A Photovoltaic Power Prediction Method Based on Data-Driven Interval Construction Belief Rule Base

Accurate prediction of photovoltaic (PV) power is crucial for ensuring grid stability. The belief rule base (BRB) is a rule-based expert system capable of effectively handling nonlinear causal relationships. Therefore, it can be applied to PV power prediction. In practical prediction scenarios, a high-quality initial model can produce more accurate predictions. However, obtaining sufficient expert knowledge to determine the structure and parameters of the BRB is usually difficult. To address this issue, a PV power prediction method is proposed based on a data-driven interval construction belief rule base (DD-IBRB), which reduces the reliance on expert knowledge during model construction. First, a fuzzy clustering algorithm is designed to construct reference intervals. Then, a Gaussian membership interval function (GIBM) strategy is proposed to initialize the belief degrees. Next, a representative point selection mechanism is designed within the reference intervals. Model inference is subsequently performed based on evidential reasoning (ER) rules. Finally, a multi-population evolution animated oat optimization with parameter constraints (MEAOO) is used to optimize the DD-IBRB model. Taking the PV power output as a case study, the mean squared error is 0.00056, indicating that the proposed DD-IBRB method can effectively complete modeling and obtain accurate prediction results.