基于正交小波和长短期记忆神经网络的用电负荷预测方法

用电负荷数据的波动性和周期性会影响电力负荷预测的准确性，针对此问题，文中提出了一种基于正交小波长短期记忆神经网络(orthogonal wavelet transform-long short-term memory, OWT-LSTM）的用电负荷预测方法。该方法对用电负荷序列进行正交小波分解，消除负荷数据的波动性，然后利用LSTM及其变种神经网络对正交小波分解后的各尺度负荷序列进行建模训练，通过各序列预测结果进行预测重构，获得最终的负荷预测结果。通过用户用电负荷数据集验证表明，该方法的预测性能优于其他模型，具有较高的预测精确性和稳定性。

Electricity Load Forecasting Method based on Orthogonal Wavelet and Long Short-term Memory Neural Networks

Due to the fact that the volatility and periodicity generated by electrical load data affect the accuracy of power load forecasting, a electric load forecasting method based on Orthogonal Wavelet Transform-Long Short-Term Memory (abbr. OWT-LSTM) was proposed. Firstly, the orthogonal wavelet transform of electric load series was performed to eliminate the volatility of load data, then the LSTM neural network was utilized to conduct the modeling and training of all scales load series after orthogonal wavelet decomposition, and through the forecasting results of all series the forecasting results the prediction reconstruction was conducted to obtain final load forecasting results. Experimental results show that through the verification by consumers’ load data set, the forecasting performance of the proposed method is evidently better than other models, and it possesses better forecasting accuracy an stability