Forecasting Power Quality Parameters using Decision Tree and KNN Algorithms in a Small Scale Off-Grid Platform

Abstract: The article presents the performance comparison results of four forecasting methods in forecasting electric power quality parameters (PQPs) in small-scale off-grid environments. The following methods were compared: Bagging Decision Tree (BGDT), Boosting Decision Tree (BODT), and K-Nearest Neighbor (KNN) algorithm with k = 5 and with k = 10. The main goal of this study is to find a relation between the input variables (weather conditions, 1st and 2nd step back of PQP, consumed power of home appliances) and the power quality parameters as target outputs. The studied PQPs are amplitude of power voltage (U), Voltage Total Harmonic Distortion (THDu), Current Total Harmonic Distortion (THDi), Power Factor (PF), and Power Load (PL). The root mean square error (RMSE) was used to evaluate the forecasting results. BGDT accomplished better forecasting results for the THDu, THDi, and PF. Only BODT achieved a good forecasting result for PL. KNN (k = 5) algorithm achieved a good result for PF prediction. KNN (k = 10) algorithm predicted acceptable results for U and PF. The computation time was considered, and the KNN algorithm took a shorter time than ensemble decision trees. Funding: This article has been supported by EU funds under the project “Increasing the resilience of power grids in the context of decarbonisation, decentralisation and sustainable socioeconomic development” (CZ.02.01.01/00/23_021/0008759), through the Operational Programme Johannes Amos Comenius.