Research On Blade Icing Identification Of Wind Turbine Based On SCADA Data Characteristics

In recent years,with the continuous adjustment of the energy structure in the world,the wind power generation as a green renewable energy has been rapidly developed.With the installation of a large number of wind turbines in different regions and different climatic zones at home and abroad,the turbine blades often freeze in certain seasons and under certain climatic conditions.The icing on the blades of the wind turbine will significantly change the shape of the blades and affect the aerodynamic performance of the wind turbine.As a result,the mechanical breakdown caused by the uneven distribution of blade loads caused by icing on the blades will not only bring damage to the life of the components of the turbine but affect the economic benefits of the wind farm,serious ice will trigger turbine safety accidents.In order to ensure the safe and normal operation of the wind turbine in winter icing conditions and to prepare for the subsequent de-icing of the turbine,this paper presents an active diagnostic monitoring method for icing of the blade.This method is based on the data of SCADA(Supervisory Control and Data Acquisition)system of wind turbine,the following research work is carried out from three aspects:the overall performance of the turbine,the efficiency of the blade absorbing wind and the vibration signal of the tower tube.1)The output power is one of the most representative performance indicators of the unit.If the icing phenomenon occurs,If there is ice accretion on the blade,relevant theoretical research,laboratory simulation and relevant field investigation can all find out that there is a big loss in the turbine output power.Based on the wind speed and power cleaning and power screening of wind turbine SCADA system,a regression prediction model mainly based on SVR algorithm is established,which makes it possible to predict the turbine power at different wind speeds in real time.Among them,the input and output of the model were determined by Pearson's correlation coefficient method,and the selection of penalty parameters and kernel-wide parameters in the model was optimized by particle swarm optimization algorithm.By comparing the power predicted value with the measured value,so judging the whole turbine is abnormal or not.2)The size of wind energy utilization coefficient(Cp)reflects the capability of wind turbine blade to absorb wind energy.Its value not only shows a certain functional relationship with pitch angle and tip speed ratio,but also depends on the turbine operating conditions influence of control strategies.The real-time data of wind speed,pitch angle and other real-time data recorded in SCADA system are used to establish the characteristic model of real-time Cp variation monitoring turbine.The operating area of the turbine under full wind speed is divided into four operating states,so four different operating conditions of the turbine Cp are established according to different operating states.The real-time Cp of the turbine under different operating conditions is compared with The normal Cp value in the working condition makes a comparative analysis to determine whether the energy efficiency of the blade is abnormal.3)Due to the irregular and inhomogeneous blade shape,the unequal blade quality results in the change of the vibration characteristics of the whole impeller system and the abnormal vibration signals of the other related components.Ensemble Empirical Mode Decomposition(EEMD)is used to decompose the nonlinear and non-stationary vibration signals of the tower.The characteristics of multiple stationary vibration components under normal conditions and abnormal conditions are compared and analyze.to make a judgment on whether the turbine is in an "abnormal state".4)The three methods for monitoring the turbine blades icing:the real-time prediction based on SVR algorithm at different wind speeds,the real-time monitoring method for using Cp at different wind speeds and the real-time analysis for the tower tube vibration signals of using EEMD method,so according to the unity of abnormal conditions to determine whether the occurrence of icing and near-anomalous time for determining abnormal to determine when the occurrence of icing.