Research And Application Of Real-time Bad Data Processing Method For WAMS

A variety of bad data generated by random abnormal events in wide area measurement system has caused serious pollution to the data quality of power system,and then threatened the production safety of power grid.Therefore,how to quickly and accurately identify and recovery the bad data in the power system is of great significance to ensure the stable operation of the power system.Howerver,The existing bad data processing methods either have problems such as weak generalization ability and easy to cause residual inundation,or have high computing costs,which is difficult to apply to online processing systems.In view of the above problems,this paper proposes a method for identifying and restoring bad data in wide area measurement system,and designs and implements a realtime processing system for bad data in wide area measurement system.The main research contents include : through the analysis of power data,it is found that normal sampling point data and adjacent sampling point data have strong spatio-temporal similarity,and abnormal sampling point data and adjacent sampling point data have weak spatio-temporal similarity,on this basis,the calculation method of spatio-temporal similarity and the bad data identification method based on the nearest neighbor of space-time are proposed,the above method can quickly and accurately detect bad data;based on the conclusion that there are similar data trends between sampling points,a bad data recovery method based on spatial trend prediction is proposed,which can quickly recover bad data with high accuracy;based on the real-time streaming data processing framework Storm,a real-time processing system for bad data is designed,which realizes data input module,data preprocessing module,abnormal data recognition module,abnormal data recovery module and data output module,so as to effectively deploy bad data processing methods.The bad data processing effect of the above method and the overall performance of the system are tested by experiments,which verifies the effectiveness and feasibility of the realtime processing system of bad data.In the bad data recognition experiment,the recognition accuracy is 95.72 %,in the bad data recovery experiment,the recovery error for missing data is 1.16 %,both methods are better than the comparison method;In the performance test of the system,the average processing delay is 228.3ms,which can meet the real-time requirements of power data processing.