Swarm and Evolutionary Computation最新文献_第2页

Deep reinforcement learning based bi-level optimization for order acceptance and scheduling: A slow-moving inventory activation policy for resilient manufacturing 基于深度强化学习的订单接受和调度双级优化：弹性制造的慢动库存激活策略

IF 8.5 1区计算机科学

Swarm and Evolutionary Computation Pub Date : 2026-04-01 Epub Date: 2026-03-31 DOI: 10.1016/j.swevo.2026.102369

Min Kong , Yexing Shen , Chao Zuo , Weizhong Wang , Muhammet Deveci , Hana Tomášková , Dursun Delen

{"title":"Deep reinforcement learning based bi-level optimization for order acceptance and scheduling: A slow-moving inventory activation policy for resilient manufacturing","authors":"Min Kong , Yexing Shen , Chao Zuo , Weizhong Wang , Muhammet Deveci , Hana Tomášková , Dursun Delen","doi":"10.1016/j.swevo.2026.102369","DOIUrl":"10.1016/j.swevo.2026.102369","url":null,"abstract":"<div><div>Order acceptance and scheduling (OAS) is a crucial operation in resilient manufacturing, involving production planning, resource allocation, inventory management, cost control, and supply chain coordination. Industrial robot manufacturers, in particular, face substantial inventory pressures when managing slow-moving materials. To address this challenge, this study introduces a bi-level programming model that integrates OAS with Slow-Moving Inventory (SMI) management. In this model, the upper level, managed by the sales department, evaluates the revenue potential of orders while considering their impact on SMI reduction. The lower level, led by the production department, schedules accepted orders to minimize makespan based on the upper level’s acceptance or rejection decisions. This research presents two novel hybrid algorithms: the Deep Q-Network enhanced heuristic (DQN<img>H) and the Deep Q-Network enhanced meta-heuristic (DQN-MEAL), which incorporate the decision-making capabilities of Deep Q-Networks (DQN) into dynamic order acceptance. An acceleration strategy is also developed to enhance computational efficiency. Numerical experiments show that DQN integration boosts the performance of DQN<img>H by up to 13.07%, and most DQN-MEAL algorithms exhibit significant performance improvements, with the exception of slight fluctuations in the Equilibrium Optimizer (EO).</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"104 ","pages":"Article 102369"},"PeriodicalIF":8.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147602320","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Parallel ant colony optimization for vehicle routing with parcel lockers 带包裹箱的车辆路径并行蚁群优化

IF 8.5 1区计算机科学

Swarm and Evolutionary Computation Pub Date : 2026-04-01 Epub Date: 2026-03-28 DOI: 10.1016/j.swevo.2026.102371

Phu An Chau , Loan T.T. Nguyen , Witold Pedrycz , Bay Vo

{"title":"Parallel ant colony optimization for vehicle routing with parcel lockers","authors":"Phu An Chau , Loan T.T. Nguyen , Witold Pedrycz , Bay Vo","doi":"10.1016/j.swevo.2026.102371","DOIUrl":"10.1016/j.swevo.2026.102371","url":null,"abstract":"<div><div>The recent surge in Vietnam’s e-commerce has significantly strained urban infrastructure, leading to increased last-mile delivery times, higher transportation costs, and diminished customer satisfaction. In response, many countries are adopting sustainable solutions like parcel lockers, which facilitate contactless deliveries. However, integrating parcel lockers introduces a new variant to the classic routing problem: the Vehicle Routing Problem with Parcel Lockers (VRPPL), which demands effective management of multiple delivery options. To address this, this research proposes a novel 3D Enhanced Parallel Ant Colony Optimization (3D-PACO) model. Our core contribution is the introduction of a pioneering multidimensional pheromone matrix to extend the traditional Ant Colony Optimization (ACO) for the VRPPL’s multiple delivery options. Further enhancing this, we propose an adaptive mechanism for dynamic ant allocation per thread for economical resource utilization. Finally, we adopt a master–slave parallel schema to the VRPPL enabling deployment of a large total number of ants. The experimental results on various VRPPL benchmark datasets indicate statistically significant improvements in solution quality with tremendous savings in computational runtime of 46.4% to 83.5% across different dataset sizes, compared to the original work.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"104 ","pages":"Article 102371"},"PeriodicalIF":8.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147602322","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A clustering-based improved simulated annealing algorithm for collaborative logistics vehicle routing problem with time windows and demand priority rules 基于聚类的改进模拟退火算法求解具有时间窗和需求优先规则的协同物流车辆路径问题

