Random Forest-Based Approach for Maximum Power Point Tracking of Photovoltaic Systems Operating under Actual...

Area of Science:

• Renewable Energy Systems
• Machine Learning Applications in Power Electronics

Background:

• Photovoltaic (PV) energy generation relies on Maximum Power Point Tracking (MPPT) algorithms.
• Existing MPPT algorithms struggle with robustness due to dynamic environmental factors and steady-state accuracy limitations.

Purpose of the Study:

• To introduce a novel Random Forest (RF) model for enhanced MPPT performance in PV systems.
• To address the limitations of current MPPT techniques in terms of efficiency, accuracy, and dynamic response.

Main Methods:

• A Random Forest (RF) model was developed to capture complex, non-linear relationships between environmental variables (irradiance, temperature) and the maximum power point.
• A 3 kW peak capacity PV system comprising 25 SolarTIFSTF-120P6 modules was modeled and simulated in MATLAB/SIMULINK using 300,000 data samples.
• The RF-based MPPT tracker was implemented using two high-speed sensors and validated through 24 days of real-world environmental testing.

Main Results:

• The proposed RF-based MPPT demonstrated significant performance improvements compared to Artificial Neural Network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS) algorithms.
• The RF model achieved high accuracy and dynamic response, validated by real-world testing.
• The RF model successfully passed the Bland-Altman test with over 95% acceptability, indicating excellent agreement and reliability.

Conclusions:

• The novel RF-based MPPT approach offers a robust and accurate solution for maximizing PV energy generation.
• This method effectively handles dynamic environmental changes, outperforming traditional and other advanced MPPT techniques.
• The RF model presents a promising advancement for efficient and reliable solar energy harvesting.