Research On Trend Prediction Method Of Running Stability Deterioration For Wind Turbine Transmission System

Whether the large rotating electromechanical equipment running in safe and stablestate has an important impact on the economy and society, so, researched on trendprediction method of running stability deterioration is significant. Wind turbines are typicallarge rotating machinery, and running state deterioration of its transmission system is a timecourse with the development of faults. If the state evolution form fault appearance to faultmanifestation macroscopically can be revealed by trend prediction of running statedeterioration, it will provide scientific means to implement predictive maintenance, avoidoccurrence of fatal accidents.This research mainly makes an emphasis on trend prediction method of runningstability deterioration for wind turbine transmission system. The main research contents:(1) Stability deterioration feature of wind turbine transmission system is weakgenerally during the early deterioration process, and deterioration feature was oftensubmerged by variable condition information. To this end, a novel weak informationpretreatment algorithm based on Birgé-Massart penalty function strategy is proposed.Signal and noise wavelet transform modulus maximum have different propagationcharacteristic with the increasing scale in multi-scale space, obtaining suitable modulusmaximum sequence through Birgé-Massart penalty function, the algorithm achieved noisesuppression effectively by using this characteristic. Simulation data and experimental datawere used to validate the proposed algorithm, and the results showed that this algorithm notonly highlighted state feature, but also obtained a smaller Root mean square error andhigher Signal to Noise Ratio.(2) Researched on feature extracting method based on higher-order cumulant diagonalslice, and an evaluation method using sensitivity property, trend property, differenceproperty, and consistency property is proposed. Experiment data of varying degrees ofdeterioration under various types of deterioration were carried out to validate the proposedmethod, and the results showed that feature extracting method can separate the deteriorationcharacteristic from non-degradation characteristics, different feature extracting method have diverse performance to stability deterioration, and the evaluation method can providetheoretical basis for selecting feature extraction method for trend predictionof runningstability deterioration.(3) HMM was introduced to stability deterioration trend prediction of wind turbines,the concept of frequency band mean energy based on1.5dimension spectrum was proposed,and a trend prediction method of transmission system state deterioration based on thisconcept was proposed. The feature vector of state deterioration was obtained by featureextraction method fusion of1.5dimension spectrum and fourth-order cumulant diagonalslice spectrum, then a variety of diagnostic models of transmission system were established.Based on state deterioration diagnosis results, state deterioration squence was obtainedaccording to the concept of frequency band mean energy, and the trend prediction methodwas applied to predict deterioration trend. Real state vibration data were collected on theindustrial scene, and the data analysis results showed that this trend prediction method isfeasible and effective.(4) Characterization parameters of stability deterioration, such as stability deteriorationmean, stability deterioration variance, were difined. Based on those concepts, a stateassessment method of running stability deterioration faced to wind turbines transmissionsystem was proposed. Real state vibration data of wind turbines were given to illustrate theperformance of the state assessment method. The results showed that the proposed stateassessment method can describe the evolution track of stability deterioration accurately.