Multi-task learning based deep neural network for short term power and thermal energy prediction of hybrid PV–Thermal systems

Abstract

Accurate forecasting of power output and thermal energy in hybrid Photovoltaic–Thermal (PVT) plants is critical for optimising energy management and maximising efficiency. Traditional models often struggle to capture complex temporal dependencies and non-linear patterns in such systems. We propose the Inception Embedded Dual Attention Bi-LSTM (IDAMBi-LSTM) Model for enhanced power forecasting and thermal energy prediction in hybrid PVT plants to address these challenges. The model integrates inception layers to capture multi-scale temporal features, dual attention mechanisms to emphasise significant data points, and Bidirectional Long Short-Term Memory (Bi-LSTM) networks for handling sequential dependencies in both forward and backward directions. We conducted a comparative analysis with an Inception Embedded Attention Mechanism with GRU (IAM-GRU) and Inception Embedded Dual Attention GRU (IDAM-GRU) model to evaluate the performance and accuracy of our proposed approach. The evaluation was based on several performance metrics, including Root Mean Square Error (RMSE), R-squared (R2), Normalised Mean Square Error (NMSE), Mean Absolute Percentage Error (MAPE), and Mean Absolute Error (MAE). The results demonstrate that the proposed Inception Embedded Dual Attention Bi-LSTM Model outperforms the IAM-GRU model across all metrics, providing more accurate and reliable forecasts for both power output and thermal energy in hybrid PVT systems. This research highlights the potential of advanced deep learning techniques in improving the efficiency and predictability of renewable energy systems.