Chalmers researchers Nikolce Murgovski held a SEC webinar May 18. Focus was recent developments in electromobility and eco-driving.
Hi Nikolce! What are the major conclusions you made?
Research on fully electric vehicles (EV) is currently in a full swing and has superseded the research on, e.g., the hybrid electric vehicles (HEVs). EV powertrains are generally less complex than HEV powertrains. However, there are still many design possibilities. In EV powertrains many mechanical connections and components are being replaced by electrical counterparts. This gives a greater possibility to move things around and I believe we will see increased research in optimal powertrain design of EVs. Moreover, in EV powertrains, there are benefits from replacing one large motor with several smaller ones, something we generally do not see in combustion driven vehicles. Finding the optimal placement and dimensioning of the motors will depend on the particular application and requires diligent research activities.
Hydrogen powered vehicles are potential candidates for some vehicular applications. For e.g., fuel cell electric vehicles are being seriously considered for heavy-duty trucking. One of the reasons is that battery technology and availability of chargers have not yet reached a sufficient cost benefit level for an immediate transition to fully eclectic vehicles. To further improve energy efficiency, fuel cell vehicles may also be supported by an energy buffer, a battery, a supercapacitor, or something in between, to enable regeneration of braking energy and storing it for a future use.
There is also ongoing research on hydrogen combustion vehicles, where the internal combustion engine is kept, but its fossil fuel is replaced by hydrogen. As of now, too little data is available to evaluate the benefit of these vehicles compared to other alternatives.
Charging infrastructure must continue improving along with the increased share of electric vehicle fleets. We may see more research on improving the cost benefit of such fleets, whether this is in public transport or delivery services. Such research goes well in hand with the advancements in autonomous driving. These are large-scale optimization problems that involve research on routing, scheduling, design questions, etc.
Unavoidably, we will see significant research activities in algorithmic developments. Vehicles are getting access to ever increasing amounts of data and better algorithms are needed to process that data efficiently. For e.g., vehicles are being equipped with systems that collect information from various sources to estimate the traffic for multiple kilometers ahead. Vehicles and infrastructure may also communicate to enable better prediction of future events and foster the development of cooperative strategies. The access to big data can be utilized in many ways. We already mentioned the ability of making better predictions, but such data can also be utilized to learn and reduce model uncertainties. This is useful for enhancing energy efficiency and safety, while reducing emissions and congestions.
In the last two decades we have seen great improvements in the computational efficiency of optimization solvers. Some of these improvements are due to faster hardware, but many of them are algorithmic and software improvements. For e.g., about 1-2 decades ago we have seen a rapid progress in the optimal control community with the development of very efficient tools targeting various convex problems. Nowadays, there are many publicly available tools for solving nonlinear and non-convex optimization problems, including tools based on machine learning. Further and more rapid developments of tools and algorithms for attacking generic types of problems will certainly come, but also in algorithms that are dedicated to specific problems. The latter means that we must thoroughly study the physical principles underlying the problem at hand, rather than hoping that one magical tool will solve all our problems. To provide a simple analogy between the intelligence of a machine with that of people, then artificial intelligence may not be at all a single almighty entity, but rather a community of different individuals.
How was the discussion afterwards? What caught the audience’s attention?
While discussing the research area on charging infrastructure design, I provided an example of a recent investigation with researchers from Beijing Jiaotong University. This peaked the audience interest. Investigations are being performed to update the Haizhu tram line in Guangzhou, China. The goal is to reduce electricity grid costs by removing overhead wires and let trams charge only at the stations. Clearly the trams will have to be equipped with onboard energy storage. However, even stationary energy storage systems may need to be installed at certain stations or at a separate grid node to support the grid when multiple trams connect to charge simultaneously. We are speaking of electric power in the order of megawatts. I am not aware of such research performed on tramlines in Göteborg, but similar activities are ongoing on the charging infrastructure for electric city buses. With the increased share of electric vehicles in the public, commercial and private sectors, the electric grid will be subject to disruptive and as yet unseen electric loads. To enable a smooth and cost-effective operation, researchers will need to analyze multiple charging concepts in order to choose the most viable alternative, similarly as the study we performed with our Chinese colleagues.
You discussed a few research areas that are active at Chalmers right now. Which are these and how can collaboration within SEC contribute to utilization within these areas?
Right now, there is a massive undertaking within electromobility from multiple departments at Chalmers University. From the viewpoint of the division of Systems and Control, it is possible to cluster the research into three areas: plant design and control, traffic level and multi-vehicle planning and control, and optimal control within a single vehicle. These areas can further be divided, often in a nested, onion-like structure with multiple layers. For e.g., within the area of plant design and control, investigations are performed on charging infrastructure design and vehicular powertrains. These layers can further be divided into sub-layers, including topology generation and optimization, electric storage technology, dimensioning of components, etc. The remaining two areas on planning and control include even more nested layers.
Common for all research areas is that decisions need to be taken to optimize certain performance criteria, typically energy efficiency, reduction of emissions, operational and ownership cost, components’ wear, etc. Achieving such goals requires the development of efficient optimization algorithms and models for making reliable predictions. The modelling and optimization involve knowledge and collaboration from multiple parties, and requires access to big data, in general. This is where the collaboration within SEC can be very helpful. By design, SEC is a hub where different partners from industry, institutes and academia meet to share knowledge and initiate collaboration. Concerning big data, it is usually the industry that is the primary stakeholder. Such data has often a great value and is unfortunately kept confidential. This significantly slows down the research progress in academia. I hope that in a near future SEC will take the role of a mediator for data sharing.