A real time neural energy management strategy for a hybrid pneumatic engine*

Various energy management strategies for a hybrid pneumatic engine are reviewed and a real time neural control strategy proposed. This Neural Network strategy learns off line the optimal control given by Dynamic Programming and the resulting control model is applied on line. The diferent strategies are simulated with a backward vehicle model for various driving cycles and their fuel consumptions compared. The results show that the Neural Network strategy is better than a classical Equivalent Consumption Minimization Strategy (ECMS) and equivalent to a Variable Penalty Coefficient Strategy with Driving Pattern Recognition.