Online Parameter Tuning Methods for Adaptive Ecms Control Strategies in Hybrid Electric Vehicles

This paper presents two adaptive ECMS schemes which compared to the standard ECMS formulation are extended by an online learning algorithm based on the so-called regret minimization paradigm. While in the first approach the Shrinking Dartboard algorithm is applied to tune the parameter values of the ECMS adaptation rule, in the second approach the Weighted Fractional algorithm averages the results of numerous ECMS strategies in order to obtain an optimal power split. Thereby, an expert represents a particular parameter setting of an ECMS adaptation rule which is assigned to an ECMS strategy. The simulation results obtained when ECMS strategy is used in combination with the online learning procedures are compared to the results obtained when no method for parameter tuning is applied and when the parameters of the adaptation rule are set to constant values for the entire driving cycle. In both cases an adaptation law (P or PI) is employed in order to obtain the equivalence factor in the ECMS strategy in each time step. The proposed adaptive control schemes achieve good performance even if the driving conditions are not known a-priori. Simulation results indicate that when applying the ECMS approach together with regret minimization procedures reduced fuel consumption can be achieved compared to the case when constant parameters are used in adaptation laws of the ECMS strategy.