基于小波包分解与循环神经网络的综合能源系统短期负荷预测

Short-Term Load Forecast of Integrated Energy System Based on Wavelet Packet Decomposition and Recurrent Neural Network

文章提出了基于小波包分解(wavelet packet decomposition,WPD)与循环神经网络的电冷热综合能源短期负荷预测方法。利用能够突出负荷细节特征的小波包对电冷热负荷进行频段分解,分析每一频段中电冷热负荷的互相关性。为体现每一频段中电冷热负荷的互相关性对预测结果的影响,将频段中互相关性较强的负荷类型放入同一处理负荷自相关性的循环神经网络模型中进行预测;频段中互相关性较弱的负荷类型则单独进行预测。与直接将电冷热负荷放入同一个循环神经网络进行预测相比,以及与将电冷热负荷通过同一个反向传播神经网络进行预测相比,所提方法考虑了综合能源在不同频段内电冷热负荷的互相关性和电冷热负荷本身的自相关性,能够有效降低负荷预测的平均绝对百分比误差。

This paper proposes a short-term load forecast method for electric,cooling and heating loads on the basis of wavelet packet decomposition(WPD)and recurrent neural network.Wavelet packets that can highlight the detail characteristics of the load are used to decompose electric,cooling and heating loads,and analyze the cross-correlation between the electric,cooling and heating loads in each frequency band.In order to reflect the influence of the cross-correlation of electric,cooling and heating loads in different frequency bands on the forecasting results,those loads with strong cross-correlation in a frequency band are put into the same recurrent neural network model that can handle the autocorrelation of the load;while those with weak cross-correlation in a frequency band are forecasted separately.Compared with directly placing the electric,cooling and heating loads into the same recurrent neural network for forecasting,and compared with placing the electric,cooling and heating loads into the same back propagation neural network for forecasting,the method in this paper considers the cross-correlation of the electric,cooling and heating loads in each frequency band and the autocorrelation of the electric,cooling and heating loads,which effectively reduce the mean absolute percentage error of load forecast.