Key Technologies For Operational Condition Monitoring And Health Maintenance Of Wind Turbines

While reducing operation and maintenance costs has become a key factor in the long-term healthy development of wind energy.Aiming at the problems of wind turbine power generation lower than the rated value,the unclear comprehensive performance of the whole machine,and sudden occurrence of abnormal events of key components,this paper analyzes the operation status of a wind turbine from three perspectives:"power generation performance of the whole machine","comprehensive performance of the whole machine"and "key subsystem"."This paper analyzes the key technologies of wind turbine operating condition monitoring and health maintenance from three perspectives to improve the operating condition of wind turbines.The following are the paper’s primary research components.1.To address the issue of wind energy’s random volatility resulting in a wide range of output power variation,masking the trend of the unit’s power generation performance,a standard power curve-based method for evaluating the whole unit’s power generation performance is proposed.Using rated power as a reference,the output power is converted to instantaneous efficiency,providing a standardized standard for evaluating power generation efficiency at various time points.Since instantaneous efficiency has a different distribution in different wind speed zones,the quantile approach is used to grade it to meet the same rating criteria in all wind speed zones.The generation performance of the unit is quantitatively evaluated online by developing a generation index model to assess current-generation performance and a generation potential model to assess the unit’s generation upside within the achievable range.2.For the problem of unclear comprehensive performance of the whole machine and wide thresholds of operating parameters,this paper proposes a model for evaluating and optimizing the entire machine’s total performance based on the benchmark space.According to the loss mechanism of energy in the wind energy conversion stage,this paper determines the index system to evaluate the overall performance of the whole machine.For the problem of a large amount of SCADA data and lack of labelling,this paper uses an unsupervised algorithm to quantify the similarity of operation states consisting of multidimensional variables.By analyzing the distribution characteristics of the similarity of operating conditions,this paper determines the benchmark space of wind turbines.By quantifying the similarity between the online data and the fixed benchmark space,the overall comprehensive performance of the wind turbine is determined.To improve the turbine’s performance,the operation optimization model of the wind turbine is established using the reference space to adjust the operation parameters.3.To address the problem that some actions of the yaw system have no actual output,but these actions reduce the stability of the yaw system,this paper proposes the idea of using a data mining approach to minimize the invalid yaw actions.The multidimensional variables associated with the start and stop of the yaw system are extracted by considering the effects of the continuity of yaw actions and the randomness of wind energy.A decision tree algorithm with good adaptability to noisy data is used to filter the multidimensional variables to reduce the input parameters of the yaw optimization model.a yaw system optimization model is developed based on historical data to reduce the number of yaws and yaw times.4.For the problem that the anomalous characterization of key subsystems of wind turbines is not evident in the early stage and only appears in particular wind speed intervals,this paper proposes an anomaly identification model that integrates thermal characteristics and data mining.Based on the thermal factors of key components,this paper identifies the mechanism of variation of essential parameters during thermal equilibrium.To match the rate of change of key parameters and the rate of change of wind speed in the same period,this paper uses the intra-regional extreme value method and data smoothing method to obtain the rate of change of wind speed and key parameters in different time scales.Anomaly identification models of key subsystems are constructed using average historical data and the change rates of parameters.By making the abnormality identification model of each key parameter simultaneously,this paper completes the localization of abnormal equipment and determines the cause of the abnormality,which provides guidance for the operation and maintenance of wind turbines.