Online dynamic working-state recognition through uncertain data classification

The satellite must continue working properly under different environments and loads. power system is an essential component. Due to tasks, loads, attitudes, a has many diverse states. Therefore, it necessary accurately recognize the state online for fault diagnostics health management. However, measurement errors, environmental noise, interference, other uncertain factors, output voltage value of levels uncertainties. If these uncertainties various states are not considered, recognition results can be low quality. To address this problem uncertainty we present dynamic working-state systems based on data classification . In system, first explore uncertain-data clustering center model state. Then, with slide-window processing strategy, compute distances between cluster centers online. Thus, obtain more accurate results. evaluation real demonstrate that presented valid applied system.