Comparison of Direct and Indirect Approaches to PV Power Estimation

This study compares direct and indirect approaches for estimating the output power of solar photovoltaic plants. The effectiveness of the direct approach is assessed through a case study of the Bichurskaya PV power plant, utilizing a supervised machine learning model, specifically the MLP model. Data from two grid-connected PV power plants, based on PVOD data, is used to evaluate the performance of the indirect approach, which involves sequential mathematical modeling of PV plant components using models from the pvlib library. Results indicate that, with sufficient input data, the direct approach is able to surpass traditional mathematical modelling in PV plant output power estimation. The MLP model achieved an nRMSE metric of 3.3%, while the model chain demonstrated a 5.3% performance on the same dataset.