Implementing the functional requirements for determining the optimal arrangement of a distributed charging infrastructure

4th IFAC Workshop on Dependable Control of Discrete Systems The International Federation of Automatic Control September 4-6, 2013. University of York, York, UK

Implementing the functional requirements for determining the optimal arrangement of a distributed charging infrastructure Tam´ as Kurczveil ∗ Eckehard Schnieder ∗ ∗

Technische Universit¨ at Braunschweig Institute for traffic safety and automation engineering Langer Kamp 8 38106 Braunschweig, Germany e-mail: [email protected], [email protected] Tel: +49 531 391-3317 Fax: +49 531 391-5197 Abstract: The optimized and reliable operation of future traffic by intelligent control systems will need to take into account boundary conditions that arise from alternative drive concepts. New challenges will need to be mastered when it comes to corresponding energy systems, control of operations, and communication interfaces, such as needed for the sufficient energy supply of traffic participants. However, they will need to be conformed to existing systems, technologies, and infrastructure to allow the common operation and positioning of charging elements with minimum interference between different modes of transport. Funded by the German Federal Ministry of Transport, Building and Urban Development (Bundesministerium f¨ ur Verkehr, Bau und Stadtentwicklung) the project emil (Elektromobilit¨at mittels induktiver Ladung − electric mobility via inductive charging) will integrate an inductive vehicle charging system and a compatible prototype bus fleet into Braunschweig’s traffic infrastructure. This paper describes the methodic approach and the implementation of functional requirements in a traffic simulation tool that are required for an evaluation of future urban road traffic with an increased rate of electric vehicles. The modifications can subsequently be used to determine the optimal placement of the corresponding charging infrastructure with consideration of conventional traffic demand. Keywords: traffic control, energy management systems, electric vehicles, vehicle dynamics, physical models, numerical simulation, iterative methods 1. INTRODUCTION

at mittels the funding of the project emil (Elektromobilit¨ induktiver Ladung − electric mobility via inductive charging). It includes the evaluative implementation of an inductive energy transfer system for public transportation and a compatible prototype bus fleet in the city of Braunschweig. Fig. 1 shows the bus lines M19 and M29 that circle Braunschweig’s city center clockwise and counterclockwise, respectively. They carry the highest percentage of passengers with a high frequency from and to major traffic nodes and landmarks such as the central train station and the university and shall there fore be equipped with an inductive charging system and compatible electric buses. The goal of the project emil is to analyze and optimize the operation and economic feasibility of an inductive electric charging system and to develop suitable operating strategies. One focus of research lies in the analysis of integrative aspects that allow for the common utilization of the charging and road transport infrastructure by public and private transport with minimum obstruction of public and total traffic. The methodic approach for these analyses and its implementation shall be presented in this paper.

The prospective post-oil era and rising fuel prices have lately resulted in several global trends towards alternative drive technologies. The main advantage of gasoline fuel over other energy carriers is its high specific energy of up to 44 MJ kg . Current projections for the development of equally convenient alternative energy storages go far beyond 2030. Thus, the utilization of alternative energy sources, such as the electrochemical energy stored in Lithium-Ion batteries (0.544 MJ kg ), will result in high vehicle masses and low ranges for the coming decades, as stated in Winter [2012] and Wansart [2012]. Measures that aim to counter these deficits include the application of light-weight materials, energy/time-optimal routing, intelligent control of traffic light-signal systems, and government regulations that introduce (operational, financial and/or infrastructural) incentives for buyers of vehicles with alternative drive concepts. The German Federal Ministry of Transport, Building and Urban Development (BMVBS) has therefore granted 978-3-902823-49-6/2013 © IFAC

