ANN Based Model for Current Transformers’ Saturation Error Compensation in Medium Voltage Switchgears

Abstract

Current Transformers’ saturation has been a major challenge facing protection engineers’ to-date even with the advent of differential protection system. This saturation causes erroneous measurement of the service currents which can lead to malfunctioning of protection relays and consequently cause false or delayed system trips with severe consequences to the power plant. This research presents an artificial neural networks (ANN) based approach for the compensation of these saturation errors caused by secondary current waveform distortions under transient or fault state. In this research, an ANN algorithm to be implemented in the numerical protection device systems to compensate for these errors was developed. This ANN model applies the multi-regression technique to map a point-to-point compensated waveform referred from the saturated waveform data and ideal/calculated current waveform from the current transformer. The model is developed and trained on python platform with validation and tests to efficiently work with accuracy under the various simulated extreme circuit conditions that the MV system could experience. Electrical transient analyzer program and electromagnetic transient’s program-restructured version software were used to model and simulate current transformer’s transient scenarios generating sufficient waveform data to train, validate and test the ANN algorithm. The model’s processing speed and accuracy was found to be satisfactory for real time application in digital protection devices.