风电机组故障智能诊断技术及系统研究

风电机组的状态监测和故障诊断是保证机组长期稳定运行和安全发电的关键。基于风电机组的基本结构,介绍了机组的故障类型和机理,论述了实际应用中机组的状态监测和故障诊断技术;基于BP神经网络的原理和优点,深入讨论了如何应用人工神经网络构建风电机组智能诊断系统,并给出了可行的系统设计方案和软件实现流程图。     

基于风电机组状态的超短期海上风电功率预测

提出一种基于风电机组状态的超短期海上风电功率预测模型。首先,综合考虑海上环境因素以及风电机组部件间的相互作用建立指标的预测模型,以长短期记忆神经网络的预测误差作为监测指标的动态劣化度;然后采用模糊综合评价法对风电机组的运行状态进行评估,依据评估结果对风电机组历史运行数据进行划分;最后根据分类后历史运行数据建立基于机组状态的超短期风电功率预测模型。结合国内某海上风电场实例数据进行分析,算例结果表明所提方法可有效提高风电功率预测精度。     

风电机组振动监测系统对齿轮箱故障的检测与诊断

着重介绍了本特利内华达ADAPT.wind风电机组在线状态监测系统边带能量率的算法原理,并以在风电机组齿轮箱故障诊断中的应用实例,充分说明其对风电场运营业主提高风电机组管理水平和运营效率的有效性。     

基于卷积神经网络的风电机组齿轮箱状态监测方法

为了实现对风电机组齿轮箱的状态监测,文章提出了一种基于卷积神经网络的风电机组齿轮箱状态监测方法。首先,提取风电机组数据采集与监视控制(SCADA)数据和振动信号作为参数,组成齿轮箱状态矩阵。其次,建立了一种卷积神经网络模型,该模型针对输入数据设计了特定结构和池化层规则,提高了计算效率,能够从齿轮箱状态信息中提取特征并判断其状态。最后,利用实际运行的风电机组数据对卷积神经网络模型进行了训练和验证,最终取得了96.3%的识别精度。同时,将该模型应用于对同一风场其他机组的状态监测,结果验证了卷积神经网络模型对齿轮箱状态监测的有效性。     

基于数据驱动的风电机组状态评估方法研究

为了延长风电机组平稳运行时长和减少故障停机次数,文章基于有监督主成分分析(SPCA)的Hotelling-T~2和Q统计量控制图,提出了一种风电机组状态监测与评估方法。首先,根据风电机组SCADA历史数据提取正常状态数据。然后,训练集成学习模型拟合主要状态变量,采用贝叶斯优化算法优化其中的超参数。最后,在移动时间窗内利用SPCA方法将监测数据分解到主成分空间与残差空间,计算真实数据与参考状态数据的Hotelling-T~2和Q统计量,并同时求取两种统计量的斯皮尔曼系数,通过划定阈值对机组进行状态评估。将该方法用于某风电场1.5 MW级风电机组,结果表明,该方法能够有效地对机组当前状态进行监测并识别出功率输出故障。     

风电场风电机组机载风速仪状态自确认

以风电机组机载风速仪为例,提出一种传感器状态自确认方法。利用多台风电机组风速的相关性,通过动态时间规整算法,选定一组风电机组群。构建基于自联想神经网络的风电机组群风速仪预测模型,采用历史正常数据通过麻雀搜寻优化算法对模型进行训练,根据实际值与预测值的关系对风速仪状态进行识别。通过仿真实验证明该方法可识别风速仪模拟异常状态,最后对某风场实际风速进行检测,结果显示能有效识别出风速仪的状态,实现风电机组风速仪状态的自确认。     

