Power Quality Parameters Forecasting Based on SOM Maps with KNN Algorithm and Decision Tree

This study tested four forecasting models combined with 3x3 SOM maps for predicting power quality parameters (PQPs) named decision tree (DT), KNN algorithm, bagging decision tree (BGDT), and boosting decision tree (BODT). The input variables used are weather conditions (air temperature, wind speed, air pressure, Ultraviolet, solar irradiance) with states of four types of home appliances (AC heating, light, fridge, TV) represented by one decimal number. Target Outputs are Power Voltage (U), total harmonic distortion of voltage (THDu), total harmonic distortion of current (THDi), power factor (PF), and power load (PL). The experiments were carried out in two stages: in the first stage, clustering dataset using self-organizing maps (SOM), 3x3 SOM in total nine hexagon nodes was used. In the second stage, inside each node builds four forecasting models: decision tree (DT), K-Nearest Neighbor(KNN) algorithm, bagging decision tree (BGDT), and boosting decision tree (BODT). Root Mean Square Error (RMSE) was used for evaluating the performance of studied models.