Electricity Price Gaming Based on Personalized Load Forecasting for Energy Saving

Jie Zou, Siya Xu, Cheng Zhou, Hai Wu, Zeng Zeng, Peng Yu

Abstract: With the development of smart grids and the increasing demand for electricity in society, the stability of the grids and energy saving have gradually become a key concern. Time-of-use (TOU) price is one of the ways for grids to balance power supply and demand. However, it is difficult to improve system stability and utility through TOU pricing due to the volatility of electricity load and the complexity caused by multi-party participation. To this end, we propose an electricity price gaming method based on personalized load forecasting for energy saving. First, we design a personalized federated learning framework for load forecasting based on collaborative domains to address the issues of data heterogeneity and resource heterogeneity. Further, on the basis of load forecasting, we establish a multi-party game model based on Stackelberg game. Then, we propose a joint optimization mechanism and equilibrium solving algorithm to obtain the optimal TOU price. The experimental results show that the method proposed in this paper outperforms other benchmarks in terms of forecasting accuracy, system utility and stability.