IF 8.5 1区计算机科学

Swarm and Evolutionary Computation Pub Date : 2026-04-01 Epub Date: 2026-04-11 DOI: 10.1016/j.swevo.2026.102388

Yong Wang , Jing Liu , Xiaoqin Zhao , Yuanhan Wei , Lu Zhen , Jingxin Zhou

{"title":"A clustering-based improved simulated annealing algorithm for collaborative logistics vehicle routing problem with time windows and demand priority rules","authors":"Yong Wang , Jing Liu , Xiaoqin Zhao , Yuanhan Wei , Lu Zhen , Jingxin Zhou","doi":"10.1016/j.swevo.2026.102388","DOIUrl":"10.1016/j.swevo.2026.102388","url":null,"abstract":"<div><div>The logistics sector is critically constrained by limited transportation resources, struggling to balance the scarcity of resources with the competing imperative of superior customer service. However, collaborative logistics models can offer a promising approach to enhancing logistics network efficiency. This study addresses these challenges by proposing and solving a collaborative logistics vehicle routing problem with time windows and demand priority rules. First, a novel customer service priority division scheme based on time windows is developed to characterize the urgency of customer demands, facilitating the categorization of customers into priority tiers to improve service efficiency and responsiveness. Then, a bi-objective optimization model is formulated to minimize the total operational cost (TOC) and minimize the number of vehicles (NV). A hybrid heuristic algorithm, integrating 3D <em>k</em>-means clustering and an improved multi-objective simulated annealing with hill climbing algorithm is devised to solve the proposed model, with Pareto optimal solution selected based on non-dominated sorting and the crowding distance criterion. The efficacy of the proposed algorithm is demonstrated through comparative analysis against the CPLEX solver and seven multi-objective optimization algorithms. A case study in Chongqing City, China, validates the effectiveness of the proposed approach, as the TOC is $7176 and the NV is 12, supported by high statistical confidence intervals. Furthermore, the proposed approach enables reasonable allocation of transportation resources with limited logistics resources and enhances service efficiency through priority decision-making. This study provides valuable guidance for the development of urban logistics systems, improving overall operation performance and sustainability of logistics networks.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"104 ","pages":"Article 102388"},"PeriodicalIF":8.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147650303","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

GDG: An evolutionary oversampling framework integrating Gaussian mixture modeling and genetic algorithm 融合高斯混合建模和遗传算法的进化过采样框架

IF 8.5 1区计算机科学

Swarm and Evolutionary Computation Pub Date : 2026-04-01 Epub Date: 2026-04-06 DOI: 10.1016/j.swevo.2026.102375

Yelin Zhang, Dongmei Wang, Yuehua Yu, Chen Chen, Chengwang Xie

{"title":"GDG: An evolutionary oversampling framework integrating Gaussian mixture modeling and genetic algorithm","authors":"Yelin Zhang, Dongmei Wang, Yuehua Yu, Chen Chen, Chengwang Xie","doi":"10.1016/j.swevo.2026.102375","DOIUrl":"10.1016/j.swevo.2026.102375","url":null,"abstract":"<div><div>Class imbalance induces significant bias in machine learning classifiers. While oversampling mitigates this, a critical knowledge gap exists: conventional generative methods often assume unimodal distributions and fail to address complex boundary overlap, leading to noisy, low-fidelity synthetic samples. To bridge this gap, we propose GDG, a novel framework integrating Gaussian Mixture Model (GMM) and Genetic Algorithm (GA). First, GMM clusters minority samples to accurately capture intrinsic multi-modal structures. Subsequently, an innovative global–local mechanism adaptively allocates synthetic samples based on boundary complexity, effectively minimizing overlap. Lastly, the GA performs a nonlinear search within superspheres, utilizing adaptive fitness weights to balance exploration and exploitation for high-quality generation. Extensive experiments on 21 benchmark datasets demonstrate that GDG significantly outperforms nine state-of-the-art baselines, improving average Accuracy by 1.9%, G-mean by 6.0%, and AUC by 1.2%. Rigorous non-parametric statistical analysis confirms these differences (<span><math><mrow><mi>p</mi><mo>=</mo><mn>1</mn><mo>.</mo><mn>78</mn><mo>×</mo><mn>1</mn><msup><mrow><mn>0</mn></mrow><mrow><mo>−</mo><mn>7</mn></mrow></msup></mrow></math></span>), with post-hoc Nemenyi testing verifying that GDG achieves the superior average rank of 2.17. These findings establish GDG as a robust, statistically validated solution for tackling complex class imbalance problems.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"104 ","pages":"Article 102375"},"PeriodicalIF":8.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147650308","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Multi-objective optimization for integrated energy systems using non-dominated sorting genetic and swarm intelligence methods 基于非支配排序遗传和群体智能方法的综合能源系统多目标优化

