Research On Data Mining Methods For Electric Energy Data Quality Analysis

Metering automation systems in intelligent electric grids can collect the electrical energy data of generation sides, power supply sides, power distribution sides and marketing power sides from power plants,substations, public transformers, special transformers, lower voltage centralized meter reading and others and transmit the data through the GPRS network to the data centers. As the collected electric energy data is an important basis for user payment and planning of urban electricity consumption, it is particularly important for power companies to improve their data quality.This paper aims to analyze the data of electric energy collected by the metering automation system and repair the anomaly data with reasonable and efficient data mining technology, in order to ensure the integrity and accuracy of the data. The main work of this paper is as follows:(1) The methods of statistical analysis and clustering are used to discover the missing data of electric energy data and its patterns, which can be used to assist the operation and maintenance of power grid.(2) Based on the two evaluation indexes of power and current voltage angle difference and power moving mean difference,a data accuracy evaluation model is proposed. The model can score every energy data collected and estimate the starting time and end time of data anomaly.(3) The comparative experiment of electric energy data repair is completed for three algorithms of the mean value method, local weighted linear regression algorithm and support vector machine. The results show that the local weighted linear regression algorithm has the highest accuracy.The paper analyzes the quality of the data, a total 550 million records,from more than two kinds of electric energy data in measurement automation system, such as table code and instantaneous quantity. It also analyzes the data missing patterns from two aspects of statistical analysis and clustering and proposes two accuracy evaluation indexes. In this paper,the local weighted linear regression algorithm is applied to the field of electric energy data restoration, which improves the accuracy of data restoration from 87.78% of the existing mean method to 92.09%, ensuring the accuracy and efficiency of data restoration. The experimental results show that the work of this paper is helpful to the operation and maintenance of urban electric power, and it can promote the technical development of intelligent electricity and system operation and maintenance level.