How to Predict the PV Module with Maximum Power Output Predictingalgorithm Based on Artificial Intelligence Technology

In general, PV modules varies depending on the manufacturer, but in the case of ground-based PV modules are guaranteed to have a power output of 25 to 30 years. In addition, power output of PV module is guaranteed to be approximately 80~85% of the initial level within 25 years (annual power output drops of 0.5-0.6% excluding the first year). However, since PV modules are installed in the field, the lifetime of PV modules may very depend on outdoor conditions, dust, and manufacturing quality. Additionally, even within the same PV string, PV modules indicate different power output degradation. Deterioration of PV modules can reduce the lifetime or power output of the entire system in a photovoltaic system, so detection technology is needed. Current-Voltage (I-V) inspection and electroluminescence (EL) inspection, etc. can be generally performed under standard test condition. It is impossible to simply measure it because there is a prerequisite that a solar irradiance meter is required to measure the I-V curve in a photovoltaic plant. In the case of EL inspection, there is a disadvantage that a camera that can measure the entire module is needed because the light emission that occurs when electricity is applied under certain conditions to the solar module must be observed. In this paper, we studied a measurement technology that can predict the power output of PV modules by applying constant current in a PV module using a supervised learning-based machine learning model. A total of 10 PV modules were used in the experiment, and the I-V and EL measurement data of the modules were used to compare the performance of the machine learning model and analyze the results. A total of three types of regression-based machine learning models were used, and the imbalanced data problem was attempted to be solved through SMOTE.