Optimizing power loss in mesh distribution systems: Gaussian Regression Learner Machine learning-based solar irradiance prediction and distributed generation enhancement with Mono/Bifacial PV modules using Grey Wolf Optimization

This work implements a solar-powered Distribution Generation (DG) system using mono-facial and bifacial PV modules within mesh distribution networks to minimize power loss. Machine learning techniques, such as Gaussian Regression Learner, Linear Regression, and Artificial Neural Networks, predict Global Horizontal Irradiance (GHI) for efficient solar energy utilization. The models are evaluated using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Coefficient of Determination, with the Gaussian Regression Learner showing the highest accuracy. A unique aspect of this work is integrating both mono-facial and bifacial PV modules. Bifacial modules, which capture sunlight from both sides, generate more energy. The novelty lies in using Gaussian Regression Learner for GHI prediction and generating solar power with bifacial modules, then feeding it into the mesh distribution system. Integration into IEEE-33 and IEEE-69 mesh distribution systems focuses on optimizing DG unit placements using the Grey Wolf Optimization (GWO) technique, known for its simplicity and effectiveness. Post-integration analysis shows significant reductions in active power loss: mono-facial PV modules reduce losses by 32.62% and 23.77% for IEEE-33 and IEEE-69 systems, respectively, while bifacial modules achieve reductions of 49.50% and 40.56%.