Holistic Rooftop Solar Energy Management Using Machine Learning Models – A Study of Sai Mitra
The study involves the holistic management of more than 60 rooftop solar energy facilities installed in Prasanthinilayam as part of ‘Sai Mitra’ project. using sophisticated machine learning models
The analysis in this study was designed to explore the relationships between environmental factors, such as temperature and humidity, seasonal variations, and the Capacity Utilisation Factor of solar facilities. A combination of regression, non-linear modelling, and mixed-effects approaches was used to address the research objectives, ensuring that the analyses aligned with the nature of the data.
Research methods
ML methods used:
- Multivariate linear regression
- Cluster analysis
- ARIMA models
- Decision tree and random forest analysis
- Two stage DEA-OLS analysis
Findings and recommendations
- Operationally, solar facility managers should integrate short-term CUF forecasting models, such as ARIMA and Random Forest, into their daily workflows to proactively adjust for weather-induced variability in performance.
- Facilities demonstrating high sensitivity to thermal conditions, would benefit from deploying adaptive infrastructure such as panel cooling systems or solar tracking mechanisms.
- Aligning routine maintenance schedules with seasonal CUF patterns, especially ahead of the monsoon season, can help mitigate efficiency losses due to precipitation and panel soiling.
- Implementation of IoT-based systems for maintenance optimization
- Dynamic reconfiguration of photovoltaic arrays to minimize mismatch losses
- Fault detection employing artificial neural networks