| China has built an AC-DC hybrid power grid with the largest scale,the strongest transmission capacity and the highest voltage grade in the world.Its complex operation,monitoring and control are highly dependent on the dispatch and control system,which is the "nerve center".Millions of automation equipment generate massive real-time data,forming a complex interactive operation system.The traditional management and control mode,which mainly focuses on manual fault handling,is difficult to meet the needs of continuously improving data reliability of the future power grid,and it is also difficult to give full play to the overall effect of massive real-time information.Therefore,it has gradually become a short board in the development of operation control of large power grid.Therefore,for the specific research object of the smart grid dispatching and control system,this paper studies the abnormal intelligent detection method and fault location technology of the power grid dispatching and control system.Dispatching control system considering the operational relation is complex,there are many different kinds of operation and other characteristics,the research content mainly includes the dispatching control system knowledge map construction,dispatching control system data traffic anomaly detection,disk failure prediction and dispatching control system fault location method and platform construction and application of five aspects,such as research,the main work is as follows:(1)Research on the construction method of knowledge graph of dispatching systems of smart grid.In view of the complex operation relationship and various kinds of operation in dispatching control system,a knowledge graph construction method of dispatching control system was proposed based on the combination of bottom-up and top-down.Through the acquisition,storage and application of knowledge,the analysis of structured and semi-structured data sources,based on clustering algorithms for the semi-automatic construction of ontological library,the formation of system structure and detailed business knowledge and other information,so as to build a vertical knowledge map covering the basic platform of the scheduling control system and business logic,improve the understanding and understanding of the complex system internal logical relationship,for the follow-up scheduling control system anomaly detection,fault positioning and other related research to provide a scientific basis.(2)Research on anomaly detection method of dispatching and control system operation flow data.In order to solve the problem of"concept drift" in massive multi-dimensional real-time dispatching data anomaly detection,a fault anomaly detection method for dispatching and control system flow data based on isolated forest algorithm was proposed.Using the idea of unsupervised incremental learning,several subforest anomaly detectors are established to form the basic forest anomaly detectors.Judging the abnormal situation of entering the sliding window data;The stream data entering the sliding window is systematically sampled to judge whether it is stored in the buffer or not.Detector updates are triggered when the sliding window exception rate is too large or the buffer is full.Introduce new anomaly detector update strategy,abandon the abnormal rate of the deflection forest anomaly detector and added into the new child son anomaly detector,solved the low state of the single threshold detection accuracy,static offline data analysis cannot adapt to the operation problem such as dynamic change,for the subsequent determination of fault types and provide evidence for the cause of the problem analysis and accurate positioning.At the same time,in order to model the time characteristics and laws of flow data,the self-supervised learning framework is introduced to further improve the performance of the detector.The comparison between the simulated data set and the real data set of dispatching control system verifies the advanced performance of the proposed method in the comprehensive performance of anomaly detection recall rate and false alarm rate.(3)Research on device disk fault detection method of power grid dispatching and control system.Disk is an indispensable core component of scheduling automation equipment.Aiming at its characteristics of high dimension of associated attribute of monitoring state data,high coupling degree between attribute and model,and serious imbalance of disk positive anomaly data,a disk fault detection method based on adaptive weighted Bagging-GBDT algorithm was proposed.Through many times of health disk samples based on clustering stratified under-sampling in order to reduce the loss of useful information,most of the class sample and the sample after sampling combined with samples of all the minority class to get more than one training sample subset,according to the adaptive weighted voting method to carry on the multiple GBDT classifier integration,realize the accurate detection of the disk failure.At the same time,on this basis,the value function estimation to disk prediction model is introduced to improve time generalization.The validity and advance of the proposed method are verified on Keel unbalanced dataset and disk public dataset.(4)Study on fault location method of smart grid dispatching and control system.Dispatching control system there is a operation constraint relationship,process complex coupling between high characteristic,in view of the existing fault location method function independently,the problem of high false alarm rate,in constructing the system of foundation platform and operation logic,on the basis of knowledge map,put forward a kind of through the operation logic,system software and hardware resource constraints mapping decoupling fault location method of dispatching control system.Through system according to the constraint relation and the operation logic for decoupling,establishing the fault tree model of each operation and internal constraint mapping table and operation between constraint mapping table,put forward the operation internal source between abnormal process tracking and operation process of fault location method,through the operation logic constraint location abnormal original within the operation process,through software and hardware resource constraints positioning fault the original process,improve the dispatching control system and the accuracy of fault location.(5)Construction and application of anomaly detection and fault location platform for dispatching control system.In a provincial power grid dispatching institutions set up a production environment dispatch control system anomaly detection and fault location platform:integration modeling technique was used to construct panoramic monitor model of dispatching automation system and asset maintenance model,based on the distributed collection information to dispatch control system has been put into operation in equipment monitoring and management;The platform data transmission and storage are supported by the efficient transmission technology of stable long connection,the efficient processing technology of cache and the automatic storage technology based on database and table.Based on this platform,the application and analysis of abnormal detection of dispatching and control system operation flow data,disk fault prediction and system fault location are carried out,and the validity of the research results is verified. |
PDF Full Text Request