Scholarlink

<p>This paper presents a robust hybrid optimization algorithm that combines artificial intelligence with genetic algorithms (GA) to maximize revenue for electricity generation plants, addressing the challenges posed by wind generation uncertainty in liberalized power markets. The novel method leverages the prediction capabilities of the long short-term memory algorithm, a deep learning methodology, to forecast superior genetic traits. These enhanced individuals are then incorporated into the population, accelerating the evolutionary process of the GA and improving its ability to achieve local optimality. The effectiveness of the proposed algorithm is demonstrated through experimental validation of the revenue optimization problem, aiming to increase wind energy utilization by reducing compensation risks related to electricity production fluctuations in the market. The performance of the algorithm in recommending wind power bidding capacity is evaluated using the IEEE 30-bus and 118-bus power system model. Comparative analysis shows that auctioned wind power consumption increased by 12% compared to the traditional GA and by more than 20% compared to the mixed integer linear programming (MILP) method, with a corresponding revenue increase of 7% compared to the MILP scenario. Furthermore, comparison with previous advanced GA research on the optimal power flow problem indicates not only a reduction in the number of generations but also significant savings in computation time; the effectiveness of the approach is confirmed by a more than 22% reduction in the NFFE index (the number of fitness function evaluations).</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

A novel hybrid algorithm: long short-term memory-genetic algorithm for optimizing uncertain revenue of wind farms in electricity markets

  • 期刊论文

  • 发表日期：

    2025年8月28日

查看全文
  • Ngoc Sang Dinh,
  • Le Song Binh Dinh,
  • Viet Anh Truong

摘要

This paper presents a robust hybrid optimization algorithm that combines artificial intelligence with genetic algorithms (GA) to maximize revenue for electricity generation plants, addressing the challenges posed by wind generation uncertainty in liberalized power markets. The novel method leverages the prediction capabilities of the long short-term memory algorithm, a deep learning methodology, to forecast superior genetic traits. These enhanced individuals are then incorporated into the population, accelerating the evolutionary process of the GA and improving its ability to achieve local optimality. The effectiveness of the proposed algorithm is demonstrated through experimental validation of the revenue optimization problem, aiming to increase wind energy utilization by reducing compensation risks related to electricity production fluctuations in the market. The performance of the algorithm in recommending wind power bidding capacity is evaluated using the IEEE 30-bus and 118-bus power system model. Comparative analysis shows that auctioned wind power consumption increased by 12% compared to the traditional GA and by more than 20% compared to the mixed integer linear programming (MILP) method, with a corresponding revenue increase of 7% compared to the MILP scenario. Furthermore, comparison with previous advanced GA research on the optimal power flow problem indicates not only a reduction in the number of generations but also significant savings in computation time; the effectiveness of the approach is confirmed by a more than 22% reduction in the NFFE index (the number of fitness function evaluations).