A Fault Early Warning System For UHV DC Converter Station

In the context of the construction of the national smart grid,UHV DC converter stations,as an important part of long-distance power transmission,require smarter operation and maintenance solutions.Fault early warning,as an important branch of intelligent operation and maintenance,can predict the potential failures of the devices in the future based on existing information,and assist management personnel to locate system instability factors as soon as possible to avoid losses caused by system failure.Starting from the actual application scenario of a domestic UHV DC converter station,this article conducts the research and application realization of the key technologies of the power equipment fault warning platform according to the current operation and maintenance status of the converter station.The main research content of this paper is as follows(1)Starting from the industrial site of the converter station,the current situation of the power equipment distribution and the deployment of different monitoring systems are studied.Aiming at the problems in the current operation and maintenance management for devices,the requirements for the fault early warning platform are analyzed.A fault early warning layered model with the characteristics of cloud edge combination is proposed,which consists of four layers named as the data integration layer,the edge prediction layer,the cloud alarm layer and the human-computer interaction layer.(2)In the edge part of the fault early warning framework,a forecasting method based on data sharing among the devices is proposed.The edge node network model is constructed according to the equipment distribution in the field.On the basis of "sender-receiver" communication mode,the device parameter prediction for multi-node data sharing at the edge side is realized.Among them,the senders use auto-encoder structures to extract the key information of the node,while the receiver uses the multi-granularity time-series forecasting model and feature engineering to realize the time-series forecasting of the device parameters,and transmit the prediction results to the cloud side.(3)In the cloud part of the fault early warning framework,a method to generate early fault alarms for power equipment based on fuzzy neural networks is proposed.The membership functions are designed for the predicted values of equipment parameters,and then the structure of the fuzzy neural network is determined according to the existing expert rules,and finally the equipment fault classification is completed through classification prediction.After fault classification prediction,an RBF neural network false alarm filter based on online learning is proposed,which uses an online learning algorithm to adaptively learn the false alarm samples feedback from users and intelligently filter the false alarm information.(4)Taking a converter station of State Grid as the application background,a set of converter station equipment fault early warning platform software is designed and developed.Firstly,based on the intelligent early warning hierarchical model,the overall design framework of the intelligent early warning platform and the overall solution of the functional modules are proposed.Then the database design,technical framework selection,module business logic and other aspects of the system development are introduced respectively.Finally,the functions and test results of the intelligent early warning system are shown,which verifies the effectiveness and practicability of the platform.