37

10.3182/20130904-3-UK-4041.00003

2013 IFAC DCDS September 4-6, 2013. York, UK

(1) the time-discrete and quantized replication and emulation of the continuous movement (behavior) of individual vehicles (2) along static/dynamic routes (3) in predefined networks and (4) their location-discrete control, such as at intersections, light-signal systems, or bus stops. Reasonable simplifications of the kinematic and interaction models are inevitable to allow for acceptable computing times when simulating large traffic networks (Detering [2011]). The goal of possible optimization measures lies in finding an optimum of the considered system in regard of the energy consumed. This evaluation requires a suitable simulation tool that allows the implementation of custom traffic scenarios, including traffic demand, a charging infrastructure, customized vehicles, prioritization, different light-signal schedules, and the generation of a customized simulation output regarding the required energy of individuals and the entire system as the desired optimization criterion. Fig. 1. Braunschweig’s urban road network with the route c of bus lines M19 and M29 (red) OpenStreetMapContributors

Since no simulation tool exists that allows all the mentioned requirements to be modelled, the best choice for this task (due to its license under GPL and high compatibility with data formats of many commercially available tools) is the traffic simulation tool SUMO (Simulation of Urban MObility) by Behrisch et al. [2011]. Its development was initiated by the Institute of Transportation Systems of the German Aerospace Center (DLR), in 2001. It has evolved into a simulation tool, high in features, functionality and interfaces. Even though instantiated vehicles follow a simplified behavior, traffic simulation tools like SUMO allow the realistic replication of prevailing traffic in arbitrary road networks.

2. METHODIC APPROACH Due to its hybrid nature, the optimization of traffic by (intelligent) traffic-control systems is a highly complex matter. Next to road traffic itself, optimization measures and methods need to consider factors such as the schedule and position of light-signal systems, reliability and failure of critical components, and driving behavior. Complexity additionally increases when the evaluation of new technologies, such an alternative charging infrastructure, requires further decompositions and models to be regarded, e.g. the energy supply of corresponding vehicles.

The approach is to model current representative traffic scenarios that take into account the existing infrastructure, a time-varying traffic demand model (differentiating between traffic intensities, such as work-days and weekends), and light-signal schedules. Many of these scenarios and its describing parameters are outlined in FGSV [2001] that mainly focuses on conventional traffic. Meanwhile, the implementation of new functionalities in SUMO on the system-function level will allow for the instantiation of inductive charging stations and compatible vehicles. In a final step, traffic parameters of the inductive charging system can be incorporated, instantiating the previously implemented functions with charging stations and compatible vehicles. An external optimization framework will initiate and iterate through simulation instances by analyzing the simulation results and adapting parameters until the requirements for a minimum energy consumption have been satisfied. Fig. 2 depicts the above described method and the required additions regarding new functionalities, infrastructure components, and traffic demands.

Increasing urbanization and motorization (at much higher rates than infrastructure development) has been accompanied by problems such as major traffic disruptions in consequence of congestions, strengthening the need to develop, apply and/or optimize new traffic-control strategies to counter these problems. Plenty of the models for a multitude of applications and considerations have emerged between 1970 and 1990 that are described in Helbing [1997]. Only the availability of capable computing and communication technologies allowed for the integration of these models into on-line traffic management systems, laying the path for the new research and development field of Intelligent Transportation Systems (ITS) in the past decade, bringing forth works such as Hounsell and McDonald [2001] and Gradinescu et at. [2007]. While current applications of (mostly commercial) traffic simulation tools lie in planning and evaluating new infrastructure, their engines will be a key component in future traffic management and coordination systems when it comes to creating a reliable forecast from fusioned data of distributed sensor measurements, as described in Boxill and Yu [2000]. They consist of 38

2013 IFAC DCDS September 4-6, 2013. York, UK

acceleration g, time variant vehicle altitude h [k], and moment of inertia of internal rotating elements Jint . Eveh [k] = Ekin [k] + Epot [k] + Erot,int [k] m 2 J = · v [k] + m · g · h [k] + int · v 2 [k] 2 2

(1)

In consideration of energy losses ∆Eloss [k] caused by air, rolling, and curve resistance and constant consumers (e.g. air conditioning), the energy gain between time steps k and k + 1 can be calculated by equation 2.

∆Egain [k] = Eveh [k + 1] − Eveh [k] − ∆Eloss [k]

(2)

The energy loss is made up of the components in equation 3, with the variables air density ρair , vehicle front surface area Aveh , air drag coefficient cw , covered distance ∆s [k], rolling resistance coefficient croll , centripetal force Frad , curve resistance coefficient crad , and the (average) power of constant consumers Pconst .

