Parameter Optimization of Hybrid Drive Trains

Motivation

In recent years, many automobile manufacturers have built prototypes of hybrid drive train passenger cars. Some hybrid car concepts are already in series production. The reasons for the success of passenger cars with hybrid drive trains lie in the synergy effects of internal combustion engine propulsion and electric motor propulsion.

By using fossil fuels with a high energy density as primal energy carrier, long driving distances and low refueling time are ensured. The disadvantages of any internal combustion engine are its poor efficiency at low loads as well as its lack of an ability to recuperate energy during negative-torque driving situations. Using an electric machine as a motor or generator and an electric storage system in addition to the internal combustion engine, these disadvantages can be eliminated.

The combination of the two propulsion systems results in additional degrees of freedom for the design of the drive train. The scope of this research is finding a generic method for deriving an optimal configuration of the hybrid drive train with respect to well-defined constraints and optimization objectives.

Optimization Objectives and Constraints

In order to optimize the hybrid drive train configuration it is necessary to define the optimization objectives. The most important objective is the fuel consumption of the vehicle. However, this project focuses on combining fuel consumption with other optimization objectives, such as battery lifetime and emissions. Some of these objectives contradict each other to a certain extent, so that it is necessary to use appropriate weighting factors for each objective. These objectives are then included in the integral cost function of the supervisory control algorithm.

The supervisory control algorithm also has to take into account the predominant way the vehicle will be used. For example, a passenger car that is mainly used for city driving will be optimized for a typical driving cycle such as the “FTP-city” cycle. Such a cycle will be modeled using a quasi-static ‘backward’ simulation. Additionally, different peak power definitions (such as acceleration from 0 to 100 km/h) can be used as constraints for the choice of the component dimensioning. This generalized setup is intended to lead to a customer-oriented drive train design.

Structure, Dimensioning of Components and Supervisory Control

For hybrid electric vehicles (HEV), various structures are possible for combining electric machine, internal combustion engine and electric energy storage. For each structure (i.e. parallel hybrid, complex hybrid and mild hybrid), it can be shown that the aforementioned constraints form a convex subset of admissible combinations for the dimensioning of the drive train components.

For every chosen configuration, it is possible to determine the optimal supervisory control algorithm for a given cycle using the non-causal dynamic programming (DP) method. Consequently, there is a global optimal solution for the configuration if the cost function is convex with respect to the parameter space.

Such a parameter optimization is shown in FIGURE 1 for a compact car with a parallel hybrid drive train. Here, the only optimization objective is the fuel consumption. It was determined for each configuration using an ECMS supervisory control algorithm, embedded in a quasi-static simulation. The colored areas represent the convex subset of admissible configurations for the chosen power requirements.

Goal

The individual optimization of structure, component dimensioning and supervisory control cannot lead to a globally optimal configuration of the drive train. This is due to the strong interdependency of the three aspects.

Therefore, a more universal approach is necessary. The goal of this project is to find a unified strategy for the parallel optimization of structure, dimensioning of components and (causal) supervisory control.

Methods

The techniques used for obtaining this strategy will be dynamic programming, quasi-static simulation, convex optimization methods, among others.