Market-oriented electricity price prediction method based on CNN-BiLSTM multi-feature fusion

Abstract

In recent years, with the continuous development of global electricity market reforms, electricity has become a freely tradable commodity. Due to the non-storage characteristics of this special commodity, its price will be affected by many complex factors and change in real time, such as time, load, weather and other factors. It is precisely because of this particularity that the electricity price reflects the operating conditions of the electricity market and is a core indicator for evaluating the efficiency of market competition. Therefore, effective forecasting of electricity prices is necessary. Due to the development of computing power and large-scale data storage technology, machine learning has a significant effect on processing time series tasks with nonlinearity and volatility aggregation. This paper proposes a real-time market-based electricity price forecasting method based on CNN-BiLSTM multi-feature fusion, mining the factors that affect electricity price changes, and incorporating external factors such as weather into the forecasting model to effectively improve the forecasting accuracy. Our proposed method has been experimentally evaluated on the Spanish electricity data set, and the experimental results show that the proposed method has a good performance.