On Dynamic Programming Technique Applied to a Parallel Hybrid Electric Vehicle

Abstract: Lack of fossil fuel supplies as well as greenhouse gases e®ect on the environment, havemotivated car manufacturers to introduce new generations of cars in order to cope withfuel consumption and emissions issues. One of the most interesting structures that isintroduced recently to the production lines belongs to hybrid electric vehicles.A hybrid electric vehicle powertrain, generally, contains an electric energy bu®er and anelectric machine as well as the conventional internal combustion engine that can worktogether in several di®erent architectures known as series, parallel and series-parallel de-pending on how the electric machine is coupled with the internal combustion engine. Thisextra degree of freedom in the powertrain has raised several di®erent research routes onhow to optimize the power split between the electric machine and the conventional internalcombustion engine.

The present work presents a Dynamic Programming approach that solves the optimalpower split between the internal combustion engine and the electric machine in parallelhybrid electric vehicles in an e±cient way, taking minimal fuel consumption considerationsinto account. The power split must be carried out in such a way that in every moment thedemanding power on the ¯nal drive is ful¯lled by either the internal combustion enginealone, the electric machine alone or both together. Another important characteristic ofhybrid electric vehicles is the possibility to regenerate breaking energy that is dissipatedin conventional vehicles, by e±ciently using the electric machine in generator mode whilebraking, and storing this energy into the electric bu®er for further use. This is also takeninto account while designing the optimal controller in the presented work.

This optimal control problem is complicated in the sense that the future driving demandsare not known a priori to the controller and hence the decision making is impossible if wetreat the problem in a simple way. What can be done to cope with this issue is to dividethe problem into two di®erent cases. The ¯rst and the most straightforward case is thedeterministic case in which the whole driving cycle is known to the controller beforehand,as is considered through the whole thesis. This can be applied e±ciently to vehicles thatare driven through a speci¯c route many times and have stop-and-go driving cycles such ascommuter busses or refuse vehicles. The second case that is much more complicated andcan be applied to any vehicle is the stochastic case that contains no prede¯ned driving cycleand instead uses di®erent methods to predict or to estimate the future driving demandsdepending on the speci¯c technique that is applied. This can easily be realized having thenew GPS and GIS equipments in hand, though it is not covered in the presented work.

A Dynamic Programming-based (DP-based) algorithm has been developed as the con-trol system design and applied to the derived vehicle model based on fully deterministicdriving pro¯les. The employed controller shows highly satisfactory reductions in the fuelconsumption compared to a simple non-hybrid model in simulations. This algorithm hasthen been used as the heart of a newly developed toolbox to be working together with QSSToolbox under Matlab and Simulink environments for much easier further case studies.