<del id="rl5r5"><ruby id="rl5r5"></ruby></del>
<nobr id="rl5r5"></nobr>

    <font id="rl5r5"><strike id="rl5r5"><big id="rl5r5"></big></strike></font>

      <em id="rl5r5"></em>

      <dfn id="rl5r5"></dfn>

      Wind Systems: Critical Components Essential as Wind Power Scales Up

      2018-01-08 10:00:11 dmsw 20

      SILENCEDIMSEN bearing early wind turbines generally require 20,000 rpm servicerequirements. We are more concerned with wind turbine operators orpower grid dispatching departments on the safety of the whole machine,operational reliability, power generation capacity, operating statechange trend and remaining time of service. Therefore, it is necessary to carry out research on the comprehensive condition monitoring method of the whole machine. At present, most of the state method researches on the whole machineare carried out on the basis of the operating data of the wind turbinebearing data acquisition and monitoring control (SCADA) system, and thethree directions of fault prediction are reviewed separately.
      1.Based on statistical analysis, the comprehensive state monitoring andevaluation of wind turbine bearing sets utilizes statistical analysismethods to analyze offline performance data of various types of windturbine bearing group condition monitoring (power, wind speed, speed,temperature, vibration, etc.). , to extract some regular indicators, compare with the factory designstandard values, or through the comparison between multiple units, toachieve the purpose of monitoring the state of the wind turbine.
      Atpresent, there are many statistical analysis of power operation data,such as the power curve of two 1.5 MW wind turbines. It obtains themeasured wind speed and power data reflecting the running performance ofthe unit, and uses the Bin method to statistically process the data. The power curve of the unit. Throughthe power curve, wind energy utilization curve and standard deviationof the two units, the performance of the unit was compared and analyzed.The actual power curve of the unit 2 shown is lower than that of theunit 1 in the range below the rated wind speed, and at higher than therated wind speed, the standard deviation of the power in the partial Bininterval is too large, and the operating state is not stable.
      The above research is to achieve the state monitoring of the wholemachine by statistical analysis of the power information. Whether thestatistical results of other feature quantities can be used to bettercharacterize the operating state of the entire wind turbine is worthy offurther exploration.
      2. Comprehensive monitoring and evaluation of wind turbines based on multi-parameter fusion
      In this research direction, most of the research was carried out based on the operational data of the wind power SCADA system. The physical characteristics of the wind turbine SCADA system include:angle, pressure, temperature, speed, cabin vibration, electrical, etc.By analyzing these operational data, the operating status of the wholemachine can be reflected.
      At present, the state monitoring and evaluation methods formulti-parameter fusion of wind turbines include artificial neuralnetwork, Gaussian mixture model parameter estimation, matter elementanalysis and fuzzy comprehensive evaluation.
      3. Fault prediction method for wind turbines
      Fault prediction refers to predictively diagnosing the state in which acomponent or system performs its function (future health state) basedon the current or historical performance state of the system, includingdetermining the remaining life or normal operating time of the componentor system.
      Thethree methods of fault prediction are: model-based approach,statistical reliability-based approach, and data-driven approach. At present, the research direction of model-based wind turbine faultprediction is relatively rare, and the other two research directionshave appeared in the relevant literature.
      a. Wind turbine fault prediction based on statistical reliability.
      Atpresent, with regard to wind turbines that have been out of warranty orhave been in service for a long period of time, their operationalperformance is degraded and the deterioration of various components isincreased, resulting in a decrease in reliability and a gradualshortening of Mean TimeBetween Failures (MTBF). Fault prediction studies are relatively rare. However, there have been a few reports on the prediction of windturbine MTBF during the test run, which is generally based on theassumption that the wind turbine reliability obeys a certaindistribution (such as Weibull, non-homogeneous Poisson distribution).
      b. Data-driven wind turbine fault prediction research.
      Inthis part of the research, SCADA data is used to conduct faultprediction research on key components of wind turbines (such as gearbox,generator, main shaft, etc.). The existing fault prediction methodsinclude support vector machine, ARMA method, multiple linear regressionmethod, Artificial neural network and other methods.
      Thebasic idea of most research is to realize fault prediction throughresidual trend distribution. As shown in Figure 4, the fault predictionframework uses SCADA monitoring data as input to the prediction model,which is established by artificial neural network or support vectormachine. Theprediction model obtains the predicted value, and then combines theactual monitored value with the predicted value to obtain the residual,and combines the residual threshold determined by the expert experienceor the normal distribution in advance, by detecting whether thethreshold is exceeded or by residual trend analysis. Forecast the failure.

      Online fault diagnosis of key components of wind turbine

      Thewind turbine is composed of multiple components, and carries outresearch on online condition monitoring and fault diagnosis of its keycomponents. It can identify fault signs in time, grasp the progress offault gradual change and save troubleshooting time in real time, andoptimize the operation and maintenance strategy to improve theoperation. The operational reliability of the machine has important academic significance and engineering practical value. Current status of online fault diagnosis research on five keycomponents including wind turbine impellers, gearboxes, generators,converters and pitch systems.
      1. Impeller
      The impeller is a key component in capturing wind energy, including blades and hubs. At present, there are many researches on blade aging and damage, andimpeller imbalance faults. The existing online condition monitoring andfault diagnosis methods are now in the laboratory simulation stage,which is rare in practical applications.

      a. Blade aging and damage. Bladeoperating environment is harsh, acid rain, freezing and other erosionand impact damage caused by blade rotation, causing blade cracks andeven cracks. Onthe dynamic non-destructive online monitoring technology of blades,there are technologies such as acoustic emission, ultrasonic wave, fibergrating and vibration analysis. Compared with ultrasonic and fiber grating technology, acousticemission can obtain more comprehensive defect information on the blade,with relatively high sensitivity and resolution, and can accuratelymonitor the location of weak areas.

      b. Impeller imbalance failure. Asthe capacity of the single machine increases, the diameter of theimpeller becomes longer and longer, and the flexibility of the windturbine is stronger. Especially in the northern winter, the icing of theblade causes the impeller imbalance to further increase the overallstructural vibration, which will cause fatigue in the transmission chaincomponents. Stress, which seriously affects unit life. Most of the existing research is to extract fault characteristics from the electrical signals of the generator.

