Neural Network Algorithm With Reinforcement Learning for Parameters Extraction of Photovoltaic Models

Area of Science:

• Electrical Engineering
• Computational Intelligence
• Renewable Energy Systems

Background:

• Accurate parameter extraction for photovoltaic (PV) models is essential for efficient PV system control and optimization.
• Existing algorithms face challenges in achieving reliable and precise PV model parameter extraction.
• The neural network algorithm (NNA), inspired by artificial neural networks (ANNs), offers strong global search capabilities but suffers from slow convergence and local optima stagnation.

Purpose of the Study:

• To develop an improved algorithm for accurate and reliable parameter extraction of photovoltaic (PV) models.
• To address the limitations of the standard Neural Network Algorithm (NNA), specifically its slow convergence and tendency for local optima stagnation.
• To enhance the performance of metaheuristic algorithms in the context of PV model parameter identification.

Main Methods:

• Introduction of the Reinforcement Learning Neural Network Algorithm (RLNNA), an enhanced version of NNA.
• Integration of three novel strategies within RLNNA: modification factor with reinforcement learning (RL), a transfer operator utilizing historical population data, and a feedback operator.
• Application of RLNNA to extract parameters for three distinct PV models.

Main Results:

• RLNNA demonstrated higher accuracy in parameter extraction compared to the standard NNA.
• RLNNA exhibited stronger stability in parameter extraction across different PV models.
• The proposed RLNNA outperformed 12 other established algorithms in PV model parameter extraction tasks.

Conclusions:

• The RLNNA effectively overcomes the limitations of the standard NNA, achieving superior accuracy and stability.
• The integrated strategies (RL, transfer operator, feedback operator) significantly enhance the performance of the Neural Network Algorithm.
• RLNNA represents a promising advancement for precise parameter extraction in photovoltaic modeling, contributing to improved PV system performance.