The expected service life of photovoltaic systems and their components significantly determines their economic efficiency and is thus a critical factor for their use. Most photovoltaic system failures are caused by inverter failures. High-power inverters connect large PV plants to the grid, while lower-power inverters connect households with small systems. The sudden failure of a high-power inverter usually causes significant disruptions in the connected distribution grid, e.g., due to frequency disturbances.
If it can be detected in advance that a failure is imminent from the data that an inverter records and constantly feeds into the manufacturers database (Internet of Things), the inverter can be repaired or replaced before the actual failure occurs.
By providing failure forecasts for high-performance inverters, the inverter manufacturer can offer the distribution grid operator an additional service to guarantee stable supply security in power grids that are heavily penetrated by PV.
The VORAUS PV project investigates whether and how modern data analysis methods (machine learning) can be used to detect and predict failures of photovoltaic inverters.
The goal of this project is to develop and evaluate algorithms that are able to predict individual failure events in the future based on the existing data of the project partner and inverter manufacturer SMA. The VORAUS PV project enables for the first time a systematic evaluation of a very large amount of non-standardized data from the production and several decades of operation of different photovoltaic inverter generations in worldwide use.
Machine Learning, Big Data, Predictive Maintenance, Condition Monitoring.
J Wibbeke, P Teimourzadeh Baboli, S Rohjans; Energies ; 0Jan. / 2022