基于征兆驱动和专家推理的水电机组轴承状态分析

导轴承是维持水电机组主轴在轴承间隙范围内稳定运行的重要部件,其工作性能的好坏将对机组安全稳定性产生重要影响。为准确监测水电机组轴承运行状态,提出了一种基于征兆驱动的轴承状态诊断模型,通过建立相应的基础数据库,运用时域和频域分析方法提取故障征兆,并结合故障诊断专家推理机制实现对轴承状态的分析诊断。应用表明,该方法能对轴承运行状态实时集中监测,同时实现对轴承故障的及时诊断,可为电站机组的运行检修提供参考。     

基于线路态势感知的台区检修决策优化技术研究

传统方式下低压台区采用计划检修以及事后抢修的模式,运维检修决策模式固化、滞后的弊端明显。以当前低压配电网数据为基础,以用户回路阻抗计算为手段,提出台区线路状态感知评估模型,支撑台区检修模式转型。基于台区简化网络拓扑构建用户回路阻抗方程,并采用约束线性最小二乘模型进行求解;考虑线路阻抗状态与故障类型的关联关系,创新性提出台区线路随机故障与累积故障状态判定模型;构建考虑检修成本的经营成效评估模型,并以经营成效最大化为目标,结合供电可靠性约束条件,建立基于非线性规划方法的台区检修决策优化模型。以某个含376个公变台区的供电所为例,进行了算例分析,验证了所提模型与算法的有效性。     

基于改进深度森林算法的风电机组故障诊断技术研究

为了及时有效地检测出风电机组发生的具体故障,同时克服传统故障诊断方法的局限性,文章提出一种基于改进深度森林算法的风电机组故障诊断方法。首先,利用有效的数据预处理方法处理SCADA原始数据并提取故障特征;然后,基于深度森林算法对风电机组具体故障进行诊断,同时,针对深度森林算法在故障诊断领域存在的缺陷,对算法提出改进;最后,利用河北某风场1.5 MW风电机组实际运行数据对文章提出的故障诊断算法进行验证,通过正确率、AUC等指标验证了所提故障诊断算法相比传统机器学习算法的有效性和优越性。该研究为风电机组运行和维修提供了依据,同时也为故障诊断领域提供了新的方法和思路。     

永磁同步风电机组发电机定子温度建模方法

针对山地风电场大型永磁同步风电机组发电机定子温度的变化特点及传热过程进行分析,提出一种基于风电机组SCADA数据的发电机定子温度建模方法。首先,分析风电机组在实际运行过程中定子温度的变化情况;然后,根据永磁同步发电机内部的传热过程,对发电机定子温度建模;最后,采用风电机组正常运行时的SCADA数据求解模型参数。实例分析表明,风电机组状态正常时,定子温度估计的平均误差绝对值为0.59 ℃,模型精度较高;风电机组状态异常时,模型温度估计的平均误差绝对值为5.17 ℃,精度显著降低。     

某风电场风力发电机组故障诊断

为了准确判断风电机组的运行状态及故障,提出了基于常规分析—振动幅值分析—波形频谱分析的故障诊断流程,阐述了针对风电机组的幅值分析方法和波形频谱分析方法,并通过对某机组异响的根源探究实例,准确地诊断出机组异响来源于齿轮箱太阳轮,可为风电机组故障诊断技术提供依据。     

基于SOM网的风电变流器故障诊断

我国新疆、甘肃、宁夏、内蒙、浙江、黑龙江、江苏、广东等都在大规模建设风电场,这些风电场建成后,其故障维护就有了很大市场.以新疆风电场为基础,尝试开发用于风力机故障智能诊断的系统.首先介绍了风力机及其变频器系统的结构,分析了变频器的故障机理.使用SOM神经网络对风机变流器进行了诊断,用数据验证了诊断结果.把传统的电力电子设备故障诊断技术与新疆风力机变频器的故障诊断相结合,为风电大面积推广应用产生了积极作用.     

