Abstract: Previous Wind Turbine(WT)generation forecasting methods rarely considered the influence of its health status,which reduced the accuracy of the prediction.Therefore,a novel power generation quantity forecasting model for WT considering the health status degradation process was proposed.The combined weights were introduced to describe the correlation between multiple performance observation sequences of WT,and the health status model based on the newly-modified multi-observation sequence HMM was put forward.By using Hidden Markov Model(HMM),the health status degradation process was evaluated by the health status rank sequence and the degradation probability respectively.Using the pre-processed wind speed and power data from Supervisory Control and Data Acquisition(SCADA)system,Wind Turbine Power Curve(WTPC)model under different health status was constructed with Bins method.Health status degradation process was integrated into WTPC in time series to establish Wind Turbine Dynamic Power Curve(WTDPC)model by considering health status.Combined with Weibull-distribution-based wind speed model,WTPDC was used to predict the power generation.The proposed model was validated by SL2000/100 WT and the results demonstrated its effectiveness.

Key words: health status evaluation, multi-observation sequence hidden Markov model, wind turbine power curve, power generation forecast

摘要： 针对现有风电机组发电量预测方法很少考虑机组健康状态的影响,从而降低了预测准确性的问题,提出一种考虑健康状态的风电机组发电量预测模型。首先引入组合权重描述机组多个性能观测序列之间的相关性,建立基于改进多观测序列隐Markov模型(HMM)的健康状态模型,分别用健康状态等级序列与劣化概率描述健康状态劣化过程。采用预处理后的机组风电场数据采集与监控(SCADA)系统风速和功率数据,利用比恩法建立不同健康状态下的风电功率曲线模型。将机组健康状态劣化过程按时间序列融合到风电功率曲线,建立考虑健康状态的风电机组动态功率曲线模型,并结合基于Weibull分布的风速模型实现发电量预测。以SL2000/100风电机组发电量预测为例,验证了该方法的有效性。

关键词: 健康状态评估, 多观测序列隐Markov模型, 风电功率曲线, 发电量预测