A Novel Adaptive Parallel Model with Knowledge-Aided Conversion Efficiency Assessment for Wind Turbine Condition Monitoring

Monitoring wind turbines is essential for their safe operation on wind farms. However, the majority of data-driven monitoring strategies do not consider expert knowledge. Consequently, they cannot simultaneously monitor statistical and physical characteristics and have poor monitoring results. This study proposes a knowledge-aided adaptive parallel monitoring strategy to monitor the process by evaluating the physical and statistical characteristics using two submodels. First, we propose a novel knowledge-aided monitoring statistic to characterize energy conversion efficiency, thus monitoring both the conversion efficiency using one of the submodels and the physical performance of the wind turbine. Subsequently, the process characteristics covering both steady and varying states can be monitored with another submodel using two monitoring statistics, which can accurately detect unusual behaviors from a statistical perspective. Generally, we can use three statistics to monitor the process from two perspectives. With the physical and statistical characteristics captured, we propose a novel adaptive monitoring strategy to adjust the model performance and accurately detect fault conditions. Real-world experiments demonstrate the effectiveness of the proposed method. Among the four monitoring methods, the monitoring strategy aided by knowledge showed the highest detection accuracy.