Scholarlink

<p>The Coat is Optimization Algorithm (COA), proposed by Dehghani et al., is a metaheuristic inspired by two natural behaviors of coatis, exhibiting strong global optimization capabilities. However, as the number of iterations increases, population diversity gradually decreases and the search range shrinks, which weakens global exploration, slows convergence, and increases the likelihood of premature convergence to local optima. To address these limitations, this paper proposes an improved version of the COA, named AICOA, aiming to enhance global search ability and avoid local stagnation. First, a Random Opposition-Based Learning strategy is introduced to increase initial population diversity and improve early-stage exploration. Second, a Gaussian walk strategy is applied during the exploration phase to enhance search perturbation and maintain diversity. Third, a Levy flight mechanism is integrated into the exploitation phase to allow large-range jumps and escape from local traps. The effectiveness of AICOA is validated through comparisons with five baseline algorithms on the CEC2017 and CEC2022 benchmark functions. Ablation studies and Wilcoxon rank-sum tests further demonstrate that the proposed strategies significantly improve performance over the original COA. Finally, AICOA is combined with the Extreme Learning Machine (ELM) to construct a short-term power load forecasting model. Experiments on actual 2024 load data from a Chinese province show that the AICOA-ELM model outperforms other models in both prediction accuracy and convergence speed, effectively reducing forecasting errors and enhancing short-term load prediction performance.</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

Multi-strategy improved coatis optimization algorithm: enhancing global exploration and local exploitation abilities

  • 期刊论文

  • 发表日期：

    2025年12月29日

查看全文
  • Wenhui Liu,
  • Yu Zhang

摘要

The Coat is Optimization Algorithm (COA), proposed by Dehghani et al., is a metaheuristic inspired by two natural behaviors of coatis, exhibiting strong global optimization capabilities. However, as the number of iterations increases, population diversity gradually decreases and the search range shrinks, which weakens global exploration, slows convergence, and increases the likelihood of premature convergence to local optima. To address these limitations, this paper proposes an improved version of the COA, named AICOA, aiming to enhance global search ability and avoid local stagnation. First, a Random Opposition-Based Learning strategy is introduced to increase initial population diversity and improve early-stage exploration. Second, a Gaussian walk strategy is applied during the exploration phase to enhance search perturbation and maintain diversity. Third, a Levy flight mechanism is integrated into the exploitation phase to allow large-range jumps and escape from local traps. The effectiveness of AICOA is validated through comparisons with five baseline algorithms on the CEC2017 and CEC2022 benchmark functions. Ablation studies and Wilcoxon rank-sum tests further demonstrate that the proposed strategies significantly improve performance over the original COA. Finally, AICOA is combined with the Extreme Learning Machine (ELM) to construct a short-term power load forecasting model. Experiments on actual 2024 load data from a Chinese province show that the AICOA-ELM model outperforms other models in both prediction accuracy and convergence speed, effectively reducing forecasting errors and enhancing short-term load prediction performance.