Hierarchical Cluster Balance Representation and Temporal-Spatial Collaborative Modeling for Nonstationary Wind Turbine Monitoring

Reliable monitoring of the wind power generation process is considerably challenging in the following two aspects. Temporally, the process exhibits significant nonstationary characteristics, posing difficulties in recognizing faults from wide-range varying variables. Spatially, the fault of wind turbines would be reflected in the seriously inconsistent behavior compared to other surrounding turbines, which have barely been discussed before. In this study, a hierarchical temporal-spatial collaborative (TSC) monitoring framework is proposed to provide holistic fault detection for the wind turbine cluster. On one hand, the temporal characteristic of each specific turbine is captured by utilizing the individual nonstationary variable partition (INVP) strategy. In this way, the inherent operation and nonstationary time-varying behavior can be analyzed and modeled separately. On the other hand, for the first time, the health statuses of the turbine cluster are described by cluster balance feature extraction (CBFE) strategy, providing the balance representation of the wind turbine cluster to recognize the inconsistent behavior of turbines from the spatial analysis perspective. Furthermore, to establish the monitoring of the whole turbine cluster and specific turbine individuals concurrently, we fuse the INVP and CBFE strategies and develop the hierarchical monitoring framework in a temporal-spatial collaborative way. The effectiveness of the proposed method is demonstrated by a real wind power generation process of a turbine cluster.