Advancing solar energy integration: Unveiling XAI insights for enhanced power system management and sustainable future.

Solar energy has emerged as a vital renewable alternative to fossil fuels, enhancing environmental sustainability in response to the pressing need to reduce carbon emissions. However, the integration of solar power into the electrical grid faces challenges due to its unpredictable nature, as a result of solar energy production variability. This research presents an advanced Explainable Artificial Intelligence (XAI) framework to explicate machine learning models decision-making processes, thereby improving the predictability and management of solar energy distribution. The influence of critical parameters such as solar irradiance, module temperature, and ambient temperature on energy yield is studied using the Local Interpretable Model-Agnostic Explainer (LIME). Rigorous testing using four advanced regression models identified Random Forest Regressor as the superior model, with an R² score of 0.9999 and a low Root Mean Square Error (RMSE) of 0.0061. Furthermore, Partial Dependency Plots (PDP) are used to emphasize the intricate dependencies and interactions among features in the dataset. The application of XAI techniques for solar power generation extends beyond explainability, addressing challenges due to various parameters in solar radiation pattern analysis, error estimation in solar performance, degradation of the battery function, and also provides interpretable insights for enhancing the lifespan of solar panels, contributing to advancements in sustainable energy technologies. The results of this study show how XAI has the potential to transform power system management (PSM) and strategic planning, propelling us toward a future of energy that is more resilient, efficient, and environmentally friendly. • The Purpose of the paper is to provide an Explainability to the productivity of solar energy to the end-user and to perform solar energy predictions. • This work performs regression analysis of various parameters related to solar power distribution and picks the model that is more reliable and delivers the maximum performance which is the Random Forest Regressor. • The selected model is applied with the Explainable Artificial intelligence with tools like Local Interpretable Model Agonistic Explainer (LIME) and the Partial Dependency plots (PDP). • They not only provide the prediction but also provide the feature importance, feature weights, and what features provide positive and negative impact towards the target prediction. • The future work includes the application of federated learning for Solar power generation and distribution prediction. [ABSTRACT FROM AUTHOR]