Short-Circuit Incipient Faults Detection from Single Phase PWM Inverter using Artificial Neural Network

Solar Photovoltaic (PV) system consists of three main parts which are PV array, PV inverter and utility grid. From the three, PV inverter is considered the weakest link in the solar PV system. Insulated Gate Bipolar Transistor (IGBT) is the most critical component in an inverter and is often blamed for the failure of inverters. If the incipient faults of the IGBT can be detected, the breakdown possibility of the solar PV system can be improved. However, before the incipient faults can be detected, it needs to be first generated before further analysis and improvements can be made. This paper proposes a process on how to generate the incipient faults which are caused by the short-circuit fault of a single phase PWM inverter. The single phase PWM inverter consists of four IGBTs and there is a total of six parameters that need to be observed for each IGBT. The parameter that could cause short-circuit fault to the IGBTs is identified by modifying the parameters of IGBT one at a time. The response at the inverter output is observed and recorded after each modification to the parameter value is done. From the results, it shows that parameter Threshold voltage (V_ge(th)) is identified to be able to generate the short-circuit incipient faults. For the application of detection the incipient faults using neural network, a total of 100 short-circuit incipient faults and one set of normal condition waveform are collected at the output of the single phase PWM inverter. These waveforms are then used to train the feedforward backpropagation neural network. One hidden layer feedforward backpropagation neural network of 7 neurons was trained and MSE of was obtained. It was shown that the trained feedforward backpropagation neural network was able to detect which IGBT component of the single phase PWM inverter produced the short-circuit incipient faults.