基于多目标蜂群进化优化的微电网能量调度方法-《郑州大学学报(工学版)》

文章信息/Info

Title:: A Novel Multi-objective Artificial Bee Colony Algorithm for Microgrid Energy Dispatching Model

Author(s):: Li Jiahua 1; Malembo 1; Wang Xingwei 1; Cheng Shi 2; Shao Yichuan 3; 1. School of Software, Northeastern University; 2. School of Computer Science, Shaanxi Normal University; 3. School of Information Engineering, Shenyang University

Keywords:: National Natural Science Foundation of China (6177021519; 61503373); Central Fundamental Research Fund (N161705001)

摘要:: 针对微电网能源调度优化问题，提出了使微电网系统运行的经济和环保的双重优化模型，根据调度系统的评估结果对调度方案进行优化。为求解该模型，提出了基于指标化拥堵距离的多目标蜂群算法（ICABC），通过建立外部档案（EA）来保存搜索过程中的非支配解；同时，为了保持解集的多样性，改进了NSGA-II的拥堵距离策略，基于指标计算拥堵距离能够避免删除密集区域的精英个体，有效地改善了Pareto前沿的分布特性。为验证所提算法的性能，将ICABC与经典的NSGA-II, MOCLPSO算法在ZDT测试集上进行了性能比较与分析。在验证实验中，将所提的模型和ICABC算法应用于解决含有多种分布式电源的微电网能量动态调度中。仿真结果表明，通过合理安排微电源的出力，所提的方法能够有效降低系统总成本。

Abstract:: Aiming at the optimization of the energy in microgrid, a bi-objective optimization model for economic and environmental operations of the microgrid systems was proposed. Then this model was assessed based on the evaluation results of the dispatching system. To be specific, in order to optimize this model, a novel multi-objective artificial bee colony algorithm called ICABC was decised, based on crowding-distance with performance indicators. This algorithm incorporated an external archive (EA) to preserve non-dominated solutions; and a novel crowding-distance method called DICC was used to maintain the diversity of solutions. DICC was essentially a updated version of traditional crowding-distance strategy in NSGA-II. It amied to avoid the removal of elite individuals in dense areas effectively while enhance the diversity of obtained non-dominated solutions, With rigor expermental evaluations on a set of benchmark problems, it showed that ICABC had more powerful performance compared with NSCA-II and MOCLPSO. Then, ICABC was applied to resolve the proposed model for dynamic dispatching of microgrid with muliple distributed generations. Simulation results exhibited that the proposed method and model effectively reduced the total system cost by rationally arranging the output of the distributed generations.