Real-time Cluster Analysis based Fault Detection for Dc Pulsed Power Load in the All-Electric Ship

Many of the new electronic loads being placed on Naval ships draw pulsed currents with a rapid and large transient from the dc power system. This creates difficulty in distinguishing between shunt faults and normal pulsed-power operation. This paper addresses the issues of real-time DSP implementation of a cluster analysis based fault detection algorithm. The data driven short-time Fourier transform based feature vectors corresponding to repetitive normal transients can be extracted and stored in the DSP during the training phase of the algorithm for any type of load. Once the statistical database is populated with those features, each subsequent occurrence of the transient event can be identified. Any shunt fault or disturbance that creates an unexpected event in the load profile will therefore be reliably diagnosed and a trigger can be sent for protective action. The performance, limitations and challenges of real-time implementation using a single core Texas Instrument DSP have been explored and discussed. The methods are demonstrated using an experimental setup and compared to the simulation results.