Fig. 2. Interfaces between the basic functionalities to model vehicle movement (conventional - gray), required additional functionalities to model (electric vehicles) and compatible charging infrastructure elements (emil - green), and an external optimization framework (blue)

∆Eloss [k] =

3. IMPLEMENTATIONS IN TRAFFIC SIMULATION TOOL SUMO

(3)

= ∆Eair [k] + ∆Eroll [k] + ∆Ecurve [k] + ∆Econst [k] 1 ∆Eair [k] = · ρair · Aveh · cw · v 2 [k] · |∆s [k]| 2 ∆Eroll [k] = croll · m · g · |∆s [k]| m · v 2 [k] ∆Ecurve [k] = crad · · |∆s [k]| r [k] ∆Econst [k] = Pconst · ∆t

In order to create a simulation output about the consumed energy, an energy model will have to be implemented for instantiated vehicle objects. Without affecting the vehicle’s behavior, this model will evaluate its energetic state and calculate the variation of its energy content. 3.1 Vehicle Kinematic Model

Depending on its sign, ∆Egain [k] is the amount of energy, the vehicle has consumed or regained resulting from its movement. The variation of the vehicle’s energy content (e.g. in its battery) can be calculated by equations 4 and 5 by introducing efficiency factors for recuperation ηrecup (∆Egain [k] > 0) and propulsion ηprop (∆Egain [k] < 0).

The change of the energy content can be calculated by summing the kinetic, potential, and rotational energy gain components from one discrete time step to the following, and subtracting the losses caused by different resistance components that are described in Mitschke and Wallentowitz [2004]. Fig. 3 schematically illustrates the vehicle and relevant parameters for the calculation of its energy at discrete time steps.

EBat [k + 1] = EBat [k] + ∆Egain [k] · ηrecup −1

EBat [k + 1] = EBat [k] + ∆Egain [k] · ηprop

(4) (5)

It can be shown, that this simplistic vehicle model is capable of calculating the trend of a vehicle’s energy content with adequate accuracy. To visualize the functionality of the new implementations, the speed profile of the New European Driving Cycle (NEDC) with no curves and a non-regular incline of 1% is applied for a vehicle with the parameters listed in table 1. The NEDC is a defined driving cycle (speed profile) with the goal of providing a common emission and fuel economy test basis for all passenger cars and is specified in UN-ECE Regulation No 101 [2012]. Whereas the basis for the evaluation of the fuel economy for light trucks and commercial vehicles is specified in DIN 70030-2 [1986], the NEDC is used as a reference in the context of this work due to its widely established application. Fig. 4 illustrates the resulting individual cumulative energy components, where Eloss represents energy losses in the vehicle’s drive train.

Fig. 3. Vehicle movement and relevant variables at discrete time steps The calculations of the kinetic, potential, and rotational energy follow the straight-forward application of the wellknown equations from classical mechanics. The vehicle’s energy Eveh [k] at the discrete time step k can thus be calculated by equation 1, with the known variables vehicle mass m, time variant vehicle speed v [k], gravity 39

2013 IFAC DCDS September 4-6, 2013. York, UK

Table 1. vehicle parameters applied for a New European Driving Cycle (NEDC) Parameter vehicle mass front surface area air drag coefficient rolling resistance coefficient moment of inertia of int. rot. elements power of constant consumers drive efficiency recuperation efficiency

mveh Aveh cw croll Jint Pconst ηprop ηrecup

Value 14 200 kg 5.00 m2 0.62 0.012 kg 37.0 m 2 259 W 0.72 0.81

Fig. 5. Petri net of charging station functional states with transitions component failures, with their only power source being the battery. The charging system needs to be designed with appropriate redundancy, ensuring a high availability at an acceptable installation and operation cost. For the instantiation of compatible vehicle objects, a corresponding vehicle type needs to be defined. The MSVehicleType class was therefore modified to support the required vehicle parameters. Listing 1 shows an example of a vehicle type definition. Calculations of the vehicle’s energy content have been allocated to the MSVehicle class in its method MSVehicle::moveChecked(). Listing 2 shows an excerpt of the implementations.

