Enhancing agent-based electricity purchase forecasting accuracy comprehensively supports power supply and electricity spot trading

Guided by the State Grid Corporation's 'Guidelines for Excellent Performance Management and Evaluation,' we apply the principles of excellent performance evaluation and lean management. Our main focus is on analyzing and forecasting the company's electricity sales market and power supply-demand balance, with the aim of enhancing the overall accuracy of the company's power purchase agency.On the basis of the LSTM model, this article adopts a method called Adaptive Noise Complete Set Empirical Mode Decomposition (CEEMDAN) to decompose the original power sequence into multiple intrinsic mode components and residuals. An improved CNN layer is added to further extract data features, which are then input into the improved LSTM model for prediction.By the end of December 2023, it is anticipated that the annual forecast error rate for residential agricultural power supply will be no higher than 4.8%, the forecast error rate for power purchase agency will be no higher than 5.1%, and the comprehensive error rate will be no higher than 5%, ranking among the top in the province.