Single-phase photovoltaic inverter on-line state monitoring and residual life prediction method

The invention provides a single-phase photovoltaic inverter on-line state monitoring and residual life prediction method, and the specific steps include: first collecting input voltage, input current, output voltage, output current and power MOSFET drain and source electrode current and junction temperature on line in real time when a photovoltaic inverter is a healthy state; then, adopting a depth neural network model to obtain a state parameter reference value of the photovoltaic inverter, calculating to obtain an on resistance reference value of the power MOSFET, and determining a failure threshold value of the single-phase photovoltaic inverter; and finally, based on the depth neural network model, obtaining a state parameter value of a tested single-phase photovoltaic inverter and performing state evaluation based on the reference value, and at the same time, adopting a Gaussian process regression model to perform multistep prediction on the tested single-phase photovoltaic inverter, thereby realizing residual life prediction. Based on multiple performance parameters of the single-phase photovoltaic inverter and taking junction temperature of the power MOSFET into consideration, on-line state monitoring and residual life prediction of the single-phase photovoltaic inverter are realized, thereby providing a theoretical basis for reasonably performing health management on a photovoltaic power generation system.