Piecewise just-in-time data recovering and fault detection method for time-varying wind power generation process with missing data

The presence of missing values in the data poses challenges for fault detection tasks in wind power processes. The conventional data filling methods commonly focus on the process data with a single mode, disregarding the multimodal properties arising from time-varying characteristics in wind power processes. In this paper, to address the challenge of recovering data in time-varying and nonlinear wind power generation processes, a piecewise data recovering method with a just-in-time fault detection strategy is proposed. By utilizing the nonlinear matrix completion method to analyze data recovery ability, the process can be segmented into multiple modes, each exhibiting distinct characteristics. Then, using a just-in-time learning strategy, the missing value of the data can be accurately recovered in the corresponding mode in a short time. Finally, the kernel principal component analysis model is employed to monitor the process based on the recovered data. Compared with the existing methods, the proposed method considers the characteristics of nonlinear and high-rank or even full-rank matrix characteristics and recovers the missing data. In this way, it can achieve an accurate fault detection result for the wind power generation process with missing data. The effectiveness of the proposed method is validated on a real wind power generation process dataset.