Daily electricity price forecasting using artificial intelligence models in the Iranian electricity market.

The structure of the electricity market in Iran is based on a pay-as-bid auction mechanism. In such a market, hydropower generators need to have accurate estimates of energy price in peak hours of the day-ahead market to optimally operate the reservoir and maximize the revenue. This paper aims at providing a robust model with the best predictors for forecasting the maximum daily electricity price (MDEP) in Iran's electricity market. To reach the goal, hourly electricity prices in 2020 and 2021 were used and several artificial intelligence models were employed to predict the MDEP and ADEP (average daily electricity price). A sensitivity analysis of the inputs showed that in most of the models for forecasting the day-ahead MDEP (P t max), the best predictors were P t − 1 max , P t − 7 max , and P t − 30 m a x . Also, the convolutional neural-long short-term memory network (CNN-LSTM) had the best performance for forecasting both MDEP and ADEP in Iran's energy market. Compared to the multivariate linear regression model, the CNN-LSTM dealt well with sinusoidal characteristics and fluctuation of electricity prices. Therefore, it could improve the accuracy and correlation of the MDEP forecasts by 31% and 3.5%, respectively. • The maximum daily electricity price (MDEP) is very important for hydropower plant's operation. • Nonlinearity, instability, and hourly fluctuations of electricity price make it difficult to forecast. • Several artificial intelligent models, SVM, ANN, ANFIS, ANN-GA, and CNN-LSTM, were used for forecasting MDEP. • The best predictors (P t − 1 max , P t − 7 max , and P t − 30 max) were determined in a sensitivity analysis. • The CNN-LSTM model improved the accuracy of forecasts for MDEP by 31% in comparison with the conventional MLR model. [ABSTRACT FROM AUTHOR]