Secondary Measurement Loop Error Assessment in Substations with Application of Deep Learning Networks

Measuring and protection devices in substations are susceptible to errors in the monitored current variations due to environmental and wear and tear, which leads to the risk of inadvertent refusal of the measuring circuits, and it is difficult to detect such amplitude variations by the conventional monitoring methods, based on which, this paper proposes a conditional generative adversarial network (CGAN) and an improved long and short-term memory network (STMN) for the error assessment method. Firstly, the current data of the measurement loop under normal operation is obtained, and the CGAN method is introduced to enhance the generation of error data; secondly, the EMD decomposition of the generated data is performed to form samples and the optimal set of features is selected; in order to further evaluate the error state, the improved LSTM algorithm is used to train the model; finally, a PSCAD/EMTDC simulation model is constructed to verify the reliability and accuracy of the methodology presented in this paper. Finally, a PSCAD/EMTDC simulation model is built to verify the reliability and accuracy of the proposed method. Finally, a PSCAD/EMTDC simulation model is constructed to verify the reliability and accuracy of the proposed method. The test results show that the new method adopted in this paper can reliably evaluate the error state of 2% of the measurement loop of the secondary system.