风电机组轴承健康状态评估和劣化趋势预测方法的研究

发布时间：2018-08-06 12:46

【摘要】：风电机组运行环境恶劣,受气候等多种不确定因素的影响,容易出现性能与状态劣化。关键部件一旦失效,检修时间较长,增加风电场的运维成本。轴承作为风电机组中关键的部件,其运行状况对整台设备的可靠性具有重要的影响。本文基于风电机组的监控与数据采集(Supervisory Control And Data Acquisition,SCADA)系统运行数据,从健康评估模型和劣化趋势预测两方面展开对轴承的状态的研究。建立轴承健康评估模型和趋势预测模型。本文以风电机组轴承温度为研究对象,考虑轴承温度受风速和功率的影响,以Bin方法进行划分工况,利用相对评价标准遴选轴承各工况的健康状态样本集;利用最小二乘拟合健康样本数据,形成轴承温度健康状态评估模型,在此基础之上,结合实际运行状态的上下阈值,引入劣化度概念。考虑到风电机组轴承劣化趋势非线性的问题,进一步应用时间序列神经网络,建立风电机组轴承的劣化趋势预测模型。以风电场实际数据为例,对模型进行验证并与其它模型进行比较。应用前面建立模型评估时,还存在风电机组轴承劣化趋势非稳定性的问题,这将会影响预测结果。于是在预测之前,应用EEMD(Ensemble Empirical Mode Decomposition)方法,将具有非平稳性特性的劣化趋势分解为一系列相对平稳的分量,利用时间序列神经网络对各分量开展预测,将所有分量的预测结果叠加得到最终的预测结果。研究结果表明,对于非线性数据,时间序列神经网络预测模型更具有一定的优势,精度得到提高,通常能满足风电机组轴承对监测参数的需求,对发现早期机组潜在的故障具有很好的实用性。对于具有较强非线性和非稳定的时间序列,本文的组合预测模型能更有效地跟踪风机轴承健康状态劣化趋势,且能明显提高预测的精度。
[Abstract]:Because of the bad operating environment and the influence of many uncertain factors such as climate, wind turbine is prone to performance and condition deterioration. Once the key components fail, the overhaul time is longer, which increases the cost of operation and maintenance of wind farm. As a key component of wind turbine, the operation condition of bearing has an important influence on the reliability of the whole equipment. In this paper, based on the monitoring and data collection of wind turbine, the operation data of (Supervisory Control And Data requirement SCADA system is collected, and the research on the state of bearing is carried out from two aspects: health assessment model and deterioration trend prediction. Establish bearing health assessment model and trend prediction model. In this paper, bearing temperature of wind turbine unit is taken as research object, bearing temperature is affected by wind speed and power, working conditions are divided by Bin method, and healthy state sample set of each condition of bearing is selected by relative evaluation standard. Based on the least square fitting of health sample data, an evaluation model of bearing temperature health state is established. Based on this model, the concept of deterioration degree is introduced in combination with the upper and lower threshold of actual operation state. Considering the nonlinear problem of bearing deterioration trend of wind turbine, a prediction model of wind turbine bearing deterioration trend is established by using time series neural network. Taking the actual data of wind farm as an example, the model is verified and compared with other models. The instability of bearing deterioration trend of wind turbine is also existed when the model is used to evaluate the model, which will affect the prediction results. Before prediction, EEMD (Ensemble Empirical Mode Decomposition) method is used to decompose the deterioration trend with non-stationary property into a series of relatively stationary components, and the time series neural network is used to predict each component. The final prediction results are obtained by superposing the prediction results of all components. The research results show that the time series neural network prediction model has some advantages and the accuracy is improved for nonlinear data, and it can usually meet the needs of the monitoring parameters of wind turbine bearings. It has good practicability to discover the potential fault of the early generation unit. For the time series with strong nonlinearity and instability, the combined prediction model in this paper can more effectively track the trend of deterioration of the health state of fan bearings, and can obviously improve the accuracy of prediction.
【学位授予单位】：华北电力大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM315

【参考文献】