The Prediction Of Remaining Useful Life Based On Data-Driven For Storage Battery Of Unmanned Submarine

Storage battery is an important energy device of unmanned submarine. It is also an important electrical power system of unmanned submarine. Tracking and predicting the health status of storage battery is an important method to ensure the navigation safety of unmanned submarine. It can provide plenty of time for preventing and excluding faults by predicting the remaining useful life (RUL) of storage battery exactly, which can avoid the disasters happened on unmanned submarine to some degree.There has been a long period of rapidly developing the data-driven prognostics method for estimating the RUL of storage battery for unnamed submarine, therefore, there are still many problems unsolved. The work which has been done is presented a hybrid prognosis approach of grey model (GM) and failure function model for the estimation of the RUL of storage battery. The major research work is as follows:In order to solve the problem of system uncertainty in determined degradation state, a hidden Markov model (HMM) has been built due to its well capability in pattern classification. HMM is used to identify the degradation state of system. Then an aging factor is introduced to characterize the deterioration of system, thus the conditional reliability function and RUL model were obtained, in which the aging factor was taken as the covariance.To overcome the problem of data scarcity, a hybrid prognosis approach of grey model and failure rate function model are presented for prediction of engineering asset health. Failure rate function allows modeling the average remaining useful life of each observation and therefore is capable of prognosis. After that, taking the average remaining useful life of each observation of the system as a sequence of grey dynamic systems to build the grey prediction model, then, the RUL of next observation is obtained by predicting the degradation of the system. Finally, the method is applied to typical real experimental data related to the accelerated life tests of storage battery.