BETTER FAULT DETECTION AND DIAGNOSIS WITH ARTIFICIAL INTELLIGENCE: METHODS, EXAMPLES AND BUSINESS CASES

摘要

Based on a sample of plants monitored by 3E, we see more than 20% of professional photovoltaic plants performing more than 10% below expectation. The main cause for these cases of underperformance are lack of time and experience. Especially not so obvious flaws and failures often persist for weeks to months before they are noticed and fixed eventually. This paper shows how artificial intelligence can be used for PV operations and maintenance to detect faults automatically and formulate recommendations on their most probable root causes. It shows several examples, namely, performance and loss analysis with limit checking and degradation analysis with trend checking. In conclusion, with the methods presented here, PV plant operators and asset managers can identify faults early and draw conclusions on the underlying root causes. By estimating the losses and comparing them to the losses as expected from a simplified model, we can associate them with the different failure events and compute the gains to be expected when the fault had been detected and corrected early. 3E is currently offering the methods presented here under the name PV Health Scan.