNSCC) correlates to poor prognoses and influences treatment strategies. Deep learning may yield promising performance of predicting ENE in HNSCC but lack of transparency and interpretability. This work proposes an evolutionary learning method, called EL-ENE, to establish a more interpretable ENE prediction model for aiding clinical diagnosis.MethodsThere were 364 HNSCC patients who underwent neck lymph node (LN) dissection with pre-operative contrast-enhanced computerized tomography images. All the 778 LNs were divided into training and test sets with the ratio 8:2. EL-ENE uses an inheritable bi-objective combinatorial genetic algorithm for optimal feature selection and parameter setting of support vector machine. The diagnostic performances of the ENE prediction model and radiologists were compared using independent test datasets.ResultsThe EL-ENE model achieved the test accuracy of 80.00%, sensitivity of 81.13%, and specificity of 79.44% for ENE detection. The three radiologists achieved the mean diagnostic accuracy of 70.4%, sensitivity of 75.6%, and specificity of 67.9%. The features of gray-level texture and 3D morphology of LNs played essential roles in predicting ENE.ConclusionsThe EL-ENE method provided an accurate, comprehensible, and robust model to predict ENE in HNSCC with interpretable radiomic features for expanding clinical knowledge. The proposed transparent prediction models are more trustworthy and may increase their acceptance in daily clinical practice.

Developing an analytical bond-order potential for Hf/Nb/Ta/Zr/C system using machine learning global optimization

Despite being effective in unveiling the complicated nanostructure-property relation of (HfNbTaZr)C high-entropy carbide (HEC), applications of atomistic simulations are restricted by the lack of reliable many-body interatomic potentials. In this study, we propose a machine learning (ML) global optimization framework to parameterize an analytical bond-order potential (ABOP) with massive internal parameters for the Hf/Nb/Ta/Zr/C system. The combined distributed breeder genetic algorithm (DBGA), extensive ab initio training data and a normalized objective function markedly accelerate the convergence towards global optimum. By performing a cross-validation for the established potential, we demonstrate that the ABOP can accurately predict fundamental energetic, structural and mechanical properties of metals, intermetallic compounds, alloys and carbides in both the training and test datasets. Moreover, the potential is employed in dynamic simulations such as thermal annealing, simple loadings, nanoindentation and irradiation collision, from which some typical experimental observations are successfully reproduced. This not only further verifies the reliability of ABOP, but also implies its broad future applicability in nanoscale simulations.

A Hybridization Approach to Solve the Capacitated Asymmetric Allocation Hub Location Problem

This paper deals with a capacitated asymmetric allocation hub location problem (CAAHLP). We determine the number of hubs, the locations of hubs, and asymmetric allocation of non-hub nodes to hub with the objective of minimum total transportation costs satisfying the required service level. To solve the problem optimally, we present a 0-1 integer programming model and find an optimal solution using CPLEX. As the CAAHLP has impractically demanding for the large-sized problem, a solution method based on combined ant colony optimization algorithm and genetic algorithm is developed which solve hub location problem and node allocation problem respectively. We investigate performance of the proposed solution method through the comparative study. Key Words: Hub Location, Asymmetric Allocation, Ant Colony Optimization, Genetic Algorithm

Fast power density aware three‐dimensional integrated circuit floorplanning for hard macroblocks using best operator combination genetic algorithm

AbstractIn this article, we propose a fast three‐dimensional integrated circuit (3D‐IC) floorplanning method for hard macroblocks that includes a thermal management scheme. It applies a genetic algorithm constituted by an optimal combination of crossover and mutation operations to identify the optimal solution for design variables, namely, total wire length, number of through‐silicon vias (TSVs), and maximum average layer power density. The proposed method additionally makes use of a unique TSV placement scheme that arranges TSVs next to their respective functional blocks. To enable efficient heat transmission to the ambient environment, layers with higher power densities are placed closer to the heat sink. The proposed 3D‐IC floorplanning approach provides the fewest TSVs, the lowest peak temperature, and promising values of wire length within the least amount of computation time. Compared to the recent fast thermal analysis for fixed‐outline 3D‐floorplanning, it generates 13.14% shorter wire length, 39.27% lower peak temperature, and 34.35% lesser number of TSVs on average with significant improvement in computation time, while analyzing GSRC thermal benchmark circuits.

