Research On Hierarchical Early Warning Of Wind Turbine Operation State Based On SCADA Data

The SCADA system provides a reliable source of big data for the evaluation and prediction of wind turbines operating status.How to effectively analyze these data,carry out unit operating status evaluation and prediction,and reduce wind turbine operating costs and downtime is an urgent problem to be solved.Based on SCADA data,a hierarchical early warning model to evaluate and predict the operation status of wind turbines was established.The main research contents include:(1)Data preprocessing and feature parameter extraction.The SCADA system has a huge amount of data and strong randomness,including many useless data points.In order to reduce the impact of useless data points on the reliability of the research results,cleansing out the data during shutdown,failure and braking of wind turbines,and then combining the least square method were imposible.The filtering method with dispersion analysis filters out the "dirty data" far away from the main data band of wind speed/power,and finally normalizes to eliminate the influence of the dimensions of different types of operating parameters on the accuracy of the model.By analyzing the operation and control principles of wind turbines,using mutual information algorithms to quantitatively analyze the correlation between operating parameters,extract characteristic parameters,reduce redundancy,and lay the foundation for the subsequent establishment of predictive models.(2)Establish a hierarchical early warning strategy and neural network prediction model for the operation status of wind turbines.By analyzing the correlation between the operating parameters of wind turbines,firstly establish a neural network-based early warning model for the overall operating status of the wind turbines.If abnormalities are found,the subsystem levels,namely wind wheel system,transmission subsystem,and generator subsystem are carried out.The early warning model of the operating state locks specific abnormal subsystems.It is convenient for later maintenance and management.In order to reduce the impact of individual data on the accuracy of early warning,the sliding window model is used to analyze the error between the predicted value of the forecast model and the actual value in each window,calculate the evaluation index,and then according to the kernel density estimation method,the threshold value of the evaluation index under the normal operation state of the wind turbine is counted as the evaluation standard for evaluating the operation state of the unit.(3)Based on the SCADA data of a wind farm in North China,an example of hierarchical early warning of the operation status of the wind turbine and its subsystems was analyzed to verify the effectiveness of the early warning model.The on-site measured data of wind turbines and after the failure in normal operation are selected as examples for verification,and the data preprocessed by the wind turbine data in this period of time are substituted into the neural network prediction model to calculate the evaluation index.When the evaluation index exceeds the threshold for three consecutive times,it is determined that the wind turbine is operating abnormally;when the evaluation index is lower than the threshold,it is determined that the wind turbine is operating normally.And use different neural network models for comparative analysis.(4)Based on Matlab,design the GUI interface of the wind turbine operating status warning.In order to facilitate users’ use and product promotion,a operating status hierarchical warning GUI interface for wind turbine based on actual applications and visual interface requirements,which can visually observe the operating information of the unit was designed.