基于特征选取与TSO-BP短期电力负荷预测研究

打开文本图片集

中图分类号：TM715 文献标志码：A 文章编号：1672-1098（2025）01-0057-07

Abstract：Objective In order to reduce the influence of environmental factorson the power load forecasting and improve the accuracy of short-term load forecasting.Methods In this paper，a short-termed power load prediction model basedon Pearson correlation coefficient（PCC）principal component analysis（PCA）and the tuna optimization algorithm（TSO）to improve BP neural network was proposed.Firstly，to eliminate the influence of irrelevant variables，PCC was used to select features and select meteorological atributes related to load forecasting. Secondly，the key influencing factors inthe meteorological feature sequence were extracted with theuseof PCA to eliminate the correlation and redundancyof the original sequence，reduce the input dimension of the model and improve the training eficiency.Finally，the improved model was obtained byusing TSO to search for theoptimal solution instead of therandomparameters，inorder tosolvethe problemof randomness inthe initial weights and threshold parameters of the traditional BP neural network and.Results The average absolute error percentage of the constructed model reached 0.52% when the simulation analysis was carried out with the use of the power load data of a certain region.Conclusion The model has higher prediction accuracy after the feature selection and

TSO optimization.

Key Words： feature selection ; power load forecasting;tuna swarm algorithm; Optimal parameters

随着分布式电源和智能设备在电网环境中的广泛应用，负荷数据呈现出更复杂的变化规律和特征，而传统方法相对简单，已经不能满足非线性时间序列的要求，所以神经网络逐渐取代传统方法[1]。（剩余7071字）