Topology Considerations in Hybrid Electric Vehicle Powertrain Architecture Design.

Abstract

Optimal system architecture (topology or configuration) design has been a challenging design problem because of its combinatorial nature. Parametric optimization studies make design decisions assuming a given architecture but there has been no general methodology that addresses design decisions on the system architecture itself. Hybrid Electric Vehicle (HEV) powertrains allow various architecture alternatives created by connecting the engine, motor/generators and the output shaft in different ways through planetary gear systems. Addition of clutches to HEV powertrains allows changing the connection arrangement (configuration) among the powertrain components during the vehicle operation. Architectures with this capability are referred to as multi-mode architectures while architectures with fixed configurations are referred to as single-mode architectures. HEV architecture optimization requires designing the powertrain’s configuration and its sizing simultaneously. Additionally, evaluation of an HEV architecture design depends on a power management (control) strategy that distributes the power demand to the engine and motor/generators. Including this control problem increases the complexity of the HEV architecture design problem.

This dissertation focuses on a general methodology to make design decisions on HEV powertrain architecture and component sizes. The representation of the architecture design problem is critical to solving this problem. A new general representation capable of describing all architecture alternatives is introduced. Using the representation, all feasible configurations are generated where these feasible configurations are used to create single- and multi-mode HEV architectures. Single-mode and multi-mode architecture design problems considering fuel economy, vehicle performance and architecture complexity are formulated separately and solution strategies are developed. The high complexity of the resulting optimization problem does not allow us to claim true optimality rigorously; therefore, the terms ``promising" or ``near-optimal" are more accurate in characterizing our results. The results show that different architectures must be designed for different applications. The case studies designing architectures for some available vehicles from the market find the architectures already implemented in these vehicles under some design constraints. Alternative architectures that improve these designs under different design constraints are also demonstrated. Architectures for a new application that is not available in the market are also designed.