Simulation-Based Optimization of Injection Molding Process Parameters for Minimizing Warpage by ANN and GA

Plastic injection molding is one of the most used methods for producing plastic products because it can be produced at a high production rate, low cost, and ease in manufacturing. However, one defect that affects product quality is namely warpage. To reduce plastic product warpage, the injection molding process is required optimal process control to increase plastic product quality. The objective of this paper is to optimize injection molding process parameters for minimizing the warpage of plastic glass. The optimization process is divided into two phases. The Finite Element Method (FEM) was employed in the first phase to simulate 32 experiments under various parameters. The parameters of this process consist of melt temperature ranging from 180 to 230 °C, mold temperature in the range of 20 – 45 °C, filling time from 0.82 to 0.92 s, packing time ranging from 5.88 to 7 s and cooling time of 14 to 18 s. In the second phase, Artificial Neural Network (ANN) combined Genetic Algorithm (GA) was developed to predict the warpage and solve the optimization process to find optimal parameters. Combining the intelligent method shows that ANN and GA effectively find the optimal process parameters that can reduce the warpage of the product by 35.73% from the maximum value.

Hollow polyhedral structures and properties of Ag2n-1Sn- (n = 2-11) clusters: A theoretical study.

The structures of Ag2n-1Sn- (n = 2-11) clusters are obtained by the combination of genetic algorithm (GA) and density functional theory (DFT). All the global minimum structures prefer hollow polyhedral structures, in which S-Ag-S element, triangular Ag3S3 and tetragonal Ag4S4 units present to stabilize the structures. The S atoms in the structures appear in μ3-S or μ4-S form. Adiabatic and vertical electron affinities of the clusters have been obtained, which reveals that they increases as cluster size. Stability analysis shows that Ag9S5- and Ag19S10- have special stability. The HOMO, LUMO orbitals of the clusters are obtained and the orbital components of them are calculated. The HOMO orbitals are mainly from the p orbitals of S atoms, whereas the s, p and d orbitals of Ag atoms contribute much bigger than the p orbitals of S atoms for LUMO orbitals. The orbital delocalization indexes (ODI) of the HOMOs and LUMOs are calculated, and the small ODIs of the HOMOs and LUMOs for n = 4-10 reveal that these orbitals are highly delocalized. By studying the projected density of states and molecular orbitals of Ag9S5- and Ag19S10- clusters, it is found that their molecular orbitals have superatomic properties. Superatomic properties play an important role in stabilizing clusters. This work used combined genetic algorithm and density functional theory (GA-DFT), and PBE0/Lanl2tz(Ag)/6-311G(d,p)(S) method to optimize the structures. Gaussian 16 program, Gauss view 6.0.16 program and Multiwfn 3.8 code are the softwares used.

Multi-gene GP and GA-FIS Models to Deal with Scaling Problem in the ANFIS model for Estimating Roughness Coefficient in Erodible Channels

Estimation of the roughness coefficient is important for reliable hydraulic design in erodible channels. In this paper, the capability of multi-gene Genetic Programming (GP), a combined Genetic Algorithm and Fuzzy Inference System (GA-FIS) model, and Multi Regression (MR) methods are employed to estimate the roughness coefficient. These methods try to extract either an explicit or an implicit relationship between the roughness coefficient and input variables. In addition, traditional GP, widely used by researchers, and conventional empirical formulas are implemented to evaluate the models. Results show that the employed methods are more accurate than empirical methods. In addition, the effects of some other parameters, such as non-dimensional water depth and shear Reynolds number, are highlighted over the roughness coefficient while previously ignored in the empirical methods. Also, findings prove that the GA is a helpful tool to optimize a FIS compared with gradient-based models like ANFIS, while the scale of input variables is not in the same order. The R2 for multi-gene GP and GA-FIS are 0.8504 and 0.8842, respectively, while this value for the most accurate empirical method (Yalin, 1992) is 0.6286.

