Power System Control and Protection Models Based on Artificial Intelligence – A Tensorflow Approach

Abstract

Abstract Artificial intelligence (AI) and Deep learning (DL) methods in power systems are being tested and prepared for practical use in many applications. In this work an artificial neural network models for fault identification and classification and switching logic control in middle voltage (MV) power electricity network is presented. Models are implemented in Google’s Python based tool Tensorflow with belonging program libraries. For fault detection and classification example a few thousand simulations are conducted in order to obtain enough fault current and voltage samples for high accuracy artificial neural network (ANN) with backpropagation model. Achieved accuracy and speed of presented deep learning model, open a possibility for application in digital relay protection devices. Second example is implementation of switching control rules in HV/MV substations. Presented models are patterns for power system controlling centres as part of broader controlling and protection logic.