A health status assessment model for hydropower units integrating KPCA feature extraction and adaptive weighting

Abstract The effectiveness of health status assessment (HSA) methods for hydropower units (HUs) is crucial for accurately determining the operating status. Traditional weighting methods often have limitations that can cause assessment results to deviate from the actual unit status. This paper proposes a novel adaptive weighting method. In this method, an HSA model is constructed and optimized based on multiple known datasets corresponding to different operational states of units. The optimized model is then applied to assess the health status of HUs, supporting the formulation of maintenance strategies. First, evaluation indicator data are processed to derive a health assessment index (HAI). Initial weights are assigned to the indicators. Then, using genetic algorithms as the optimization method, the difference between the unit health assessment score derived from the fuzzy comprehensive evaluation and the corresponding actual status value is employed as the objective function to optimize the weights of the indicators, obtaining an optimal weight combination. This method is applied to the health assessment of a specific HU. Results show that the constructed HAI enabled fault detection 93 h earlier than the actual incident report. Compared to traditional weighting methods, the proposed adaptive weighting method minimizes the assessment error, resulting in a more accurate reflection of the actual unit status.