IF 8.5 1区计算机科学

Swarm and Evolutionary Computation Pub Date : 2026-04-01 Epub Date: 2026-04-13 DOI: 10.1016/j.swevo.2026.102384

Feng Li , Yonghui Chen , Ying Hu , Yang Yu , Lining Xing

{"title":"Multi-objective optimization for integrated energy systems using non-dominated sorting genetic and swarm intelligence methods","authors":"Feng Li , Yonghui Chen , Ying Hu , Yang Yu , Lining Xing","doi":"10.1016/j.swevo.2026.102384","DOIUrl":"10.1016/j.swevo.2026.102384","url":null,"abstract":"<div><div>This paper investigates a multi-objective optimization scheme for electricity-gas-heat integrated energy systems (EGH-IES) involving wind power and photovoltaic uncertainty by using non-dominated sorting genetic and swarm intelligence methods. Firstly, power output scenarios of wind and photovoltaic power are analyzed through Frank-Copula function-based probability density. Additionally, the <em>k</em>-means clustering technique is introduced for reducing power output scenarios to obtain representative uncertainty outputs. Furthermore, stepped carbon trading mechanism is considered for guiding EGH-IES to control carbon emissions, then a multiple objective optimization model is constructed by mechanism model and operation constraints of EGH-IES, which involves comprehensive operation cost and carbon emissions cost. Finally, integrating the advantages of non-dominated sorting genetic technique with elite strategy and wolf pack algorithm, a multi-objective optimization scheme, i.e., non-dominated sorting genetic with elite strategy-wolf pack (NSGES-WP) algorithm, is carried out, the diversity of solutions is maintained and the convergence speed is improved. The simulation results demonstrate that compared with the NSGA-II method, the proposed method reduces comprehensive operation costs by 22.75 % and carbon emissions by 3.28 %. The presented optimization methodology could provide a feasible solution for reducing the operation cost and carbon emissions of EGH-IES.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"104 ","pages":"Article 102384"},"PeriodicalIF":8.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147703361","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

DSPDE: A synergistic integration of dynamic stratification and probabilistic escape for enhanced differential evolution optimization DSPDE：动态分层和概率逃逸的协同集成，用于增强差分进化优化

IF 8.5 1区计算机科学

Swarm and Evolutionary Computation Pub Date : 2026-04-01 Epub Date: 2026-04-08 DOI: 10.1016/j.swevo.2026.102380

Chiwen Qu , Juchuan Yuan , Lupeng Zhang , Xinyue Zhang , Zhengxin Huang

{"title":"DSPDE: A synergistic integration of dynamic stratification and probabilistic escape for enhanced differential evolution optimization","authors":"Chiwen Qu , Juchuan Yuan , Lupeng Zhang , Xinyue Zhang , Zhengxin Huang","doi":"10.1016/j.swevo.2026.102380","DOIUrl":"10.1016/j.swevo.2026.102380","url":null,"abstract":"<div><div>The lack of explicit metrics for online population state assessment makes it challenging to dynamically balance exploration and exploitation in Differential Evolution (DE). To address this, this paper proposes DSPDE, a novel DE variant that introduces a closed-loop framework to achieve the synergistic integration of dynamic stratification with a probabilistic escape strategy. This framework enables algorithm self-regulation by continuously monitoring population distribution and individual stagnation, then triggering targeted structural and behavioral adjustments. Specifically, DSPDE first employs a dynamic stratification mechanism that partitions the population into elite, ordinary, and inferior subgroups based on fitness distribution. The subgroup sizes are adaptively adjusted during the evolutionary process, ensuring resource allocation aligns with evolutionary stages. Concurrently, a probabilistic escape strategy monitors individuals using a Gaussian mixture model to quantify stagnation risk. Upon detecting stagnation, it intelligently selects between an anti-gradient escape operator (for directional exploration) and a mirror resampling operator (for diversity recovery), thereby enhancing local optimum avoidance. The synergy lies in the state-aware coordination loop: stratification results inform escape probability, while escape outcomes refine subsequent stratification. This creates an adaptive “perceive-adjust-respond” cycle for population management. Extensive experiments on the CEC2022 benchmark and practical mechanical engineering problems demonstrate that DSPDE significantly outperforms state-of-the-art DE variants in convergence accuracy, speed, and robustness, with statistical verification. Theoretical analysis of the escape strategy’s exploration capability and ablation studies validating each module’s contribution are also provided.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"104 ","pages":"Article 102380"},"PeriodicalIF":8.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147650309","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A global feedback learning-based memetic algorithm for energy-aware scheduling in collaborative heterogeneous flexible job shops 基于全局反馈学习的异构柔性作业车间能量感知调度模因算法

