Hybrid CNN-Multivariate LSTM Model for Accurate Short-Term Electricity Price Forecasting and Energy System Optimization

The electricity market is continually changing for several reasons, including the growing popularity of renewable energy sources (RESs), the opening up of markets, and numerous economic and political considerations. Forecasting wholesale electricity prices is difficult due to these characteristics. Effective short-term forecasting is essential to keep the structure in equilibrium and handle abnormalities like price drops and predicted mistakes. Effective generation prediction is becoming increasingly important for effective power distribution and energy oversight as RESs grow more and more integrated into the electrical system. Due to the shortcomings of traditional forecasting techniques, algorithms using machine learning and deep learning have developed more and more well-liked because of their capacity to extract intricate associations from information and produce accurate forecasts. This work utilizes the Hybrid methods of Convolutional Neural Networks (CNNs) and Multivariate-Long Short-Term Memory (LSTM) networks. The hybrid model CNN-Multivariate LSTM is planned to gather both spatial dependencies and temporal patterns, which helps to analyze the trends and fluctuations in the energy economy. The proposed method improves the energy consumption or unexpected fluctuations in energy supply, efficiently optimizing the energy systems. The experimental results demonstrate exceptional accuracy in forecasting, by a Mean Absolute Percentage Error (MAPE) of 5.143%, especially in the combination CNN-Multivariate LSTM model. The results provide potential suggestions for additional studies in this sector.