Transformer short-term load forecasting method considering demand response potential

Abstract

The uncertainty of power load is one of the important research directions in demand response uncertainty. Accurate and effective power system load forecasting is an important prerequisite for ensuring the safety, stable operation, and normal production of the power grid. To improve the accuracy of short-term load forecasting in power systems under demand response scenarios, this paper proposes a Transformer load forecasting method that considers demand response potential. Firstly, the change law of response uncertainty with electricity price difference and consumer psychology principles are used to quantify the power demand response results under different probability conditions. Then, Transformer neural networks are used to extract features from user historical load, temperature, electricity price, and other time series data. Finally, a multi-head self-attention mechanism is used to pay attention to the structural relationship between time series data, analyze the importance of input variables at each historical moment on the current load, and achieve high-precision prediction of user load and demand response potential. This article takes industrial users as an example to predict the power load and demand response regulation power of the general component manufacturing industry. Through comparative analysis with actual data, the effectiveness of the proposed method is verified. Compared with other existing methods, the Transformer model that considers demand response performs well in power load forecasting, providing a certain theoretical basis for evaluating the potential of demand response. The subsequent work will study the characteristics of electric, hot, and cold loads and their coupling relationships under the difference of electricity prices, and improve the forecasting performance of user loads and Demand Response Regulation power, so as to reduce the power generation and operation costs of the grid.