On the structure and electronic properties of Ptn clusters: new most stable structures for n = 16-17.

The lowest energy structures and electronic properties of Ptn clusters up to n = 17 are investigated by using a genetic algorithm in combination with density functional theory calculations. There are several putative global minimum structures for platinum clusters which have been reported by using different approaches, but a comprehensive study for n = 15-17 has not been carried out so far. Herein, we perform a consensus using GGA (PBE), meta-GGA (TPSS) and hybrid (B3PW91, PBE0, PBEh-3c, M06-L) functionals in conjunction with the Def2-TZVP basis set. New most stable structures are found for Pt16 and Pt17, which are slightly lower in energy than the previously reported global minima. Molecular dynamics simulations show that the clusters are rigid at room temperature. We analyze the structural, electronic, energy and vibrational data of the investigated clusters in detail.

Production Scheduling Model and Machine Assignment to Automotive Parts Manufacturing for Optimum Production

PT. XYZ is a company engaged in the field of Automotive Parts Manufacturer. In carrying out its production this company uses make to order. The company carries out simple planning and scheduling, namely FCFS (First Come First Serve). The company has a problem, namely there is no production scheduling with the appropriate machine assignment. This causes the company to sometimes experience delays in completing the product. The purpose of this study is to overcome scheduling problems in companies to minimize order completion time (makespan). The method used in this study is the Heuristic Pour method, namely the development of a new heuristic algorithm in completing flowshop scheduling with the aim of minimizing makespan based on a combination approach and Genetic Algorithm which will be assisted with the help of MATLAB R2016a software which is a computational algorithm commonly used to solve a search for values in an optimization problem. Based on the data processing that has been done, the results of production scheduling are obtained using the Heuristic Pour method in the order Job 4 (J4)-Job 2 (J2) -Job 7 (J7) -Job 6 (J6) -Job 5 (J5) -Job 3 (J3) ) -Job 1 (J1) gets a makespan of 19440 minutes or equal to 16 working days with a total delay of 1 working day. While production scheduling with the selected Genetic Algorithm method getsthe order Job 4 (J4) -Job 2 (J2) -Job 7 (J7) -Job 6 (J6) -Job 3 (J3) -Job 1 (J1) -Job 5 ( J5) with the best makespan of 18660 minutes or 15 days, and there was no delay in completing requests. The company is expected to consider or review the use of the appropriate production scheduling method because it can provide a more optimum demand completion time (makespan), so the company can reduce delays in fulfilling customer requests.

Site selection for artificial recharge in Cisangkuy sub-watershed, West Java, using combined fuzzy logic and genetic algorithm

The Cisangkuy sub-watershed is part of the Citarum watershed with ±300 km of river length. The total area is in line with the potential of water resources which is quite good, but over time the hydrological condition of the site has decreased. Decreasing the hydrological state is due to changes in land use and land cover from year to year and is indicated by high discharge and flooding in the rainy season. This study aims to find the optimal location of artificial recharge to maintain water balance in the Cisangkuy sub-watershed. Fuzzy logic is used for determining the location of artificial recharge by considering the input variable DRASTIC in the form of depth to groundwater level, net recharge, aquifer media, soil type, slope, vadose zone, and hydraulic conductivity. In addition, subtractive clustering is performed to obtain the class of each DRASTIC parameter. Furthermore, a genetic algorithm is carried out to get the optimal location of the artificial recharge zone. The priority zone is indicated by the high class of each DRASTIC parameter input. Combined fuzzy logic and genetic algorithm show that the optimal location for artificial recharge is in the northern part of the Cisangkuy sub-watershed, precisely in the Banjaran area.

