Wind and photovoltaic power prediction using a combined VMD-CNN-BiLSTM model

Abstract The swift advancement of renewable energy necessitates precise forecasting of wind and photovoltaic power to ensure grid stability and effective energy management. Nonetheless, the data associated with wind and solar power is characterized by non-stationarity, noise interference, and complex temporal dynamics, which hinder the effectiveness of conventional forecasting techniques. To enhance prediction accuracy, this study integrates the effectiveness signal decomposition and denoising capabilities of Variational Mode Decomposition (VMD), the robust feature extraction potential of Convolutional Neural Networks (CNN), and the strong temporal memory capacity of Bidirectional Long Short-Term Memory networks (BiLSTM) to develop a high-accuracy combined VMD-CNN-BiLSTM forecasting model for wind and solar energy.