Discrete Operational Data Analysis in Real Time Mode to Deduce the Catastrophic Failure Causes of a Charred Central Inverter

Abstract

The central inverter plays a significant role in large-scale PV plants. It is used to convert the DC power produced by many strings of the PV modules to AC power which is injected into the grid. So, it is considered the core of the PV plant. It affects distinctly the PV plant availability. So, the partial or total failures of inverters negatively affect PV plant availability. The inverter is designed to operate under specific conditions. Also, it is designed to produce AC power with technical specifications based on the input DC power characteristics. There are two types of data monitored during the inverter operation. The first type is the instantaneous data stored in the communication board memory. It is continuous time values that represent the waveforms of inverter outputs such as voltage, current, frequency, etcetera. Often, in the event of catastrophic failures that lead to the complete burning of the inverter, the instantaneous data is lost and can’t be restored. The second type is the operational data monitored and recorded by the SCADA system for each interval. It includes discrete RMS of inverter outputs such as voltage, current, frequency, etcetera. Most SCADA systems are designed to record data per minute. Hereby, the operational data is considered the sole data source in the event of catastrophic failures of the PV inverter. This paper presents an analytical model of the inverter operational data before fault occurrence. The objective of analyzing the inverter operational data is to deduce the thread-tip observation that indicates the failure causes of the grid-tie PV inverter.