On-line Early Warning Method And Software Of Wind Turbine Operation Status Based On Kernel Density Estimation

The identification and warning of wind turbine operating status is an effective way which can improve the power generation efficiency of wind turbines,shorten downtime and reduce maintenance costs.Due to the wide application of modern large-scale wind turbine SCADA system,the feasibility of implementing online identification and early warning to wind turbine operation status is greatly improved.In addition,the research of online identification and early warning on the operating status of wind turbines based on SCADA system has become an important branch in the operation and maintenance of modern large-scale wind turbines.This paper focuses on the research of online early warning method of large-scale wind turbine operation status and the corresponding software development on the data of on-site SCADA system.The main research work is as follows:In the first step,a wind turbine operating state health indicator model is established and the sliding window parameter is optimized.Wind turbine are divided into multiple subsystems by analyzing the characteristics of wind turbine energy flow.Then the sliding window model and polynomial fitting method are applied to analyze the wind turbine SCADA data.Thus,the parameter relationship model of the subsystem is constructed and the calculation of the health status of the wind turbine operating state is established.Through the discussion of the two sets of SCADA data,the influence of sliding window width and step size on the calculation results of health indicators is analyzed.At last,the appropriate window width and step size are got,as the scale and sampling frequency of data set under the condition of minimum calculation time and variance.In the second step,an online warning method for the operating state of wind turbines is proposed based on the principle of small probability events.The randomness analysis of the operating state health indicator is carried out on the wind turbine.The normality test methods such as Kolmogorov-Smirnov test,Lilliefors test and Shapiro-Wilk test are used to verify that the probability distribution of health indicators did not obey the normal distribution.Based on the kernel density estimation method,a random distribution model of wind turbine operating state health indicators is established.Based on the principle of small probability events,an online warning method for wind turbine operation status is proposed and the early warning threshold is quickly analyzed and calculated by 0.618 golden section method.Since the early warning analysis process only depends on the SCASA data of the normal operation state of the wind turbine,it avoids the difficulty of realizing the early warning of the wind turbine operating state in the absence of abnormal operating state data.In the last step,online warning software system is developed.Qt creator integrated development environment is used to realize software development for wind turbine operating state identification and early warning.Software are designed from data sources,data analysis,data visualization and early warning data management,etc.Data acquisition and online analysis during the operation of the wind turbine are realized by connecting the database of the SCADA system.The update of the normal wind turbine operating state parameter model is facilitated by completing the basic framework of the software and designing the relevant parameter setting window.Using the rich control class provided by Qt creator,the real-time analysis of the wind turbine operating state health indicator changes and the real-time running status of each subsystem are displayed in the software corresponding interface.The wind turbine operating state is displayed in the form of a graph.