计及电价和Attention机制的LSTM短期负荷预测模型

LSTM Short-term Load Forecasting Model Considering Electricity Price and Attention Mechanism

为了提高电力市场环境下短期负荷预测精度,利用互信息法和电价负荷曲线验证电价与负荷的关系,考虑电价对负荷预测的影响,从而提出一种基于Attention-LSTM(attention long short-term memory,Attention-LSTM)网络的短期负荷预测模型。首先将考虑电价波动因素的特征向量从输入层放入LSTM模型隐藏层中进行训练,然后将训练后得到的特征向量作为Attention层的输入,生成权重向量,最后将特征向量和权重向量合并得到新的向量,通过全连接层的训练得到预测结果值。运用江苏某地市数据进行实验验证,结果表明所提方法具有更高的负荷预测精度。

In order to improve the short-term load forecasting accuracy in the power market environment,the mutual information method and the electricity price load curve are used to verify the relationship between electricity price and load,and the influence of electricity price on load forecasting is considered,a short-term load forecasting model based on Attention-LSTM network was proposed.Firstly,the feature vector considering the fluctuation factor of electricity price is put into the LSTM model hidden layer from the input layer for training.Then the eigenvector obtained after training was used as the input of the Attention layer to generate the weight vector.Finally,the feature vector and the weight vector were combined to obtain a new one.The vector is obtained by the training of the fully connected layer to obtain the predicted result value.Experiments are carried out using data from a certain city in Jiangsu Province,the results show that the proposed method has higher load prediction accuracy.