Scholarlink

<p>Due to the continuous increase in urban residents and more convenient transportation, the number of people traveling during the Spring Festival has surged, leading to higher household heating demand and drastic fluctuations in electricity load. These high dynamic loads make power generation control challenging, causing significant losses and impacting grid safety and dispatch. This paper proposes a neural network framework capable of adapting to high dynamic loads and making accurate predictions. It screens influencing factors and introduces Complementary Ensemble Empirical Mode Decomposition (CEEMD). The framework utilizes a Bidirectional Long Short-Term Memory (BiLSTM) structure with an attention mechanism for subcomponent prediction, eliminating interference and improving accuracy. Parameter optimization is achieved using logarithmic mapping and a hybrid algorithm-improved Whale Optimization Algorithm (BWOPs). The combined prediction model, CEEMD-BWOPs-BiLSTM(SA), is validated through simulations and comparative analysis with other models, demonstrating higher accuracy in electricity load forecasting with superior MAPE and RMSE 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

Logarithmic mapping and multi-algorithm collaborative optimization for high dynamic load forecasting

  • 期刊论文

  • 发表日期：

    2025年2月23日

查看全文
  • Xifeng Guo,
  • Hongye Zhang,
  • Yi Ning,
  • Di Zheng,
  • Wenzhuo Cong

摘要

Due to the continuous increase in urban residents and more convenient transportation, the number of people traveling during the Spring Festival has surged, leading to higher household heating demand and drastic fluctuations in electricity load. These high dynamic loads make power generation control challenging, causing significant losses and impacting grid safety and dispatch. This paper proposes a neural network framework capable of adapting to high dynamic loads and making accurate predictions. It screens influencing factors and introduces Complementary Ensemble Empirical Mode Decomposition (CEEMD). The framework utilizes a Bidirectional Long Short-Term Memory (BiLSTM) structure with an attention mechanism for subcomponent prediction, eliminating interference and improving accuracy. Parameter optimization is achieved using logarithmic mapping and a hybrid algorithm-improved Whale Optimization Algorithm (BWOPs). The combined prediction model, CEEMD-BWOPs-BiLSTM(SA), is validated through simulations and comparative analysis with other models, demonstrating higher accuracy in electricity load forecasting with superior MAPE and RMSE performance.