Forecasting of Power Quality Parameters Based on Meteorological Data in Small-Scale Household Off-Grid Systems

Off-grid power systems are often used to supply electricity remote households, cottages, or small industries, comprising renewable energy systems, typically a photovoltaic plant whose is stochastic in nature, without distributions. This approach economically viable and conforms the requirements of European Green Deal Fit for 55 package. Furthermore, these associated with lower short circuit as compared distribution grid traditional plants. The quality parameters (PQPs) such small-scale off-grid largely determined by inverter’s ability handle impact device; however, this makes it difficult accurately forecast PQPs. To address issue, work prediction models PQPs function meteorological conditions regarding households Central Europe. end, seven models—the artificial neural network (ANN), linear regression (LR), interaction (ILR), quadratic (QLR), pure (PQLR), bagging decision tree (DT), boosting DT—were considered forecasting four PQPs: frequency, amplitude voltage, total harmonic distortion voltage (THDu), current (THDi). computation times their accuracies were also compared. Each model was three sunny days August. As result study, most accurate methods DTs. ANN requires longest computational time, conversely, LR takes shortest time. Notably, aimed predict poor that could cause all equipment respond advance disturbances. study expected be beneficial substations included existing smart grids.