A machine learning approach for coordinated voltage and reactive power control

Abstract

Increasing penetration of renewable-based distributed generators (DGs) has transformed passive distribution networks to active distribution networks (ADNs). Therefore, traditional practices for voltage and reactive power (V/Q) control should be revised and improved. All control resources should be coordinated based on real-time information and in closed loop. To achieve this, machine learning (ML) is used to assist in making decisions by mapping the relation-ship between the selected network information and the desired control output. In this paper, setting of the shunt compensator operating in capacitive or in-ductive modes is coordinated with the tap position of the substation transformer such that all security measures are within the limits. Dataset emulating network behaviour during a year of operation is con-structed for training a ML algorithm. A multi-class classification problem is formulated. Simulation results show satisfactory accuracy for some classes.