A short and medium term electricity spot price prediction method based on extreme learning machine algorithm

Electricity, unlike other commodities, cannot be easily stored, leading to inherent uncertainty in electricity spot prices. In highly competitive markets, forecasting models for electricity spot prices have become crucial tools for companies operating in the electricity sector. To further enhance the decision-making capabilities of such companies in electricity procurement, this paper proposes a model based on electricity spot price forecasting.Firstly, an artificial neural network trained with the Extreme Learning Machine algorithm is utilized to determine the monthly average spot prices for the next six months. Secondly, a reference function is incorporated into the algorithm to return spot prices associated with a given level of risk, thereby mitigating spot market volatility. The results indicate that the proposed model achieves a root mean square error（RMSE） lower than 10% of the values associated with futures prices, demonstrating its potential to improve mid-term to short-term electricity trading decisions and reduce decision-making risks.