Artificial-Intelligence-Enabled Lifetime Estimation of Photovoltaic Systems Considering the Mission Profile of the DC-AC Inverter

This paper proposes a robust artificial neural network (ANN) model based on artificial intelligence that predicts the accumulated damage per cycle in a photovoltaic (PV) system, thus indicating their remaining operational lifetime and unreli-ability; an accurate model is paramount in the task of ensuring the reliability of power electronic systems after being exposed to varying environmental conditions and operational stresses. This study employs the thermal model of a PV system's dc-ac inverter connected to an ac grid, meticulously extracting the thermal, input power, and load data under fluctuating demands and inputs. Following this, a comprehensive analysis is enabled by interpolating a year-long dataset of each relevant signal with the results of this simulation. This process forms the foundation of deploying a Monte Carlo simulation sequence, after which a Weibull distribution is deployed to provide insights into the lifespan cycles remaining based on the accumulated damage over time and its unreliability. By leveraging this dataset, constructing an ANN capable of predicting the lifetime consumption or damage in a thermal cycle with a maximum accuracy of 78.90% is possible. The applications of this research can extend from the enhancement of maintenance schedules to real-time applications in the digital twin modeling of power electronic systems. This predictive model contributes to the ongoing efforts to improve the sustainability and reliability of power-electronic-based power systems by predicting expensive malfunctions and extending the lifetime of components critical to the power system.