To evaluate the performance of the powertrain architectures, modelling and simulation tools are indispensable. This is particularly true as prototyping and testing each design combination is cumbersome, expensive, and time consuming [26]. New hybrid powertrain configurations and controllers are also continuously being developed, thus the ability to simulate a powertrain before prototyping is important.
Simulating vehicle powertrain requires dedicated simulation software [26]. To that end, this research was inspired by the work carried out by Argonne Labs and the National Renewable Energy Laboratory (NREL) in the area of powertrain simulation and optimisation. Advanced powertrain researches from these two institutes have resulted in the creation of two simulation tools respectively; PSAT (PNGV System Analysis Toolkit) and ADVISOR (Advanced Vehicle Simulator)
Both tools have been frequently cited in literature for the purpose of system-level powertrain simulation and optimisation. Other powertrain simulation tools such as AVL Boost and GT-Suite are also available; however they have been noted to be more suited for simulation of detailed ICE attributes with the expense of greater computational time [58-60]. This includes the simulation of combustion mechanisms, exhaust after- treatment systems, and acoustics, which are not the intent of this research.
There are fundamental differences in the approach used by PSAT and ADVISOR; the former uses forward-facing models [61], whilst the latter uses a hybrid approach. In ADVISOR, the models are primarily backward-facing, with forward-facing methods only active when component performance limits are encountered; when they are not, ADVISOR operates strictly as a backward-facing model [32]. Another proponent of the backward-facing modelling method is the QSS toolbox, developed at ETH Zurich by a
team led by Lino Guzzella [27; 31; 62]. Unlike ADVISOR, the QSS toolbox is strictly backward-facing only.
Table 4 shows a list of reviewed publications that have used these simulation tools for the purpose of powertrain size optimisation. This table summarises the simulation tools employed and the types of powertrains that were analysed. Findings revealed that each simulation tool has a modular approach to powertrain modelling, and therefore provided the flexibility of simulating a wide variety of topologies. This literature survey also gave insights to the functional aspects of each simulation toolbox, which will be discussed further.
The development of PSAT was backed by the U.S. government [30] for the PNGV (Partnership for a New Generation of Vehicles) initiative. This initiative included a comprehensive forward-facing HEV simulation environment developed by a consortium of three U.S. automotive manufacturers: Ford, GM, and Daimler-Chrysler [61]. One fundamental strength of PSAT is the fact that it features modular implementation of powertrain components within a powertrain architecture. This provides the flexibility to scale the powertrain components, as well as replacing the models with different model blocks (such as proprietary blocks) if the need arises. This was made possible by strong reference to “power bonds” as seen in Bond Graph modelling techniques [63; 64].
In comparison, ADVISOR is a hybrid vehicle simulator which incorporates both forward-facing and backward-facing methods. ADVISOR compares the required values (backwards-facing results), with achievable values (forward-facing results). Nevertheless, this approach requires the definition of two models for each powertrain component, leading to larger programming overheads for introducing new components [65]. Finally, the QSS toolbox is a fully backward-facing model, and is capable of utilising a relatively larger time-step, generally in the order of 1 second.
Similar to PSAT, both ADVISOR and the QSS toolbox also follow a modular approach. The user can alter both the model inside the block as well as the MATLAB m-files associated with the block to suit their modelling needs. For example, the user may need a more precise model for the electric motor subsystem [26]. A different model can replace the existing model as long as the inputs and the outputs are the same. On the other hand, the user may leave the model intact and only change the MATLAB m-file associated with the block diagram. This is akin to choosing a different manufacturer of the same powertrain component. Therefore, all three software packages provide modelling flexibility for the user.
Table 4: Review of utilisation of powertrain simulation software in literature References [49] [66] [67] [68] [69] [27] [70] [71] [72] ADVISOR * * * * * PSAT * * * QSS Toolbox * * Single source * * Series Hybrid * * * * * Parallel Hybrid * * * * * Compound Hybrid * * CV * * HEV/PHEV * * * * * *
HEV – Fuel Cell * * *
An example of component modularity was described by Assanis et al., who used ADVISOR for simulating a hybrid powertrain, but required a customised engine model instead of the one supplied by ADVISOR. Because of its modularity, they could swap the engine module with one of their own, which included a higher fidelity turbocharged model. In contrast, if ADVISOR were to have a fixed list of components, it may cause some difficulties for the design engineer who desires to evaluate the impact of using a different, non-existing component, or wants to continuously vary component sizes in search for an optimum combination.
In addition to being modular, the powertrain components within ADVISOR are scalable. This was achieved by including routines that allow variation of component size through scaling of maps [30]. ADVISOR also follows an open-source model, and thus receives support from the industry and academia to validate and improve the model database [32]. Several publications have also used ADVISOR for powertrain simulation and validated its results favourably against real-world experiments [67; 73; 74], and ADVISOR itself has been used as a source for validation [75].
The key similarity in all three simulation tools is that they use MATLAB and Simulink as their underpinnings to run the simulations. According to Wipke et al., MATLAB and Simulink were chosen for their nearly self-documenting graphical programming environment and their wide acceptance by researchers in academia and industry [32].
However, though modular, the powertrain architecture in all three simulations tools was fixed during simulation. As a result, trying to compare the results from optimising different powertrain architectures, such as a pure EV with a PHEV, would require running two sets of optimisations separately. This is because the structure of the powertrain architecture is fixed during the optimisation run. For example, ADVISOR requires reconfiguration when comparing series and parallel hybrid, as shown by Same
et al. [38] for optimising a Formula Student vehicle.
A similar modular powertrain simulation tool with aggregated powertrain components was developed by Imperial College London in 2002 [65]. It also compares a few other backward-facing simulation tools, and mentioned that one common trait of these tools were the fixed powertrain layouts. This meant that each powertrain type had to be optimised individually. The paper above also mentions the importance of powertrain component modularisation for the purpose of comparing different powertrain architectures, thus strengthening the aims of this research.