An Electricity Load Forecasting Approach Combining DBN-Based Deep Neural Network and NAR Model for the Integrated Energy Systems

In order to improve the accuracy of electricity load forecasting in the integrated energy systems (IES), an improved electricity load forecasting method based on the NAR dynamic neural network and the deep belief network (DBN) is proposed. Firstly, the calculation method of the net load is determined for the typical topology of the IES. Based on the historical net load data, the NAR dynamic neural network model is established to forecast the time series components of the net load. Then the results of DBN network pre-training are used to initialize the deep neural network (DNN) and fine-tune the parameters with back propagation (BP) algorithm to forecast the volatility components of the net load. Finally, the predicted value of the net load is output in combination with the two-part forecasting result. The case study is given to verify the effectiveness of the proposed model.