The Nonlinear Variable Selection and Its Real-time Detection of Lead-acid Battery Capacity

Real-time detection of lead-acid battery capacity is hard for not having the suitable instrument. The soft sensing method is considered. It is studied on the basis of variable selection, the confirming of structure and parameters for the predicted model. The importance of each related nonlinear secondary variable is computed and sorted with RReliefF method. After that, the secondary variable sets are selected in turn. Soft sensing model with neural network is finally confirmed by cross-validation method. Simulation results show that the established model is effective with mean square error is 0.0011 for the testing data. It provides one way of simplifying the real-time detection of battery capacity.