Analysis and Accurate Prediction of User’s Response Behavior in Incentive-Based Demand Response

Abstract

Incentive-based demand response can fully mobilize a variety of demand-side resources to participate in the electricity market, but the uncertainty of user response behavior greatly limits the development of demand response services. This paper first constructed an implementation framework for incentive-based demand response and clarified how load-serving entity aggregates demand-side resources to participate in the power market business. Then, the characteristics of the user’s response behavior were analyzed; it is found that the user’s response behavior is variable, and it has a strong correlation on the timeline. Based on this, a prediction method of user response behavior based on long short-term memory (LSTM) is proposed after the analysis of the characteristics of the LSTM algorithm. The proposed prediction method was verified by simulation under the simulation environment setup by TensorFlow. The simulation results showed that, compared with the traditional linear or nonlinear regression methods, the proposed method can significantly improve the accuracy of the prediction. At the same time, it is verified by further experiments that the proposed algorithm has good performance in various environments and has strong robustness.