Deep belief rule based photovoltaic power forecasting method with interpretability.

Area of Science:

• Renewable Energy Systems
• Artificial Intelligence in Power Engineering
• Predictive Modeling

Background:

• Accurate photovoltaic (PV) power prediction is crucial for power grid management.
• Existing belief rule base (BRB) models struggle with rule explosion and result inexplicability in complex systems.
• The uncertainty inherent in PV generation poses challenges for reliable power grid operations.

Purpose of the Study:

• To propose a novel Deep Belief Rule Base with Interpretability (DBRB-I) model for enhanced PV output power prediction.
• To address the limitations of traditional BRB models, specifically rule explosion and lack of interpretability.
• To improve the accuracy and clarity of PV power generation predictions for better grid integration.

Main Methods:

• Development of a deep BRB structure to mitigate rule explosion.
• Implementation of sensitivity analysis for simplifying inefficient rules and reducing model complexity.
• Design of a novel optimization method using the Projection Covariance Matrix Adaptation Evolution Strategy (P-CMA-ES) algorithm to maintain model interpretability.

Main Results:

• The proposed DBRB-I model effectively predicts PV output power.
• The deep structure and rule simplification successfully address the rule explosion problem.
• The P-CMA-ES optimization ensures the interpretability of the prediction model is preserved.

Conclusions:

• The DBRB-I model offers an effective solution for accurate and interpretable PV power prediction.
• The method enhances the reliability of PV power generation forecasting for power grids.
• This approach contributes to more efficient scheduling and development management of power grids.