LSTM Short-Term Load-Prediction Model for IES Including Electricity Price and Attention Mechanism

Abstract

In order to improve the accuracy of short-term load prediction in an integrated energy system, the data of a comprehensive energy demonstration area is analyzed. In the case of similar time-sharing gas prices, the relationship between electricity price and load is verified by using the electricity price and load curve, and the electricity price and gas curve. The influence of electricity price on load prediction is considered. A short term Load Prediction model based on attention-LSTM (Attention Long short term Memory) network is proposed. First, the eigenvectors considering price fluctuations are trained from the input layer to the hidden layer of the LSTM model. The trained eigenvectors are then used as input to the Attention layer to generate the weight vectors. Finally, the eigenvectors and weight vectors are combined to get new vectors, and the predicted values are obtained by training the fully connected layers. The experimental results show that the proposed method has higher load prediction accuracy.