Photovoltaic Module Degradation Detection Using V-P Curve Derivatives and LSTM-Based Classification

Area of Science:

• Renewable Energy Systems
• Artificial Intelligence in Engineering
• Materials Science for Energy

Background:

• Photovoltaic (PV) systems are crucial for sustainable energy but suffer performance degradation due to environmental exposure.
• Current diagnostic methods for solar module aging lack real-time monitoring, quantitative assessment, and scalability for large power plants.
• Degradation leads to power loss and operational issues, necessitating advanced detection techniques.

Purpose of the Study:

• To develop a novel, real-time method for detecting and quantifying degradation in solar modules.
• To enable early identification of the number and severity of degraded solar modules within a string.
• To overcome the limitations of existing PV diagnostic tools.

Main Methods:

• Utilizing the first-order derivative of the voltage-power curve to extract key degradation features.
• Developing an AI model based on long short-term memory (LSTM) for classifying normal/abnormal states and predicting aging.
• Designing a shallow LSTM network optimized for PV time-series data to prevent overfitting and gradient vanishing.

Main Results:

• The proposed method effectively extracts features indicative of solar module degradation.
• The LSTM model accurately classifies system states and predicts aging status, demonstrating learning and convergence.
• MATLAB simulations validated the model's effectiveness and stability in training.

Conclusions:

• The novel voltage-power curve derivative method combined with LSTM AI offers a robust solution for solar module degradation detection.
• This approach facilitates early and accurate diagnosis of PV system health, improving operational efficiency.
• The study provides a foundation for advanced, real-time monitoring systems for photovoltaic power plants.