基于VMD-SWO-RF的短期电力负荷预测研究

Research on Short-Term Power Load Forecasting Based on VMD-SWO-RF

摘要: 短期电力负荷预测对智能电网的电力系统的安全稳定运行及调控具有重要意义。本文提出了一种基于“分解与优化”框架的短期电力负荷预测方法。首先，通过变分模态分解(Variable Mode Decomposition, VMD)对负荷数据进行分解，提取不同时间尺度的模态分量。然后，利用随机森林(Random Forest, RF)模型对各模态分量分别进行预测，并结合蜘蛛蜂优化算法(SWO)对模型的关键超参数进行优化，从而提升预测的精度和模型的泛化能力。最后，将所有模态分量的预测结果进行整合，得到最终的负荷预测值。实验结果显示，VMD-SWO-RF模型的4个评价指标均表现最佳，MAE为136.59，RMSE为190.52，MAPE为1.55%，R²达0.98。相比VMD-SSA-RF、VMD-RF、RF、KNN、LSTM和SVM模型，VMD-SWO-RF在MAE上分别减少了1.94 kW、8.75 kW、12.31 kW、23.66 kW、13.53 kW和19.19 kW，展现出更小的误差和更高的预测精度，充分验证了其优越性，能够更好挖掘电力负荷数据中的复杂特征关系，在短期电力负荷预测任务中表现出显著优势，具有良好的应用前景。

Abstract: Short-term power load forecasting is of great significance to the safe and stable operation and regulation of power systems in smart grids. This paper proposes a short-term power load forecasting method based on the “decomposition and optimization” framework. First, Variable Mode Decomposition (VMD) is applied to decompose the load data, so as to extract modal components at different time scales. Then, the Random Forest (RF) model is employed to predict each modal component individually, and the Spider Wasp Optimization (SWO) algorithm is integrated to optimize the key hyperparameters of the model, thereby improving the prediction accuracy and the generalization ability of the model. Finally, the prediction results of all modal components are integrated to obtain the final load prediction value. Experimental results show that the VMD-SWO-RF model performs best in all four evaluation metrics: the MAE is 136.59, RMSE is 190.52, MAPE is 1.55%, and R² reaches 0.98. Compared with the VMD-SSA-RF, VMD-RF, RF, KNN, LSTM, and SVM models, the VMD-SWO-RF reduces the MAE by 1.94 kW, 8.75 kW, 12.31 kW, 23.66 kW, 13.53 kW, and 19.19 kW respectively. It exhibits smaller errors and higher prediction accuracy, fully verifying its superiority. This model can better mine the complex feature relationships in power load data, demonstrates significant advantages in short-term power load forecasting tasks, and possesses promising application prospects.