Analysis Of Abnormity In Power Metering Automation System Based On Data Mining

The power metering automation system, as one of the most important component of smart grid,takes on the function of data acquisition and analysis. Through the metering automation terminals arranged in the power grid,the system can get data and monitor power plants, substations, transformers and users, and specifically, it has remote meter reading, electricity monitoring, load control, line loss statistical analysis, power quality analysis, customer energy assessment and other functions. As the complexity of the power grid structure, and the large number of equipment and lines,the grid often has failures and abnormalities, which is one of the main factors of the instability of grid and one obstacle to the development of the grid. In this paper, dealing with such a large-scale and widely distributed power metering automation system, a fault diagnosis system with real-time detection and abnormity analysis is designed. The main work is as follows:(1) In this paper, two methods of detecting terminal communication failure and line loss are proposed, which can detect communication and line loss anomaly in a city’s electric energy data for 2014.(2) Based on the terminal traffic data, a communication quality evaluation model is proposed, so that the detection of the terminal fault can be abstracted as the classification of the evaluation model.(3) An improved algorithm based on the kNN(k-Nearest Neighbor) classification algorithm is proposed,which gives the classification’s results and reliability scores. Using the communication quality evaluation model and the characteristics of different types of fault, this algorithm realizes the terminal communication fault detection, classification and early warning.(4) This paper proposes a line loss abnormity detection method based on data analysis. The method includes two parts: abnormal line loss detection and cause analysis.A line loss fluctuation index is designed to measure the changes in line loss, which is used to the timely detection of abnormal lines, and several typical line loss anomaly patterns are used to analyze the characteristics of different modes to determine the cause of the abnormity.The experimental results show that the proposed method can improve the potential communication faults detection ability of the system with an accuracy of 77.01%, and play an early warning role for the terminal that has not failed. The line loss abnormity detection method significantly improves the speed of line loss abnormity cause determination.Finally, this paper uses Spark engine to replicate the above algorithm on big data platform,and the application has been embedded into the big data platform of electric energy data.