Explainable AI and optimized solar power generation forecasting model based on environmental conditions.

Area of Science:

• Renewable Energy Systems
• Artificial Intelligence in Energy
• Machine Learning for Power Forecasting

Background:

• Accurate solar power forecasting is crucial for smart grid stability and efficient energy management.
• Existing forecasting models often lack interpretability and robust hyper-parameter optimization.
• Integrating explainable AI (XAI) and advanced optimization techniques can enhance solar power prediction reliability.

Purpose of the Study:

• To propose and evaluate the novel X-LSTM-EO model for reliable solar power generation forecasting.
• To leverage explainable artificial intelligence (XAI) to identify key factors influencing forecast accuracy.
• To optimize long short-term memory (LSTM) model hyperparameters using the Equilibrium Optimizer (EO) for enhanced performance.

Main Methods:

• Developed the X-LSTM-EO model combining LSTM for time-series forecasting, EO for hyper-parameter optimization, and Local Interpretable Model-agnostic Explanations (LIME) for interpretability.
• Evaluated model performance using R-squared (R2), RMSE, COV, MAE, and EC metrics.
• Compared the proposed model against conventional LSTM, Decision Tree, Linear Regression, and Gradient Boosting algorithms, with and without the EO optimizer (using PSO as a benchmark).

Main Results:

• The X-LSTM-EO model achieved high performance with R2=0.99, RMSE=0.46, COV=0.35, MAE=0.229, and EC=0.95.
• Significant performance improvements were observed compared to conventional LSTM, with rates reaching 148% for R2 and 134% for EC.
• The LSTM model outperformed Decision Tree and Linear Regression, and the EO optimizer demonstrated superior efficacy compared to PSO.

Conclusions:

• The proposed X-LSTM-EO model provides a reliable and interpretable approach for forecasting solar power generation, even with abrupt changes.
• The integration of XAI and EO significantly enhances the accuracy and robustness of solar energy predictions.
• This model can aid in optimizing the operational efficiency of photovoltaic power systems.