With the development of China's electricity spot market and proxy purchasing, sharp price fluctuations due to supply-demand uncertainties pose key risks. To improve forecasting accuracy for purchasing decisions, this study examines spot price prediction under such uncertainties. It analyzes market clearing mechanisms and identifies supply-side uncertainties (renewable output, conventional generation costs) and demand-side uncertainties (load elasticity, economic cycles). A multi-dimensional factor system covering meteorological, economic, systemic, and market aspects is established. Integrating the time-series advantages of LSTM with the uncertainty-handling capability of robust optimization, a hybrid forecasting model balancing accuracy and adaptability is proposed. Empirical tests using Shandong's spot market data show the model's MSE, MAE, and MAPE decrease to 286.32, 14.25, and 1.73%, respectively, outperforming traditional LSTM and random forest models. The model effectively addresses forecasting challenges under uncertainties and supports cost control and risk management in proxy purchasing.
Zhang et al. (Mon,) studied this question.