Wind turbines have extremely low carbon emissions throughout their life cycle,and the use of wind energy is conducive to global climate governance,which is of great significance to the construction of a resource-saving and environment-friendly green development system and the realization of sustainable development of human society.The supervisory control and data acquisition(SCADA)system is the basis of wind turbine operation and maintenance information,and with the further close integration of new energy and information technology,intelligent screening of SCADA system data is the key way to informatization and intelligence in wind power generation.For the problem of wind turbine SCADA system data intelligent screening,this paper combines wind turbine working mechanism,state parameter data correlation,mathematical statistics and machine learning related theories,proposes a wind turbine SCADA system normal data intelligent screening method,and researches and analyzes the application of normal data set to wind power prediction,and the main content is shown as follows.(1)The acquisition of wind resources during normal operation of wind turbines is analyzed,revealing the general distribution pattern of power measurement data.Through the analysis of wind resources and power curves,the influence of wind distribution on the actual power curve plotting of wind turbines was studied to prepare the screening of normal data for wind turbine SCADA system.(2)The correlation between wind turbine condition parameters was investigated.An indepth analysis of the correlation between two important parameters of the power curve characterizing the performance of wind turbines,wind speed and power,and the unit condition parameters is focused.Power data can reflect the situation of wind turbine status parameter data,and provide a grip and ideas for wind turbine SCADA system normal data intelligent screening.(3)The problem of intelligent screening of wind turbine SCADA system data is studied,and an intelligent discriminative method for SCADA system data is proposed Based on the research basis of wind resources,power curves and parameter correlation,a determinable sample category data set is obtained by using Gaussian distribution,DBSCAN clustering and curve fitting least squares method.For the existence of sample category imbalance problem,Near Miss-2 under-sampling method is used to balance the sample set categories.The main state parameters of wind turbines are considered,and the SCADA system normal data intelligent discriminative model is constructed based on random forest algorithm,which intelligently screens out the normal data of wind turbine SCADA system and constitutes the normal data set.(4)The application of normal data sets to wind power prediction model is studied and analyzed.Combined with the normal data set,the prediction model parameters of wind output power were selected to build the prediction model of wind turbine output power in the actual state,which can better predict the actual output of the wind turbine and can be applied to distinguish the abnormal sensor of the wind turbine.At the same time,the wind turbine output power prediction model based on improving power generation efficiency is constructed,which can better predict the optimal output power obtained by capturing wind energy to the maximum extent during normal operation of the wind turbine,which is of great significance for wind turbine state identification and power generation efficiency improvement. |