Analysis Of Abnormal Electro-data Based On Data Mining

Under the background of big data era,Chinese power administration pays more attention to marketing informatization.The planning development of "13th Five-Year Plan" pointed out that using big data technology will improve information capacity and business application capacity.With the popularization of the electric power information collection system,the vast amount of electricity data can be collected,which provides a solid data base for the analysis of large data in the electric power system.However,confronted with the massive increase in the use of electricity data,most of the power departments only use traditional statistical methods for anomaly analysis.In that way,abnormal data might not be extracted effectively.Therefore,it is necessary to introduce the data mining technology into the anomaly analysis and to fully exploit the abnormal information of the electricity data.First of all,considering that all the anomalies will be reflected in the power data,we use the daily load curve as a feature index of anomaly detection,since it shows a steady regularity.The outlier detection model is constructed by using the isolated forest algorithm,which has the advantages of less parameters,high accuracy and fast computation efficiency.The model outputs all the user's abnormal scores and the probability of suspected sort.The results show that,using this sort,we only need to detect a small number of outlier detection to detect most of the abnonnal users.Secondly,in order to highlight the isolated forest in the superiority of the algorithm based on electrical anomaly detection,through the establishment of anomaly detection model based on clustering analysis,local outlier factor algorithm and compared.It's found that based on the electric data to construct abnormal detection model advantage in computing efficiency is particularly significant,and the accuracy rate remained high,isolated forest algorithm for the construction of anomaly detection model the accuracy and efficiency of the electricity data proved.Third,the daily load curve took into account by users with habits,we need to combine the electrical parameters of users suspected abnormal for further analysis.Abnormal electro-data recognition model can reduce the misjudgment rate mostly.The decision tree algorithm is able to understand and achieve high efficiency,in order to realize the automatic and fast classification of the voltage of the measuring point,and to assist the judgment of the current data.In the investigation,the identification model is verified.Finally,due to the abnormal electric metering device in the presence of residual voltage,resulting in power compensation errors in the actual case.Therefore,it is necessary to calculate the correction coefficient method to improve the traditional one.Through the analysis of the freezing 96-point voltage data of electric energy data acquire system,considering the residual voltage of fault phase,and the research on electric metering device calculation principle.Adjustment coefficient method is compared with the existing methods,and it's more comprehensive and more reasonable.