A novel reflective intelligence optimizer with machine learning (RIO-ML) for parameter estimation of photovoltaic...

Area of Science:

• Renewable Energy Engineering
• Computational Intelligence
• Machine Learning Applications

Background:

• Solar photovoltaic (PV) equivalent circuit models present complex nonlinear and multimodal parameter estimation challenges.
• Conventional optimizers often struggle with the high dimensionality and multimodality of PV models.
• Accurate PV parameter estimation is crucial for efficient solar energy system design and performance analysis.

Purpose of the Study:

• To introduce a novel Reflective Intelligence Optimizer with Machine Learning (RIO-ML) for accurate PV parameter estimation.
• To address the limitations of traditional optimizers in handling nonlinear and multimodal PV models.
• To enhance the robustness and convergence speed of parameter estimation algorithms for solar cells.

Main Methods:

• Development of the RIO-ML approach, integrating a multi-leader social learning algorithm with reflective memory.
• Utilization of machine learning, specifically Multi-Layer Perceptron models, for adaptive parameter control.
• Implementation of progressive Gaussian refinement with reflective boundary treatment for improved optimization.
• Testing RIO-ML on Single Diode Model (SDM), Double Diode Model (DDM), and Triple Diode Model (TDM) of the RTC France solar cell.

Main Results:

• RIO-ML achieved minimum RMSE values of 8.739710×10^-4 A (SDM), 8.456760×10^-4 A (DDM), and 7.7546980×10^-4 A (TDM).
• Mean RMSE values were consistently low across all models, indicating high precision.
• Comparative analysis demonstrated RIO-ML's superior performance over existing metaheuristic algorithms in solution quality, convergence rate, and robustness.
• Maximum absolute current errors were maintained below 1.6×10^-3 A for all tested PV models.

Conclusions:

• The RIO-ML approach is a highly effective and efficient tool for the accurate estimation of solar PV model parameters.
• The integration of reflective intelligence and machine learning significantly enhances optimization performance for complex PV models.
• RIO-ML provides a robust solution for challenges in solar cell parameter identification, contributing to advancements in solar energy research.