IF 8.5 1区计算机科学

Swarm and Evolutionary Computation Pub Date : 2026-04-01 Epub Date: 2026-04-08 DOI: 10.1016/j.swevo.2026.102392

Hongquan Qu, Shiliang Shao, Yunhong Xu, Maolin Cai, Yan Shi, Xiaomeng Tong

{"title":"A global feedback learning-based memetic algorithm for energy-aware scheduling in collaborative heterogeneous flexible job shops","authors":"Hongquan Qu, Shiliang Shao, Yunhong Xu, Maolin Cai, Yan Shi, Xiaomeng Tong","doi":"10.1016/j.swevo.2026.102392","DOIUrl":"10.1016/j.swevo.2026.102392","url":null,"abstract":"<div><div>Industry 5.0 emphasizes the collaboration between human and robot resources in perception, decision-making, and execution, forming a dynamic and complementary cooperative mechanism. However, multi-flexible resource collaboration significantly enlarges the solution space, which increases the difficulty of resource allocation and optimization. At the same time, the matching of jobs with machines and the switching of operational states, such as startup and shutdown, directly affect energy consumption, making energy savings and cost reduction rigid requirements. To address these issues, a collaborative heterogeneous flexible job shop scheduling (CHFJS) model is formulated, with the primary objectives of minimizing makespan and energy consumption. Subsequently, a memetic algorithm based on global feedback learning (GFLMA) is proposed to solve the CHFJS problem. A total of 12 neighborhood structures are designed, and a Bayesian inference and weighting-based local search strategy is established. Additionally, energy-saving operators are specifically designed to address the problem characteristics. Finally, extensive experiments on instances of various scales are conducted to validate the effectiveness of GFLMA. The results demonstrate that the proposed algorithm outperforms state-of-the-art algorithms in more than 70% of the instances, confirming the superiority of GFLMA.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"104 ","pages":"Article 102392"},"PeriodicalIF":8.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147650389","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Knowledge-driven large-scale multi-objective evolutionary learning for interval prediction of key quality indicators in blast furnace ironmaking process 基于知识驱动的高炉炼铁过程关键质量指标区间预测的大规模多目标进化学习

IF 8.5 1区计算机科学

Swarm and Evolutionary Computation Pub Date : 2026-04-01 Epub Date: 2026-04-06 DOI: 10.1016/j.swevo.2026.102379

Jingchuan Zhang , Yaxue Liu , Xianpeng Wang

{"title":"Knowledge-driven large-scale multi-objective evolutionary learning for interval prediction of key quality indicators in blast furnace ironmaking process","authors":"Jingchuan Zhang , Yaxue Liu , Xianpeng Wang","doi":"10.1016/j.swevo.2026.102379","DOIUrl":"10.1016/j.swevo.2026.102379","url":null,"abstract":"<div><div>Accurate prediction interval (PI) modeling of key quality indicators in the blast furnace (BF) ironmaking process is crucial for maintaining its stable operations. Nevertheless, existing approaches often face challenges in accurately capturing highly nonlinear input–output relationships and sample-dependent heteroscedastic uncertainty commonly observed in industrial data, while suffering from severely degraded efficiency when the corresponding PI modeling task involves a large-scale search space. To address these challenges, this paper formulates the PI modeling task as a large-scale multi-objective optimization problem (MOP) that simultaneously maximizes PI coverage probability and minimizes interval width; and then proposes a dual knowledge learning-based large-scale multi-objective evolutionary algorithm (DKL-LSMOEA) to enable fast and robust PI modeling. In DKL-LSMOEA, high-quality datasets are first constructed by collecting and labeling the evolutionary process data, from which two complementary types of knowledge are learned, i.e., structural importance knowledge and directional distribution knowledge. Based on this knowledge, a knowledge-driven reproduction operator is then developed to enhance search efficiency and convergence performance within a more compact search space. Extensive experimental results demonstrate that DKL-LSMOEA outperforms state-of-the-art algorithms on both benchmark large-scale MOPs and the real-world PI modeling task in BF ironmaking, achieving a more favorable trade-off between interval coverage and interval width and thus showing strong potential for practical industrial applications.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"104 ","pages":"Article 102379"},"PeriodicalIF":8.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147650307","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Solving electric vehicle routing problem with heterogeneous drones and no-fly zones using enhanced adaptive large neighborhood search algorithm 基于增强的自适应大邻域搜索算法求解异构无人机和禁飞区下电动汽车路径问题