基于数据融合的风电变桨系统故障预警研究

随着风电机组装机容量的不断攀升,同时带来并网发电率低、机组故障率高等缺点,导致风电机组整体利用率较低。为此提出一种基于数据融合的风电变桨系统故障预警方法。首先结合SCADA系统中的运行统计信息和历史数据,采用Relief特征参数选择方法筛选出与风电变桨系统故障相关的特征参数;然后采用数据融合的方法,建立基于MSET技术的风电变桨系统故障预测模型,并采用滑动窗口法进行故障预警阈值的确定;最后以上海某风场1.5 MW双馈异步风电机组进行实例分析,结果表明该方法可提前发现风电变桨系统故障征兆,实现对风电变桨系统的早期故障预警。     

Response analysis and comparison of a spar‐type floating offshore wind turbine and an onshore wind turbine under blade pitch controller faults

This paper analyses the effects of three pitch system faults on two classes of wind turbines, one is an onshore type and the other a floating offshore spar‐type wind turbine. A stuck blade pitch actuator, a fixed value fault and a bias fault in the blade pitch sensor are considered. The effects of these faults are investigated using short‐term extreme response analysis with the HAWC2 simulation tool. The main objectives of the paper are to investigate how the different faults affect the performance of wind turbines and which differences exist in the structural responses between onshore and floating offshore wind turbines. Several load cases are covered in a statistical analysis to show the effects of faults at different wind speeds and fault amplitudes. The severity of individual faults is categorized by the extreme values the faults have on structural loads. A pitch sensor stuck is determined as being the most severe case. Comparison between the effects on floating offshore and onshore wind turbines show that in the onshore case the tower, the yaw bearing and the shaft are subjected to the highest risk, whereas in the offshore case, the shaft is in this position. Copyright © 2014 John Wiley & Sons, Ltd.     

Vibration‐based wind turbine planetary gearbox fault diagnosis using spectral averaging

Planetary gearboxes (PGBs) are widely used in the drivetrain of wind turbines. Any PGB failure could lead to a significant breakdown or major loss of a wind turbine. Therefore, PGB fault diagnosis is very important for reducing the downtime and maintenance cost and improving the safety, reliability, and lifespan of wind turbines. The wind energy industry currently utilizes vibratory analysis as a standard method for PGB condition monitoring and fault diagnosis. Among them, the vibration separation is considered as one of the well‐established vibratory analysis techniques. However, the drawbacks of the vibration separation technique as reported in the literature include the following: potential sun gear fault diagnosis limitation, multiple sensors and large data requirement, and vulnerability to external noise. This paper presents a new method using a single vibration sensor for PGB fault diagnosis using spectral averaging. It combines the techniques of enveloping, Welch's spectral averaging, and data mining‐based fault classifiers. Using the presented approach, vibration fault features for wind turbine PGB are extracted as condition indicators for fault diagnosis and condition indicators are used as inputs to fault classifiers for PGB fault diagnosis. The method is validated on a set of seeded localized faults on all gears: sun gear, planetary gear, and ring gear. The results have shown a promising PGB fault diagnosis performance with the presented method. Copyright © 2015 John Wiley & Sons, Ltd.     

Validation of a double fed induction generator wind turbine model and wind farm verification following the Spanish grid code

Wind turbine manufacturers are required by transmission system operators for fault ride‐through capability as the penetration of wind energy in the electrical systems grows. For this reason, testing and modeling of wind turbines and wind farms are required by the national grid codes to verify the fulfillment of this capability. Therefore, wind turbine models are required to simulate the evolution of voltage, current, reactive and active power during faults. The simulation results obtained from these wind turbine models are used for verification, validation and certification against the real wind turbines measurement results, although evolution of electrical variables during the fault and its clearance is not easy to fulfill. The purpose of this paper is to show the different stages involved in the fulfillment of the procedure of operation for fault ride‐through capability of the Spanish national grid code (PO 12.3) and the ‘procedure for verification, validation and certification of the requirements of the PO 12.3 on the response of wind farms in the event of voltage dips’. The process has been applied to a wind farm composed of Gamesa G52 wind turbines, and the results obtained are presented. Copyright © 2011 John Wiley & Sons, Ltd.     

