The Study Of Online Condition Assessment And Fault Localization For Large Scale Wind Turbine Based On SCADA System

With the aggravation of the energy crisis and the enhancement of people’senvironmental protection consciousness, wind power is being paid more and moreattention for its comprehensive advantages of technical maturity, infrastructureconstruction and cost aspects. However, the development of wind power industry isseriously restricted by the operation and maintenance cost which is high up to10%~15%. Therefore, consideration from reducing operating risks for wind turbines andthe operation maintenance cost of wind power and so on, online condition assessmentand fault diagnosis technology of large wind turbines have become an urgent subjectto be solved for wind power industry.The SCADA system equipped to wind turbine alarm through the numericalout-of-limit. The type of alarm is single and don’t have the function of preventingfault deterioration. A large number of SCADA monitoring data implies interactionand influence among components and subsystems of wind turbine. Therefore, thispaper proposed an online condition assessment method of wind turbines using datamining technology. This method integrated a regression forecast model based onSupport Vector Regression algorithm on the existing SCADA alarm system.Monitoring personnel could assess operating state of wind turbines by means of alarminformation of SCADA system and deviation and its trend of predicted values of theregression model and the actual monitoring values. The method not only couldenhance the precision of condition assessment, but also could track the developmentof faults. It provided strong technical support for reducing the risk of operation,optimizing the maintenance strategy and reducing the cost of operation andmaintenance of wind turbines.Based on the thinking of "community matching", this paper divided themonitoring items of SCADA system into six overlapping communities correspondingwind turbines sub-systems from the operation and control mechanism of each windturbines sub-system, and detailedly analyzed the criterion judged whether the sampling data of every monitoring item was abnormal. When an exception ofsampling data of some monitoring items occurred, matching these monitoring itemsto wind turbine sub-system corresponding to their community so as to achieve thepurpose of fault localization. The monitoring data and operation and controlmechanism of wind turbines were combined organically in this method, whichovercame the disadvantage that the artificial intelligence method existing rely toomuch on monitoring data, and had stronger generalization performance.In short, the method proposed in this paper about online condition assessmentand fault localization provides the guarantee of safe operation in the long cycle forwind turbines, offers the technical support of informed decision for the maintenancepersonnel, and paves the way of reducing the cost of operation and maintenance ofwind power.