Researchers at the University of Lisbon in Portugal have developed a new machine learning algorithm that can predict failures in the inverter subsystems of utility-scale PV plants. The algorithm monitors the inverter subsystems and sends alarms when maximum and minimum values are reached, thereby providing early warnings of potential failures.
The algorithm works by analyzing data and categorizing variables based on historic values. It identifies different types of failures, including grid faults, overvoltage and undervoltage issues, AC overcurrent, grid power failure, excessive stray current, and more. By monitoring these variables and detecting abnormalities, the algorithm can predict when failures may occur.
To test the effectiveness of the algorithm, researchers applied it to two ground-mounted PV systems with capacities of 140 kW and 590 kW. Both systems used inverters from German manufacturer SMA. The algorithm successfully analyzed the variables and identified failures related to inverter errors.
The classification and prediction models used in the algorithm are based on tree-based models, which divide the feature space into smaller, non-overlapping regions with similar response values. This approach allows for the identification of seasonal variations in inverter failures, providing valuable insights for reliability analysis.
Additionally, the researchers suggest using clamp circuits to the resonant capacitance in parallel to automatically protect inverters from inrush and overcurrent. This can enhance power-conversion efficiency by regenerating the clamp current to the input voltage source.
Overall, the newly developed algorithm offers a data-driven evaluation of inverter subsystems and their failure modes. By accurately predicting failures and identifying variations in their occurrence, it can help utility-scale PV plants improve reliability and maintenance strategies.
The research conducted by the University of Lisbon highlights the potential of machine learning algorithms in the renewable energy sector. By harnessing the power of data analysis and predictive models, researchers and industry professionals can optimize the performance and efficiency of PV plants, ultimately contributing to the development of a more sustainable energy landscape.
