Power Big Data Processing Method Based On Heterogeneous Data Source

With the continuous deepening of smart grid construction,the scale and type of data in power system are growing rapidly,and the complexity of data is also rapidly increasing,showing typical big data characteristics.Introducing the big data processing method into the power system application field,we can pay attention to the potential information in the power big data,further explore its value,and provide reference for grid dispatching decision.The Data Monitoring and Acquisition System(SCADA)and the phasor measurement unit(PMU)-based wide-area measurement system are two representative measurement systems that are widely used in power systems.Compared with SCADA data,WAMS data accuracy and refresh rate performance are better.Together they form a massive measurement data resource for the power system.Based on the data processing requirements in power big data,the thesis studies the SCADA system and WAMS system.Data mining methods and machine learning algorithms represented by association rules are applied to power systems to meet the requirements of bad data identification and heterogeneous data fusion in power big data,improve data quality,ensure data integrity,and fully exploit data.Value,the most powerful aid for grid decision-making.This paper proposes a method for identifying bad data based on heterogeneous data association rules mining.Considering the periodic properties of the sample,discrete SCADA data with high-precision fast-refreshed WAMS data,mining the periodic association rules between sampling time,active power,reactive power and current,constructing suspected bad data sets through rule matching,and using local The residual search method verifies the bad data set,and compares the feasibility of the verification method with the artificially set bad data and the residual search method.The superiority of the method is verified by comparing the scale of the bad data with the number of identification times.At the same time,a method based on machine learning SCADA data and WAMS data fusion is proposed.Analyze the obstacles in the process of SCADA data and WAMS data fusion and explore solutions.Focusing on the SCADA data loss problem in the fusion process,the machine learning is introduced into the missing data filling field,and the self-encoding neural network is constructed.The unsupervised network training and parameter optimization are performed layer by layer after the self-encoder is stacked to extract The change characteristics of the SC AD A data are finally connected to the top-level LR model to realize the prediction filling of the missing values,thereby overcoming the time-span inconsistency obstacles in the fusion of the two systems,achieving effective fusion,and finally,the actual SCADA value is used and compared with other algorithms to verify the feasibility of the method.