Fig. 4. Cumulative energy components (lower) of a vehicle when driving the speed profile (upper) of the New European Cycle (NEDC) with an non–regular incline of 1% 3.2 Vehicle Charging Model

Listing 1. Definition of vehicle types with additional parameters

An additional requirement includes the implementation of a new infrastructure object, that charges compatible vehicles’ batteries. The location of charging stations, the charging power and efficiency of the charging process as well as individual exponentially distributed failure and recovery rates shall be specifiable by the user. If a vehicle moves or stops above or in the system-specific proximity of an inductive charging station that is in up state, the energy content of its battery is charged according to equation 6, with charging power Pchrg , charging efficiency ηchrg , and duration between two discrete time steps ∆t. Fig. 5 depicts the functional states that charging stations can attain and the implemented functionality in form of transitions in a Petri net. With the constant failure rate λ and recovery rate µ, the probability for the transitions failure and recovery from the corresponding previous state in the Petri net can be modelled with equations 7 and 8, according to the exponential distribution.



EBat [k + 1] = EBat [k] + Pchrg · ηchrg · ∆t

(7)

−µt

(8)

in

MSVehi-

// UPDATE VEHICLE/BATTERY ENERGY if (getVehicleType().getMaxBatKap() != 0) { // Energy lost/gained from vehicle movement (via // vehicle energy model) [kWh] myState.myKap -= getPropEnergy(myState.mySpeed, myLane); // Energy gained at inductive charging station // [kWh] myState.myKap += getChrgEnergy(myState.myPos, myLane); // saturate between 0 and MaxBatKap [kWh] if (myState.myKap < 0){ myState.myKap = 0; if (getVehicleType().getMaxBatKap() > 0){ // generate Output, that this vehicle’s battery // is depleted } } else if (myState.myKap > getVehicleType().getMaxBatKap()) { myState.myKap = getVehicleType().getMaxBatKap(); } }

(6)

Probability of failure: F (t) = 1 − e−λt Probability of recovery: M (t) = 1 − e

Listing 2. Energy calculations cle::moveChecked()

With the modelling of charging station failures, reliability evaluations can be performed, such as needed to asses the charging system’s quality of service and optimize the placement of charging stations. In case of project emil these analyses play a crucial role, since the designated buses need to be able to complete their routes even in case of traffic disruptions (such as delays or detours) and

The instantiation of charging stations can be done in a SUMO configuration’s additional files with the syntax 40

2013 IFAC DCDS September 4-6, 2013. York, UK

shown in listing 3. The MSLane class has been extended by a list of structures struct MSLane::ChrgStn, containing information about instantiated charging stations. The method NLHandler::myStartElement() that initially parses through the configuration files has also been expanded to fill the newly defined list of structures struct MSLane::ChrgStn with the respective parameters of charging stations. These modifications in SUMO allow vehicles to be charged according to equation 6.

depart="10" departKap="91" charged="no">

Listing 3. Instantiation of charging stations ... ...

3.3 Simulation Output While the calculations of the required energy consist of the summation of individual components, an external framework would not require more than the total energy of individual vehicles for a (global) optimization. The goal of the project emil is to develop a prototype bus (fleet) and a compatible charging infrastructure and to analyze possible strategies for the integration of individual motor car traffic into the implemented charging infrastructure. It is therefore appropriate and preferred to place charging stations at bus stops, reducing the possibilities to 26 locations (the amount of bus stops) along the route. Relevant variables for a subsequent optimization include the arrival time, state of charge at arrival time, departure time, and state of charge at departure time.

Fig. 6. Possible visualisation of the simulation output: Energy content of a vehicle’s battery (upper part) along its designated route (lower part) The generated data can be passed to subsequent algorithms that process and evaluate the contained information and modify the parameters of relevant/representative scenarios in order to meet optimization criteria. 4. CONCLUSION AND OUTLOOK The resulting scope of SUMO functionalities along with a distributed charging network, traffic demand model for vehicles with compatible drive concepts, and the described simulation output will allow for the analysis and optimization of the traffic in regard of the arrangement of the charging stations along defined bus routes.

