Model Predictive Control for Energy Management in Electric Hybrid Vehicles
Lead Research Organisation:
University of Oxford
Department Name: Engineering Science
Abstract
In recent years, the automotive industry has seen a paradigm shift towards electric powertrains, largely driven by the considerable reduction of pollutant emissions at the point of use, something highly relevant to an increasingly environmentally conscious society. Plug in hybrid electric vehicles incorporate both a battery powered electric powertrain for typical use, and a conventional internal combustion engine for use if the electric power is completely depleted. Therefore, whilst the overall emissions of the car depend on several factors, including the source of the electric power and the cumulative distance driven, in typical use they can reasonably be expected to be less than that of a standard internal combustion car. The technological advantages these systems provide are amplified to the consumer by policies such as green subsidies, and therefore are likely to be key in the automotive market for years to come.
A further factor that influences the efficiency, and therefore cumulative emissions of a hybrid vehicle is the control system used to manage the power consumption of both the electric motor and internal combustion engine. Design of a control system to optimise the performance of the mechanical technologies is therefore a fundamental to the overall system. Several approaches have been used to tackle this issue, of which optimal control formulations have been particularly successful. These formulations in turn fall into three groups: Dynamic Programming DP, Pontryagin Minimum Principle PMP and Model Predictive Control MPC.
The proposed research project is therefore to further investigate the use of MPC to manage power in an electric hybrid vehicle, and more specifically to complete the following tasks - Model the hybrid vehicle processes including battery state, engine power maps and electric motor losses, and apply function approximation to the required level of accuracy whilst allowing desirable optimisation properties such as convexity. A particular area of interest will be the charge state of the battery, which has previously assumed to constantly decrease through the driving cycle.
Incorporate constraints on the battery state, engine power and electric motor power as well as uncertainty in the predicted driver demand into the energy management strategy. This calls for a robust optimisation with probabilistic and hard constraints.
Make the controller implementable on computing hardware typically available in production vehicles, with corresponding limits on functionality and precision. This will make it necessary to use a bespoke optimisation method in conjunction with a hierarchical control scheme that avoids large numbers of decision variables.
All of the above aims are completely novel and do not appear in the literature, and this project falls within the EPSRC energy research area.
A further factor that influences the efficiency, and therefore cumulative emissions of a hybrid vehicle is the control system used to manage the power consumption of both the electric motor and internal combustion engine. Design of a control system to optimise the performance of the mechanical technologies is therefore a fundamental to the overall system. Several approaches have been used to tackle this issue, of which optimal control formulations have been particularly successful. These formulations in turn fall into three groups: Dynamic Programming DP, Pontryagin Minimum Principle PMP and Model Predictive Control MPC.
The proposed research project is therefore to further investigate the use of MPC to manage power in an electric hybrid vehicle, and more specifically to complete the following tasks - Model the hybrid vehicle processes including battery state, engine power maps and electric motor losses, and apply function approximation to the required level of accuracy whilst allowing desirable optimisation properties such as convexity. A particular area of interest will be the charge state of the battery, which has previously assumed to constantly decrease through the driving cycle.
Incorporate constraints on the battery state, engine power and electric motor power as well as uncertainty in the predicted driver demand into the energy management strategy. This calls for a robust optimisation with probabilistic and hard constraints.
Make the controller implementable on computing hardware typically available in production vehicles, with corresponding limits on functionality and precision. This will make it necessary to use a bespoke optimisation method in conjunction with a hierarchical control scheme that avoids large numbers of decision variables.
All of the above aims are completely novel and do not appear in the literature, and this project falls within the EPSRC energy research area.
Organisations
People |
ORCID iD |
| Mark Cannon (Primary Supervisor) | |
| Sebastian East (Student) |
Studentship Projects
| Project Reference | Relationship | Related To | Start | End | Student Name |
|---|---|---|---|---|---|
| EP/N509711/1 | 01/10/2016 | 30/09/2021 | |||
| 1793329 | Studentship | EP/N509711/1 | 01/10/2016 | 31/07/2020 | Sebastian East |
| Description | A key issue with implementing model predictive control for energy management in plug-in hybrid electric vehicles is that a challenging optimization problem must be performed by the engine controller recursively throughout the journey. Previously, no optimization algorithm had been demonstrated that could solve this type of problem given the hardware and time constraints available, and the research in the literature had only implemented general purpose optimization software. My research has investigated algorithms that are specifically tailored to the structure of the problem, and I have developed two algorithms that have been published in high impact engineering journals and conferences (I have a further two under consideration). These algorithms have shown that the problem can be solved to a sufficient degree of accuracy in real time, and are a significant development towards the realisation of MPC as a real-time energy management control solution. The aformetnioned algorithms have been further extended to consider engine switching decisions, and to energy management problems in batter/supercapacitor electric vehicles. In the latter case, the energy management problem minimizes battery usage, thereby extending the life-cycle of the vahicles battery and redicing it's cost. |
| Exploitation Route | The next major step is to take the work that I have completed and implement it in an actual vehicle so that real-world performance can be demonstrated; currently, the algorithms have only been demonstrated through simulation. The algorithms do, however, also provide a benefit in a simulation environment, where because they are designed for the mathematical structure of the problem, they are faster than general purpose optimization software. |
| Sectors | Energy,Transport |