Wind Turbine Condition Modeling And Monitoring Based Big Data

The harsh environment of wind turbine and the growing wind turbines aging problem leads to the high failure rate of the sensors and other component of wind turbine.That causes high maintenance costs and low economy benefit.It is crucial to monitor the wind turbine performance in real time.The paper build the model of the wind turbine when the units are operating based the SCADA data.1.The model of wind turbine based on wind energy utilization coefficient is established to monitor the operation status of the unit.The failure will directly affect the value of wind energy utilization,so the wind energy utilization is used to represent wind turbine performance.The data of the factors of affecting the wind energy utilization(wind speed,tip speed ratio,yaw and turbulence)are extracted from SCADA data,and the model of wind energy utilization is established by nonlinear state estimate technique(NSET).The model reflects the relationship of the utilization rate of wind energy and other influence factors when wind turbines operate normally.After the establishment of the model,the real time data provides the input vector and an output vector will be estimated by the memory matrix.If the input vector is abnormal,the predicted residual will increase observably,which indicates the wind turbine has failure.Several experiments are conducted and the method works quite promising2.The sensor model of the neighbor comparison unit is established.Wind turbines that experience similar wind resources usually have similar working conditions and sensor measurements.With the anemometer as study case,the wind speed data from SCADA system were selected,and the neighbor comparison unit is determined by Pearson correlation coefficient.Two methods are used to establish the neighbor model of anemometer.The first method is to determine the standard value of the correlation coefficient of the neighbor comparison unit which is taken as the criterion to evaluate the operation condition of the wind turbine.If the value of the correlation coefficient is lower than the standard value,it indicates the anemometer has an abnormality.The other method is building the neighbor comparison model of anemometers by the NSET.The model reflects the similarity relation of the wind speed of neighboring units.When the input of wind speed data is abnormal,the predicted residual will increase observably which is an indication of the failure of this anemometer.In this paper,the two methods are simulated with the data of a wind farm,and the validity and accuracy of the methods are verified.The two methods also apply to monitor other sensors of wind turbine.