Prediction of power disturbances by artificial intelligence methods

Ensuring the continuity of power quality is the most important factor for efficient and reliable operation of a power system. In a balanced three-phase system, voltage and current are sinusoidal and nominal. When any fault is occurred on power systems, the system is unbalanced and power quality disturbances can be occurred. They cause power losses on the transmission and distribution lines, malfunctioning of devices, disasters such as fire that will affect living life and the environment. In this context, the use of predictive artificial intelligence methods in the prediction of power quality disturbances offers an innovative and supportive solution in the process of preventing the possible effects of power quality disturbances, ensuring the efficient and safe operation of the power system, and transition from traditional grids to smart grid and microgrid structures. The aim of the study is to predict the magnitude of the instantaneous voltage sag in a bus (single fault-7 ones) or buses (double fault-other 7 buses) of the IEEE9 bus test system according to 14 total short circuit faults before defined time duration by homogeneous and heterogeneous machine learning methods. While selecting the methods, firstly load flow analysis was performed to investigate the voltage sag behavior and then a short circuit fault was in order to phase a. When homogeneous (random forests) and heterogeneous (support vector regression, decision tree, K- nearest neighbors) machine learning methods are compared in terms of prediction performance and speed, it is observed that the heterogeneous ensemble training model is generally more successful. Keywords: Power Quality Disturbances, Voltage Sag, Ensemble Machine Learning, Heterogeneous Machine Learning, Homogeneous Machine Learning