Load Forecasting Considering Demand Response Mechanism

With the continuous access of generalized demand- side resources to the electricity market, in order to obtain better load forecasting results, it is necessary to accurately quantify the demand response factor. Based on this, an improved short-term load forecasting model combining Genetic Algorithm (GA) and Long Short-Time Memory neural network (LSTM) considering demand response is proposed. First of all, with the goal of maximizing Load Aggregation (LA) revenue, an optimization model is established for DR sub-signals such as interruptible load, transferable load and energy storage system according to the real-time electricity price, and the GUROBI optimization toolbox is called to solve each DR sub-signal. Then, the demand response signal is introduced into the historical load data of the implementation of LA, and the prediction models that take into account and ignore the demand response are established respectively. Finally, the effectiveness of the GA-LSTM method is proved by comparing the prediction results of ARIMA, BP, LSTM and GA-LSTM models. By taking into account the load forecasting results before and after the demand response, it is concluded that taking into account the demand response signals can improve the forecasting accuracy.