      2, gear box
      There are many researches on on-line monitoring and fault diagnosis ofgearboxes. In addition to the oil analysis methods for off-linedetection, online analysis methods mainly include: vibration analysis,temperature analysis and electrical analysis.

      a. Vibration analysis
      The fault characteristics of the gearbox are typically extracted from the time and frequency domains of the vibration signature.
      Atpresent, most of the existing wind turbine condition monitoring systemproducts mainly rely on vibration characteristic quantity analysis. Dataoff-line analysis and expert auxiliary analysis are used to obtain thecondition monitoring and fault diagnosis results of gearbox bearings andgears. However, vibration analysis has certain limitations for low frequencysignals, and the installation of sensors on the gearbox body to obtainvibration signals requires increased investment and maintenance costs.

      b. Temperature analysis
      The temperature characteristic amount reflects the operating state of the gearbox to a certain extent. Thenonlinear state estimation method is used to establish the temperaturemodel of the gearbox under normal working conditions and use it fortemperature prediction. Bysimulating the fault condition of the gearbox, the temperature offsetis added to the SCADA monitoring data to simulate the fault. Theanalysis results are shown in the figure. The upper limit of the 95%confidence interval of the mean curve exceeds the preset mean value inthe 451th sliding window. The threshold, at point 551, monitors for abnormal changes in gearbox temperature. However,since the temperature has thermal inertia characteristics, the changeis slow and susceptible to external environmental factors. When a fixedthreshold is used, when an early warning signal is issued, the componenthas been seriously deteriorated, and the fault may occur soon, which isdifficult to play an early fault diagnosis role. Therefore, it is necessary to study the dynamic thresholddetermination method of temperature characteristic quantity undervarious working conditions.

      c. Electrical analysis
      Time domain and frequency of electrical signals from generators
      The fault feature information is extracted in the domain to implement fault diagnosis of the gearbox. Comparedwith other signals, the collection of electrical characteristicquantities does not require additional sensors and does not affect theintegrity of the unit. However, the wind turbine operating environment is complex, there aremany sources of interference, and there may be overlapping of multi-partanomaly features. The actual operating environment needs to beconsidered, and the gearbox state feature extraction algorithm based onelectrical characteristics is deeply studied.

      d. Oil analysis
      Whenthe gear meshing occurs in an abnormal wear state, huge wear particlesare generated instantaneously, or the wear rate increases to cause asignificant increase in the number of wear particles. Oilanalysis is considered to be one of the most effective techniques forgearbox condition monitoring and fault diagnosis, including infraredspectroscopy, iron spectroscopy, particle analysis, gas chromatography,etc., by extracting various monitoring indicators in the oil, including Kinematicviscosity, PQ ferrography, acid value, moisture, etc., monitor andanalyze the changes of various monitoring indicators in the oil, andachieve abnormal detection of the gear box. At present, many institutions in China have established oil monitoringlaboratories, but due to the limitations of monitoring hardware(sensor) design and manufacturing technology, there are largemeasurement errors and low precision factors, and online oil monitoringhas not been realized in practice.

      3, the generator
      Most of the fault diagnosis research on generators is to monitor thestator current, rotor current and active power changes online, and todiagnose faults such as turn-to-turn short circuit, single-phase ormulti-phase short circuit, bearing damage and rotor eccentricity.

      4, converter and pitch system

      a. converter
      Asthe key control channel for the electric energy feedback to the powergrid, the converter is an important link that affects the safe andstable operation of the wind turbine and the network. In the existing literature, an intelligent fault diagnosis method based on sample training for online converters is widely used.
      Theresearch mainly focuses on the fault diagnosis of wind powerconverters. However, due to the uncertainty of wind speed, thereliability of wind turbine converter operation is affected by the lowfrequency operation of the machine side converter and the randomfluctuation of wind speed. The converter output power is serious. Thechange is very large, so that the power device operates at differentload levels, which may cause changes in the junction temperature of thedevice. The power device will withstand long-term and frequentunbalanced electro-thermal stress, causing fatigue accumulation,resulting in failure of solder cracking and wire drop. . Thefigure shows the turbulent wind speed of 11 m/s and the change of theIGBT junction temperature of the machine-side converter of thedoubly-fed wind turbine. It can be seen that the IGBT power device needsto withstand frequent fluctuations and the amplitude is 20 ° throughoutthe life of the converter. Thejunction temperature thermal cycle of C will inevitably accelerate theaging and failure rate of the device. Therefore, the on-line conditionmonitoring of the device of the wind power converter based on fatigueand failure mechanism should attract attention. Research in this field of power devices is currently rare.

      b. Pitch system. Thereare few researches on the fault diagnosis of the pitch system. Thepitch system has extremely low speed, non-continuous operation andrandom load. The online condition monitoring can be analyzed byvibration analysis or generator current signal. Inaddition, considering the absolute threshold evaluation method based onsingle parameter, it may lead to the problem that the existing pitchsystem state monitoring method cannot judge its abnormal state online intime and accurately.