Day-Ahead Spot Electricity Price Prediction Based on Hybrid Model

The spot electricity price is crucial for maintaining the supply-demand balance in the electricity market, and accurate electricity price prediction holds significant importance for electricity producers and large consumers. This article proposes a hybrid model based on FEDformer and LSTM, which utilizes the advantages of both to predict day-ahead spot electricity prices. FEDformer excels in capturing long-term dependencies, utilizing frequency domain analysis to extract long-term trends and seasonal features, which can more accurately model the long-term trend of electricity prices. LSTM can capture short-term dependencies in sequence data and is suitable for handling sudden fluctuations and short-term changes in electricity price data. We use real data from the electricity market of a certain province from 2020 to 2022 for model training and evaluation, which includes various factors that affect electricity prices, such as temperature, humidity, load, new energy, etc. We compared the model with other models and our own basic model, and the experimental results showed that the model outperformed other models in different evaluation indicators, providing a new method for predicting day-ahead spot electricity prices for participants in the electricity market.