基于EWT-GRU-RR的配电网短期电力负荷预测模型

Short-term power load forecasting model based on EWT-GRU-RR

随着间歇分布式电源的大规模并入,电力负荷的波动性和非线性特征日益明显,现有单一预测模型较难实现精准预测。本研究提出一种经验小波变换(EWT)、门控循环单元(GRU)和岭回归(RR)相结合的短期电力负荷预测模型EWT-GRU-RR。首先,应用灰色关联度选取与负荷高相关性的气象耦合因素,作为相似日的分类指标;然后,采用皮尔逊系数法对类别内的负荷进行最佳相似日选取以减小计算规模;接着,采用EWT将相似日负荷数据分解得到不同频率的负荷模态序列;最后,采用GRU与RR分别对不同频率模态序列进行多步预测,并将预测分量叠加得到最终负荷预测结果。实验结果表明,本研究所提模型的预测误差较单一预测模型GRU减少了77%以上,较支持向量机回归(SVR)减少了75%以上,较先采用经验模态分解(EMD)进行分解再采用径向基函数神经网络(RBF)和RR组合预测模型EMD-RBF-RR减少了75%以上,较先采用EMD进行分解再采用GRU和RR组合预测模型EMD-GRU-RR减少了76%以上,有效提高了负荷预测精度。

With the large-scale integration of intermittent distributed power generation,the fluctuation and nonlinear characteristics of power load are becoming obvious,and it is difficult for the existing single prediction model to achieve accurate prediction.This paper proposed a short-term power load forecasting model based on the combination of empirical wavelet transform(EWT),gated recurrent unit(GRU)and ridge regression(RR).Firstly,the meteorological coupling factors highly correlated with load were selected by using grey correlation degree and were used as the classification index of similar days.To reduce the calculation scale,the best similar days for the loads within the category were selected by using the Pearson correlation coefficient method.Then,similar daily load data were decomposed by using EWT to obtain the load modal sequences of different frequencies.Finally,the multi-step predictions on modal sequences of different frequencies were performed respectively by using GRU and RR and the final load prediction result was obtained by superimposing the prediction components.The experimental results show that the forecasting error of the proposed model is reduced by more than 77%compared with single forecasting model GRU,more than 75%compared with support vector regression(SVR),more than 75%compared with the combined prediction model EMD-RBF-RR,in which the empirical mode decomposition(EMD)was first used to decompose and then the radial basis function(RBF)and RR were used to predict respectively,and more than 76%compared with the combined forecasting model EMD-GRU-RR,in which EMD was used to decompose and then GRU and RR were used to predict respectively.The proposed model improves the accuracy of load forecasting effectively.