Wind farm maintenance optimisation strategy considering MFTS distance‐based K‐medoids clustering and mixed Weibull model

Abstract This paper proposes a wind‐resource‐weighted maintenance optimization framework to reduce operation and maintenance (O&M) costs and improve power generation efficiency. Existing clustering methods based on single distance metrics cannot fully capture the multidimensional temporal characteristics of wind‐speed data, while single Weibull models may provide insufficient fitting accuracy under low‐wind‐speed conditions. Conventional opportunistic maintenance strategies may also result in inappropriate maintenance timing and increased costs. To address these issues, the proposed framework integrates wind‐resource heterogeneity characterization, probabilistic resource assessment, and maintenance optimization. First, a wind‐speed clustering method combining the multi‐feature time‐series fusion (MFTS) distance with an optimized K‐medoids algorithm is developed to identify heterogeneous wind‐resource regions. Second, a cluster‐based mixed Weibull model is established to characterize wind‐resource distributions across different wind directions. Finally, a wind‐resource‐weighted maintenance strategy is developed, using wind‐resource information to identify critical turbines and component degradation states as the direct basis for maintenance triggering and timing optimization. Unlike conventional strategies primarily based on degradation states or failure probabilities, the proposed framework incorporates spatial wind‐resource differences into maintenance prioritization and intervention timing. Simulation results demonstrate its effectiveness in identifying heterogeneous wind‐resource patterns, improving resource characterization, and reducing maintenance costs under the investigated wind farm conditions.