Data-driven User Behavior Prediction Based on Integrated Prediction Model

Abstract

Multi-station integration of various substations to participate in grid dispatching can fully mobilize load-side resources to participate in the power market, improve grid flexibility, and reduce grid peaking pressure, but the uncertainty of user behavior limits the development of demand response services. In response to this problem, this article first builds an incentive-based demand response implementation framework, expounds how load service entities gather demand-side resources to participate in the power market business, and convert the behavior of users responding to incentive policies to demand elasticity. Then, considering that the prediction model will ultimately need to be applied online, this paper proposes a data-driven incentive demand response user behavior prediction method that integrates LSTM based on the long short memory (LSTM) algorithm. At the same time, in order to improve the performance of the prediction model, The original data is smoothed and scaled, and the weight coefficient of the loss function is added. The results of the calculation examples show that, compared with the traditional LSTM algorithm and the k-neighbor prediction method, the prediction method proposed in this paper can significantly improve the prediction accuracy, and the preprocessing of the original data has positive significance for improving the prediction accuracy.