Detection of Bad Data and Estimation of Missing Parameter Values Using System Synergism

Communication infrastructure is paramount for the optimal control and protection of modern-day distribution systems. However, inherent problems associated with the communication systems such as loss of data packets due to excessive latency and erroneous signal induction result in impairment of the network control and relaying signals. In the present article, a method based on Data Mining is proposed for the detection of Bad/Noisy data. Furthermore, the concept of system synergism is utilized to estimate the value of the signal in which Bad Data is detected. In case of Bad measurements during faults/transients or load switching, the estimates are observed using MachineLearning (ML) based regression model. XGboost has been found to have the best performance among the ML models. Modbus is considered the communication protocol for the power distribution system under the consideration. Test results are demonstrated using a standard IEEE 33 bus test system. The feasibility of the proposed algorithm in real-time is confirmed using the raspberry pi 3B+ controller using Serial Peripheral Interface (SPI) protocol.

Automated summary

Communication infrastructure is crucial for the control and protection of modern distribution systems. This article proposes a data mining method to detect bad/noisy data and utilizes machine learning models to estimate the value of the detected bad data signals.