Title:	Simulation-Based Optimization of a Series Hydraulic Hybrid Vehicle using Evolutionary Optimization Algorithms
DNr:	SNIC 2018/3-573
Project Type:	SNIC Medium Compute
Principal Investigator:	Petter Krus <petter.krus@liu.se>
Affiliation:	Linköpings universitet
Duration:	2018-12-01 – 2019-04-01
Classification:	20399
Keywords:

Abstract

Hydraulic hybrid vehicle transmissions combine a primary power source, typically a combustion engine, with a hydraulic circuit to offer the possibility of energy recuperation and prime mover operating point modulation to ultimately, improve energy efficiency or reduce emissions. As alternative to electric hybrid transmissions they are of interest for vehicles which have an existing hydraulic circuit, or are heavy enough to benefit from the hydraulic components’ higher power density. Due to the characteristic discharge profile of the hydraulic energy storage, it is beneficial to already consider control aspects in the early transmission design process. Here, dynamic simulation models are useful to capture the effects of control strategies on the operation of the transmission and to evaluate the general impact of variables onto the final design. Through simulation-based optimization (SBO), component sizing and the parameterization of a simple control strategy can be addressed simultaneously to explore and capture trade-offs and effects of different problem formulations. Due to relatively high computational loads, efficient optimization algorithms are important to provide sufficient quality of results at reasonable computational costs. Therefore, derivative-based algorithms with a multi-start approach are typically used when aiming for fast results, whereas evolutionary algorithms provide a better quality of results. Previously, SBO was conducted with non-derivative-based, non-evolutionary algorithms (Complex-RF and variants) on hardware less powerful than the NSC. The results were compared with regard to performance and robustness of results amongst themselves, and also to a limited number of experiments with a more computationally intensive evolutionary algorithm (particle swarm optimization, PSO). This work was submitted to the Journal “Engineering Optimization” as part of a manuscript addressing different angles of post-optimization analysis, and was received positively by the reviewers. The editor responsible for the publication, however, strongly encouraged to extend the study with regard to the number of optimizations included for the PSO, and also the evolutionary optimization algorithms included in the comparison. An initial, small-scale local NSC project to test operation of the simulation software, its integration with optimization algorithms under external software and basic scaling experiments showed the feasibility of the intended use of the NSC resources. This project aims to conduct all necessary optimization experiments in fairly rapid succession. Results gained from the use of the NSC resources will yield a more solid scientific foundation for the comparison of optimization algorithm performance aspects in the publication currently in the review process by including “state-of-the-art” evolutionary optimization algorithms as benchmark. The implementation can serve to illustrate the benefits and limitations of the computationally less expensive Complex-RF(P) algorithms and the use of free fluid power software (Hopsan) in combination with powerful computational resources.