IF 8.5 1区计算机科学

Swarm and Evolutionary Computation Pub Date : 2026-04-01 Epub Date: 2026-04-10 DOI: 10.1016/j.swevo.2026.102382

Wanting Chen, Shuai Zhang, Kang Jiang

{"title":"Solving electric vehicle routing problem with heterogeneous drones and no-fly zones using enhanced adaptive large neighborhood search algorithm","authors":"Wanting Chen, Shuai Zhang, Kang Jiang","doi":"10.1016/j.swevo.2026.102382","DOIUrl":"10.1016/j.swevo.2026.102382","url":null,"abstract":"<div><div>The collaboration of electric vehicles and drones in last-mile deliveries meets the demands of a booming e-economy while promoting environmental sustainability. However, challenges arise for drones, such as fixed payload compartments unsuitable for diverse package sizes and restricted flight areas due to regulatory policies. Thus, this study proposes a novel model for the electric vehicle routing problem with heterogeneous drones and no-fly zones to tackle the mentioned challenges. First, the proposed model extends the drones to heterogeneous types with different flight endurance, payload capacity, and flight speed for real-world scenarios. Second, it defines the restricted flight area as a no-fly zone, which is close to reality. A detour strategy is devised to address the drone routes that conflict with the no-fly zones. To solve this model, an enhanced adaptive large neighborhood search algorithm with three-layer coding scheme is presented. The algorithm utilizes a longest detour removal operator to avoid local optimal solutions and integrates a no-fly zone tabu strategy to enhance global search capability. Finally, the experimental results demonstrate the effectiveness of the proposed algorithm.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"104 ","pages":"Article 102382"},"PeriodicalIF":8.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147650390","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Differential evolution with a variance contribution ratio-based diversity enhancement mechanism for numerical optimization 基于方差贡献比的差分进化多样性增强机制的数值优化

IF 8.5 1区计算机科学

Swarm and Evolutionary Computation Pub Date : 2026-04-01 Epub Date: 2026-03-30 DOI: 10.1016/j.swevo.2026.102359

Liqi Zhao , Liangliang Sun , Zhenghao Song , Qichun Zhang , Natalja Matsveichuk , Yuri Sotskov

{"title":"Differential evolution with a variance contribution ratio-based diversity enhancement mechanism for numerical optimization","authors":"Liqi Zhao , Liangliang Sun , Zhenghao Song , Qichun Zhang , Natalja Matsveichuk , Yuri Sotskov","doi":"10.1016/j.swevo.2026.102359","DOIUrl":"10.1016/j.swevo.2026.102359","url":null,"abstract":"<div><div>Owing to its fast convergence speed and high optimization accuracy, differential evolution (DE) has become one of the benchmark algorithms for solving continuous optimization problems. However, when tackling complex optimization problems such as high-dimensional and multimodal tasks, DE often suffers from evolutionary stagnation, insufficient global search capability, and premature convergence. To address these challenges, this paper proposes Differential Evolution with a Variance Contribution Ratio-Based Diversity Enhancement Mechanism (DE-VCR). The proposed algorithm incorporates the following three major improvements. First, an infinity-norm-based parameter generation and update mechanism is designed, which exploits information from the current best individual to guide adaptive control parameter adjustment. Second, a gamma-distribution-based mutation strategy is proposed to enhance global exploration by transforming low-fitness individuals. Finally, a population diversity enhancement mechanism based on the variance contribution ratio is introduced and combined with principal component analysis to effectively alleviate evolutionary stagnation. To evaluate the performance of DE-VCR, extensive experiments are conducted on 376 benchmark problems from the CEC2013, CEC2014, CEC2017, and CEC2022 test suites, with comparisons against various state-of-the-art DE variants and other evolutionary algorithms. The experimental results demonstrate that DE-VCR exhibits strong competitiveness in terms of optimization accuracy. In addition, the algorithm shows good practicality and robustness in real-world applications such as parameter estimation of photovoltaic power systems.</div></div>","PeriodicalId":48682,"journal":{"name":"Swarm and Evolutionary Computation","volume":"104 ","pages":"Article 102359"},"PeriodicalIF":8.5,"publicationDate":"2026-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147602319","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0