Analysis of Inverter Efficiency Using Photovoltaic Power Generation Element Parameters

Area of Science:

• Renewable Energy Systems
• Artificial Intelligence in Energy
• Photovoltaic Performance Monitoring

Background:

• Photovoltaic (PV) power generation is susceptible to environmental variables and equipment condition.
• Maintaining optimal equipment performance, particularly inverters, is crucial for sustained energy production.
• Predictive assessment of equipment health is essential for efficient PV plant operation.

Purpose of the Study:

• To propose and validate a method for determining inverter efficiency degradation in PV systems.
• To utilize Long Short-Term Memory (LSTM) networks for predictive maintenance of PV equipment.
• To quantify the impact of operational duration on inverter efficiency.

Main Methods:

• Correlation and linear analysis were performed on PV power generation and environmental sensor data.
• A predictive model was trained using solar radiation and power data highly correlated with generation.
• The trained LSTM model was applied to analyze inverter performance data from 2020-2022.

Main Results:

• The predictive model achieved a Mean Absolute Percentage Error (MAPE) of 7.36, Root Mean Square Error (RMSE) of 27.91, Mean Absolute Error (MAE) of 18.43, and R-squared (R2) of 0.97.
• Statistical analysis revealed an average increase in error rate of 159.55W in 2022 compared to 2020.
• This indicates a 0.75% decrease in inverter efficiency over the three-year operational period.

Conclusions:

• The developed LSTM-based method is effective for analyzing inverter efficiency in operational PV plants.
• The findings demonstrate a quantifiable decrease in inverter efficiency with prolonged use.
• This predictive approach supports proactive maintenance strategies for photovoltaic systems.