考虑多维时域特征的行业中长期负荷预测方法

Medium-and Long-term Industry Load Forecasting Method Considering Multi-dimensional Temporal Features

中长期负荷预测是电力系统规划与设计的重要依据,准确的行业中长期负荷预测能为电网布局规划、检修计划制定等提供决策支撑。在此背景下,针对行业中长期负荷隐含的多维时域特征,提出一种基于分解和预测思想的行业中长期负荷预测方法。首先,构建了基于周期趋势分解算法的行业中长期负荷特征分解模型,以得到分别表征行业负荷变化趋势性、周期性以及随机性特征的趋势分量、周期分量及残差分量;接着,针对分解得到的各维度分量,分别构建了基于门控循环单元的行业负荷全局趋势特征提取与预测模型、基于卷积神经网络的负荷周期局部特征提取模型,以及基于改进自适应高斯核密度估计的负荷残差概率密度预测模型,由此形成考虑多维时域特征的行业中长期负荷预测方法。最后,以中国某市化工行业负荷数据为例,验证了所提预测方法的有效性。

Medium-and long-term load forecasting is an important basis for power system planning and design,and precise medium-and long-term industry load forecasting can provide the decision support for power system planning and maintenance programming.On this background,a medium-and long-term industry load forecasting method based on the idea of decomposition and forecasting is proposed to account for the multi-dimensional temporal features hidden in the medium-and long-term industry load.Firstly,the feature decomposition model of medium-and long-term industry load based on the seasonal trend decomposition algorithm is constructed to obtain the trend,periodic,and residual components,which represent the trend,periodical and stochastic features of the industry load,respectively.Secondly,according to each dimensional decomposed component obtained by decomposition,a global trend feature extraction and forecasting model based on gate recurrent unit,a local load feature extraction model based on convolutional neural network,and a residual load probability density estimation model based on the improved adaptive Gaussian kernel density estimation are constructed,respectively.Therefore,a medium-and long-term industry load forecasting method considering multi-dimensional temporal features is formed.Finally,the load data of the chemical industry in a city of China is employed to verify the effectiveness of the proposed forecasting method.