Scholarlink

<p>To address the serious problems caused by excessive fossil fuel consumption, countries are actively exploring various solutions, which include integrating renewable energy into existing power systems while controlling costs and optimizing the distribution of electric vehicle charging stations to facilitate the transition from gasoline-powered to electric vehicles. To address them, we proposed a data-driven evolutionary algorithm with model calibration strategy and data perturbation strategy (DDEA-MCDP). We used historical data related to these two problems to create solutions through data perturbation, which were used to train surrogate models. We also used model calibration strategies to enhance information interactive between surrogate models. Five solutions were evaluated by these models and used to train another surrogate model, which were selected based on the consistency of predictions from three other surrogate models and their distance to the population center. These new solutions were reliable configurations that enabled the surrogate model to learn more characteristics. We validated the effectiveness of DDEA-MCDP by addressing charging station distribution problem in Irish and photovoltaic-diesel generator system problem in California and Germany. To demonstrate the performance of DDEA-MCDP, we conducted ablation experiments and comparative experiments, Wilcoxon rank-sum test with similar algorithms, including SRK-DDEA, TT-DDEA, DDEA-PES, MS-DDEO, CL-DDEA, and BDDEA-LDG. The results indicate that DDEA-MCDP provided a reasonable scheduling mechanism that minimizes costs and energy utilization.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Data-driven evolutionary algorithm with model calibration strategy and data perturbation strategy for electricity resource economic scheduling optimization

  • 期刊论文

  • 发表日期：

    2026年1月15日

查看全文
  • Ruxin Zhao,
  • Jiajie Kang,
  • Lixiang Fu,
  • Chang Liu,
  • Chao Jiang,
  • Yang Shi

摘要

To address the serious problems caused by excessive fossil fuel consumption, countries are actively exploring various solutions, which include integrating renewable energy into existing power systems while controlling costs and optimizing the distribution of electric vehicle charging stations to facilitate the transition from gasoline-powered to electric vehicles. To address them, we proposed a data-driven evolutionary algorithm with model calibration strategy and data perturbation strategy (DDEA-MCDP). We used historical data related to these two problems to create solutions through data perturbation, which were used to train surrogate models. We also used model calibration strategies to enhance information interactive between surrogate models. Five solutions were evaluated by these models and used to train another surrogate model, which were selected based on the consistency of predictions from three other surrogate models and their distance to the population center. These new solutions were reliable configurations that enabled the surrogate model to learn more characteristics. We validated the effectiveness of DDEA-MCDP by addressing charging station distribution problem in Irish and photovoltaic-diesel generator system problem in California and Germany. To demonstrate the performance of DDEA-MCDP, we conducted ablation experiments and comparative experiments, Wilcoxon rank-sum test with similar algorithms, including SRK-DDEA, TT-DDEA, DDEA-PES, MS-DDEO, CL-DDEA, and BDDEA-LDG. The results indicate that DDEA-MCDP provided a reasonable scheduling mechanism that minimizes costs and energy utilization.