Scholarlink

<p>Accurate short-team load forecasting is crucial for maintaining power system stability, particularly given the increasing complexity of modern grids and the nonlinear characteristics of load data. This study proposes a hybrid model integrating Ensemble Empirical Mode Decomposition (EEMD), Sparrow Search Algorithm (SSA), and Bidirectional Long Short-Term Memory (BiLSTM). First, EEMD is applied to decompose the load time-series data into multiple Intrinsic Mode Functions (IMFs), effectively extracting the underlying patterns and trends. Then, the BiLSTM network is used to capture the temporal and nonlinear dynamics of the load data. Meanwhile, SSA is employed to optimize the BiLSTM parameters, enhancing its ability to model the temporal features of the load data. To validate the superiority of the proposed EEMD-SSA-BiLSTM model, this study compares it with BiLSTM, SSA-BiLSTM, and EEMD-BiLSTM models. Experimental results on two distinct datasets demonstrate the superior performance of the proposed EEMD-SSA-BiLSTM model. Compared to the SSA-BiLSTM model, which serves as the closest baseline, the proposed approach achieves a reduction in MAE of 58.4% and 50.6% across the two datasets, and a reduction in RMSE of 56.8% and 41.4%, respectively. Additionally, the model maintains a NSE above 0.985 in all cases, demonstrating highly accurate predictions.</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

Short-term electrical load forecasting strategy based on EEMD-SSA-BiLSTM

  • 期刊论文

  • 发表日期：

    2025年6月19日

查看全文
  • XinHao Zhang

摘要

Accurate short-team load forecasting is crucial for maintaining power system stability, particularly given the increasing complexity of modern grids and the nonlinear characteristics of load data. This study proposes a hybrid model integrating Ensemble Empirical Mode Decomposition (EEMD), Sparrow Search Algorithm (SSA), and Bidirectional Long Short-Term Memory (BiLSTM). First, EEMD is applied to decompose the load time-series data into multiple Intrinsic Mode Functions (IMFs), effectively extracting the underlying patterns and trends. Then, the BiLSTM network is used to capture the temporal and nonlinear dynamics of the load data. Meanwhile, SSA is employed to optimize the BiLSTM parameters, enhancing its ability to model the temporal features of the load data. To validate the superiority of the proposed EEMD-SSA-BiLSTM model, this study compares it with BiLSTM, SSA-BiLSTM, and EEMD-BiLSTM models. Experimental results on two distinct datasets demonstrate the superior performance of the proposed EEMD-SSA-BiLSTM model. Compared to the SSA-BiLSTM model, which serves as the closest baseline, the proposed approach achieves a reduction in MAE of 58.4% and 50.6% across the two datasets, and a reduction in RMSE of 56.8% and 41.4%, respectively. Additionally, the model maintains a NSE above 0.985 in all cases, demonstrating highly accurate predictions.