Electric Power Sensing Data Quality Classification Using Decision-Tree via IEC60870-5-101/104

Abstract

The IEC60870-5-101 and IEC60870-5-104 standards have been widely adopted for use in substation automation systems. The quality of sensing data delivered via IEC60870-5-101/104 (IEC101/104) will determine the performance of systems. This paper presents an AI-based methodology to identify the 3-phase voltage or current sensed in the process level of substation, which is delivered from an IEC101/104-based substation automation system, to generate a data quality classification. The decision tree algorithm is used to classify the data quality such as value up-limitation, down-limitation, normal, phase sequence error as well as refresh failure. Considering that the Decision Tree algorithm requires a large amount of training data, an IEC101/104 simulator is used to generate messages that include not only good quality data but also bad quality data. Based on the decision tree algorithm, a specific decision tree model is built based on the training data to classify the data quality. In the final stage, the methodology is validated using some real data. In terms of sensing data quality, the mistakes from the primary devices to the gateway in substation are identified and improved in case of poor quality. In the future, the specific model will be used in edge computing to implement the sensing data quality assessment locally in the substation.