A novel hybrid GRU-XGBoost model for day-ahead photovoltaic generation forecasting in microgrids
Year: 2025
Type: article
Abstract: The global transition toward renewable energy has made photovoltaic (PV) technology a key component in sustainable microgrid operations. However, the inherent variability of PV generation (PVGen), influenced by changing weather conditions, presents challenges for efficient energy management. This study introduces a novel hybrid forecasting model, combining Gated Recurrent Units (GRU) and Extreme Gradient Boosting (XGBoost), to address the challenges of PVGen forecasting in microgrids. The model operates in two stages: GRU captures temporal patterns in historical weather and PV data, and XGBoost refines the forecasts. Particle Swarm Optimization (PSO) is used to fine-tune hyperparameters in both stages, ensuring optimal performance. The model was evaluated under various weather conditions using RMSE, MAE, nRMSE, and R2 as performance metrics. The model achieved a notably low nRMSE of 4.17%, outperforming existing models such as GRU-Informer-SVR (33.9% improvement) and LSTM-Informer (54.5% improvement). Under sunny conditions, the model achieved an nRMSE of 2.67% (38.3% lower than standalone GRU) and an R2 of 0.9956. On overcast days, the model maintained robust performance, achieving an nRMSE of 5.12% (63.9% improvement over GRU) and R2 of 0.9670. The proposed GRU–XGBoost hybrid model significantly advances the field of day-ahead PV generation forecasting, offering a reliable, adaptable, and efficient solution for microgrid applications. Its robust performance under fluctuating weather conditions highlights its potential as a critical tool for optimizing PVGen management in real-world energy systems.
Source: Scientific African
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $590
OpenAlex ID: W4413258798