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

Area of Science:

• Renewable Energy Systems
• Artificial Intelligence
• Expert Systems

Background:

• Accurate photovoltaic (PV) power prediction is essential for grid stability.
• Belief rule base (BRB) systems excel at modeling nonlinear relationships but require significant expert knowledge for initial model construction.
• Existing methods face challenges in acquiring sufficient expert input for effective BRB parameterization.

Purpose of the Study:

• To propose a novel data-driven interval construction belief rule base (DD-IBRB) method for PV power prediction.
• To reduce the dependency on expert knowledge in constructing and initializing BRB models.
• To enhance the accuracy and efficiency of PV power forecasting.

Main Methods:

• Utilized a fuzzy clustering algorithm to establish reference intervals for data-driven model construction.
• Implemented a Gaussian membership interval function (GIBM) for initializing belief degrees.
• Employed a representative point selection mechanism and evidential reasoning (ER) rules for model inference.
• Optimized the DD-IBRB model using a multi-population evolution animated oat optimization (MEAOO) algorithm with parameter constraints.

Main Results:

• The proposed DD-IBRB method demonstrated effective modeling capabilities for PV power prediction.
• Achieved a low mean squared error of 0.00056 in the PV power output case study.
• The data-driven approach significantly reduced the need for extensive expert knowledge.

Conclusions:

• The DD-IBRB method offers an accurate and efficient approach to PV power prediction.
• This data-driven strategy overcomes the limitations of expert knowledge dependency in BRB modeling.
• The findings support the application of DD-IBRB for reliable grid integration of renewable energy sources.