Research And Implementation Of Fault Early Warning Technology For Smart Meters

As an important metering equipment of the electricity information acquisition system,the smart meter has built an important bridge between power enterprises and power users.To ensure the normal operation of smart meters and to replace the abnormal meters in time,which is of great significance to safeguard the rights and interests of grid enterprises and power users.At present,with the remarkable improvement of the functions of smart meter,the number of smart meters in operation shows explosive growth.If their running state can be accurately warned and then the staff check the abnormal smart meter according to the warning results,which will greatly reduce the operation cost of smart meters and improve the maintenance efficiency of smart meters.Based on the collected historical data of smart meters,combined with data mining algorithm,this paper has studied the fault warning technology of smart meters.The main work of this paper is as follows:Firstly,the source of the historical data of smart meters and its big data characteristics were analyzed,and the raw sample data of the smart meter was preprocessed.In view of the case of sample feature loss and data repeating in the historical data sets,the raw data were cleaned by such these methods as duplicate sample removal.In order to reduce the data dimension,combined with the correlation analysis between the various attributes and the fault state of smart meters,the related attributes of historical data sets were selected,and the irrelevant attributes were removed to improve the performance of data mining algorithm.Aiming at the shortage of abnormal state sample data in the historical data sets,the Synthetic Minority Oversampling Technique was used to solve the problem of data imbalance between abnormal and normal state sample in the historical data sets.Secondly,the data mining algorithm based on decision tree carried out the construction of smart meter fault warning model,the process of which was regarded as the process of fault state classification,and the decision tree model was used to transform the fault warning knowledge acquisition problem into the decision tree construction problem.In view of the efficiency and accuracy of classification learning of C5.0algorithm is higher than that of traditional ID3 algorithm and C4.5 algorithm,the decision tree was constructed by using C5.0 algorithm to extract the knowledge of running state evaluation of smart meter,and the classification rules of smart meter fault warning were obtained.Finally,in order to realize the integrated management of smart meter’s data and remote monitoring of its running state,a fault warning system was developed based on the VS2010 platform,the pre-warning accuracy of which fault was evaluated by test sets.Taking the DDSF885-M smart meter as an example,using the historical data to carried on the example analysis,the results show that the system can effectively assist the power grid operators to make an accurate warning of smart meter’s status.