Dynamic modelling and control of fully rated converter wind turbines

Today, many countries are integrating large amount of wind energy into the grid and many more are expected to follow. The expected increase of wind energy integration is therefore a concern particularly to transmission grid operators. Based on the past experience, some of the relevant concerns when connecting significant amount of wind energy into the existing grid are: fault ride through requirement to keep wind turbines on the grid during faults and wind turbines have to provide ancillary services like voltage and frequency control with particular regard to island operation.While there are still a number of wind turbines based on fixed speed induction generators (FSIG) currently running, majority of wind turbines that are planned to be erected are of variable speed configurations. The reason for this is that FSIG are not capable of addressing the concern mentioned above. Thus, existing researches in wind turbines are now widely directed into variable speed configurations. This is because apart from optimum energy capture and reduction of mechanical stress, preference of these types is also due to the fact that it can support the network such as its reactive power and frequency regulation. Variable wind turbines are doubly fed induction generator wind turbines and full converters wind turbines which are based on synchronous or induction generators.This paper describes the steady state and dynamic models and control strategies of wind turbine generators. The dynamic models are presented in the dq frame of reference. Different control strategies in the generator side converter and in the grid side converter for fault ride through requirement and active power/frequency and reactive/voltage control are presented for variable speed wind turbines.     

Dynamic response analysis of wind turbines under blade pitch system fault,grid loss,and shutdown events

This study focuses on the dynamic responses of land‐based and floating wind turbines under blade pitch system fault and emergency shutdown conditions. The NREL 5 MW turbine is studied. A hydraulic pitch system is considered, and the faults under study are events with a seized blade or a blade running out of control. Emergency shutdown is defined as a fast pitch‐to‐feather maneuver of the blades. Load cases with power production and grid fault with ensuing shutdown are also analysed for comparison. The fault scenarios and the blades' fast pitching activity are simulated using HAWC2 through external Dynamic Link Libraries. On the basis of the time‐domain simulations, the response characteristics of the land‐based and the floating turbines in the four design load cases are compared. The load effects from the fault conditions are compared with the operational cases. Strong system dynamics and resonant responses, such as the tower elastic mode and the yaw resonant response, are elicited during shutdown. If the pitch system has a fault and one blade is hindered from normal pitching, the uneven load distribution of the blades leads to large structural and motion responses. For both turbines, the response maxima vary cyclically with the instantaneous azimuth when the blades start pitching to feather. For the floating wind turbine, the interaction of waves and wind also affects the results. The effect of the pitch rate during shutdown is analysed. The responses of the land‐based turbine in grid loss and shutdown conditions are proportional to the pitch rate, whereas decreased sensitivity is found in the cases with pitch system faults. For the floating turbine, the effect of the pitch rate is small, and reduced pitch and yaw motion extremes are observed as the pitch rate increases. Copyright © 2013 John Wiley & Sons, Ltd.     

风速分布差异对永磁风力发电机组故障特性的影响分析

为了更有效地辨识永磁风力发电机组的故障特征,从叶轮扫掠面内的实际风况出发,综合考虑风剪切、塔影效应产生的风速时空分布差异,并研究该差异对永磁风力发电机组故障特征的影响。首先介绍考虑风速时空分布差异后的风速模型（等效风速模型）,然后基于该模型对永磁风力发电机组常见的电气和机械故障特性进行理论分析,分别推导得出考虑风速时空分布后绕组不对称故障、叶轮不平衡故障以及这2种故障复合作用下的定子电流表达式;最后利用Matlab/Simulink平台进行仿真验证。研究结果表明,与现有的基于轮毂处平均风速的故障特征相比,考虑风速时空分布差异后,永磁风力发电机绕组不对称故障特征中含有3kP(k为整数,P为叶轮转频)为主的调制频率;叶轮不平衡故障特征频率除P和3kP外,还出现了P与3kP的耦合调制;复合故障下频率成分会更复杂。