Portfolio rebalancing based on a combined method of ensemble machine learning and genetic algorithm

PurposeThis paper presents a combined method of ensemble learning and genetics to rebalance the corporate portfolio. The primary purpose of this paper is to determine the amount of investment in each of the shares of the listed company and the time of purchase, holding or sale of shares to maximize total return and reduce investment risk.Design/methodology/approachTo achieve the goals of the problem, a two-level combined intelligent method, such as a support vector machine, decision tree, network Bayesian, k-nearest neighbors and multilayer perceptron neural network as heterogeneous basic models of ensemble learning in the first level, was applied. Then, the majority vote method (weighted average) in the second stage as the final model of learning was collectively used. Therefore, the data collected from 208 listed companies active in the Tehran stock exchange (http://tsetmc.com) from 2011 to 2015 have been used to teach the data. For testing and analysis, the data of the same companies between 2016 and 2020 have been used.FindingsThe results showed that the method of combined ensemble learning and genetics has the highest total stock portfolio yield of 114.12%, with a risk of 0.905%. Also, by examining the rate of return on capital, it was observed that the proposed method has the highest average rate of return on investment of 110.64%. As a result, the proposed method leads to higher returns with lower risk than the purchase and maintenance method for fund managers and companies and predicts market trends.Research limitations/implicationsIn the forthcoming research, there were no limitations to obtain research data were easily extracted from the site of Tehran Stock Exchange Technology Management Company and Rahvard Novin software, and simulation was performed in MATLAB software.Practical implicationsIn this paper, using combined machine learning methods, companies’ stock prices are predicted and stock portfolio optimization is optimized. As companies and private organizations are trying to increase their rate of return, so they need a way to predict stock prices based on specific indicators. It turned out that this algorithm has the highest stock portfolio return with reasonable investment risk, and therefore, investors, portfolio managers and market timers can be used this method to optimize the stock portfolio.Social implicationsThe homogeneous and heterogeneous two-level hybrid model presented in the research can be used to predict market trends by market timers and fund managers. Also, adjusting the portfolio with this method has a much higher return than the return on buying and holding, and with controlled risk, it increases the security of investors’ capital, and investors invest their capital in the funds more safely. And will achieve their expected returns. As a result, the psychological security gained from using this method for portfolio arrangement will eventually lead to the growth of the capital market.Originality/valueThis paper tries to present the best combination of stock portfolios of active companies of the Tehran Stock Exchange by using the two-level combined intelligent method and genetic algorithm.

Modelling and optimization of liquid air energy storage systems with different liquefaction cycles

Liquid air energy storage (LAES) is one of the large-scale mechanical energy storage technologies which are expected to solve the issue of renewable energy power storage and peak shaving. As the main energy loss of a standalone LAES occurs in the liquefaction process, this paper focused on the thermodynamic analysis of LAES systems with different liquefaction cycles. Six liquefaction cycles were considered. To obtain a higher system efficiency and fair comparative analysis of the performance of different LAES systems, authors used a combined genetic algorithm (GA) and HYSYS process simulation method to perform the multi-parameters global optimization. Subsequently, energy and exergy analyses were conducted on the optimized LAES systems. The results show that the system efficiency of the LAES based on turbo-expander refrigeration (i.e., the Claude cycle, Kapitza cycle, and Heylandt cycle) increases with an increase in the split ratio. The LAES with Kapitza liquefaction cycle shows the most competitive system performance compared with the other liquefaction cycles based-LAES systems. The LAES with Kapitza liquefaction cycle exhibits the lowest optimal charging pressure of 10.73 MPa and achieves the highest round-trip efficiency, system storage exergy efficiency, and system recovery exergy efficiency of 60.02 %, 86.41 %, and 84.98 %, respectively. The obtained GA optimization and thermodynamic analysis results of LAES systems with different liquefaction configurations can provide significant theoretical guidance for further industrial applications of LAES.