基于特征选择策略和TCN的电力负荷预测方法  

Power load forecasting method based on feature selection strategy and TCN

作　　者：袁文辉 张仰飞[1] YUAN Wen-hui;ZHANG Yang-fei(School of Electric Power Engineering,Nanjing Institute of Technology,Nanjing 211167,China)

机构地区：[1]南京工程学院电力工程学院,南京211167

出　　处：《信息技术》2024年第4期9-14,21,共7页Information Technology

基　　金：国家自然科学基金(52107098)。

摘　　要：电力负荷由于受到多种外部因素影响,具有较大的波动性和随机性,使得高精度的负荷预测十分困难。为有效处理高维特征以提高模型预测精度,提出了一种基于特征选择策略和时间卷积神经网络的电力负荷预测方法。首先,采用基于极端梯度提升树的特征选择策略,深度挖掘与负荷关联性强的特征作为预测模型的输入;其次,构建基于时间卷积神经网络(TCN)的电力负荷预测模型,对特征选择后的负荷数据进行预测;最后,采用某市真实负荷数据进行仿真分析。结果表明,文中所提方法与传统预测方法相比,具有更高的预测精度。Due to the influence of many external factors,power load has great fluctuation and randomness,which makes it very difficult to predict the load with high accuracy.In order to effectively deal with high-dimensional features and improve prediction accuracy,a power load prediction method based on feature selection strategy and Temporal Convolutional Network(TCN)is proposed.Firstly,the feature selection strategy based on eXtreme Gradient Boosting(XGBoost)is adopted to deeply mine the features with strong correlation with load as the input of the prediction model.Secondly,a power load prediction model based on TCN is constructed to predict the load data after feature selection.Finally,the real load data of a city is used for simulation analysis,and the results show that the proposed method has higher prediction accuracy than the traditional prediction method.

关 键 词：多维特征 负荷预测 极端梯度提升树 特征选择策略 时间卷积神经网络 

分 类 号：TM732[电气工程—电力系统及自动化]