Machine learning for monitoring and classification in inverters from solar photovoltaic energy plants
Fabíola Pereira, Carlos Silva
Abstract
The efficiency of solar energy farms requires detailed analytics and information on each inverter regarding voltage, current, temperature, and power. Monitoring inverters from a solar energy farm was shown to minimize the cost of maintenance, increase production and help optimize the performance of the inverters under various conditions. Machine learning algorithms are techniques to analyze data, classify and predict variables according to historic values and combination of different variables. The 140 kWp photovoltaic plant contains 300 modules of 255W and 294 modules of 250W with smart monitoring devices. In total the inverters are of type SMA Tripower of 25 kW and 10 kW. The 590 kWp photovoltaic plant contains 1312 Trina solar 450W modules. In total the four inverters are SMA Sunny Tripower type of 110-60 CORE 2 with rated power of 440kW were analyzed and several supervised learning algorithms were applied, and the accuracy was determined. The facility enables networked data and a machine learning algorithm for fault classification and monitoring was developed, energy efficiency was calculated and solutions to increase energy production and monitoring were developed for better reliability of components according to the monitorization and optimization of inverters.