An output extension has been implemented in SUMO that lists these variables for respective vehicles, if a bus stop or charging station is reached. If a vehicle’s route includes neither a bus stop nor a charging station, it does not need further analyses in this context and is therefore not listed in the simulation output. In order to allow for a rapid implementation of evaluation functionalities, xml was chosen as the output format. Listing 4 shows the xmloutput of a simulation from an exemplary scenario. Fig. 6 illustrates one possibility for a corresponding visualization in the commonly used style of a bead diagram: the vehicle route is represented by a blue line, designated stops along the route with blue dots, and charging stations with green induction loops, respectively. Charging station failures would be indicated by a vehicle’s Charge attribute charged="no".

An optimization algorithm and framework will need to be developed, which takes into account • desired long battery lives (by maintaining high battery charging states), • high availability (by redundant configuration) • minimum required energy of the collective simulated traffic, and • minimum travel time by synchronizing charging instances according to predicted waiting times (such as for the entry and exit of passengers or delays caused by remaining traffic) and known waiting times (such as for light-signal systems or public transportation schedules).

Listing 4. Xml-output of the BusChrgStop device
Next to the variation of the arrangement of charging stations, scenarios can also be adapted by variation of parameters describing • infrastructure (e.g. schedules of light-signal systems), • traffic rules (e.g. driving behavior, restrictions, joint use of charging stations by private and public traffic participants), and/or 41

2013 IFAC DCDS September 4-6, 2013. York, UK

• communications (e.g. prior notification of a required charge).

J. Wansart. Analyse von Strategien der Automobilindustrie zur Reduktion von CO2 -Flottenemissionen und zur Markteinf¨ uhrung alternativer Antriebe: Ein systemdynamischer Ansatz am Beispiel der kalifornischen Gesetzgebung. Dissertation, Technische Universit¨ at Braunschweig, Springer Gabler, Wiesbaden, 2012. D. Helbing. Verkehrsdynamik - Neue physikalische Modellierungskonzepte. Springer, Berlin, 1997. N.B. Hounsell, M. McDonald. Urban network traffic control. Proceedings of the Institution of Mechanical Engineers, Part I: Journal of Systems and Control Engineering, 215, (4), pp. 325-334, SAGE, Bath, 2001. V. Gradinescu, C. Gorgorin, R. Diaconescu, V. Cristea, I. Iftode. Adaptive Traffic Lights Using Car-to-Car Communication. Proceedings of the IEEE Vehicular Technology Conference, Dublin, 2007. S.A. Boxill, L. Yu. An Evaluation of Traffic Simulation Models for Supporting ITS Development. Texas Southern University, Center for Transportation Training and Research, Houston, 2000. S. Detering. Kalibrierung und Validierung von Verkehrssimulationsmodellen zur Untersuchung von Verkehrsassistenzsystemen. Dissertation, Technische Universit¨ at Braunschweig, Braunschweig, 2011. M. Behrisch, L. Bieker, J. Erdmann, D. Krajzewicz. SUMO - Simulation of Urban MObility: An Overview. SIMUL 2011, The Third International Conference on Advances in System Simulation, Barcelona, 2011. Handbuch f¨ ur die Bemessung von Straßenverkehrsanlagen (HBS). Forschungsgesellschaft f¨ ur Straßen- und Verkehrswesen, K¨oln, 2001. M. Mitschke, H. Wallentowitz. Dynamik der Kraftfahrzeuge. Springer, Berlin 2004. Regulation No 101 of the United Nations Economic Commission for Europe (UN-ECE). Official Journal of the European Communities - L 138, Brussels, 2012. DIN 70030-2:1986-11: Road vehicles; determination of fuel consumption; goods vehicles and buses. Deutsches Institut f¨ ur Normung, Beuth, Berlin, 2012.

In project emil, charging stations shall only be placed at bus stops, of which nBusStops = 26 are located along the route of designated bus line. With a maximum amount of nChargingStations = 4 charging stations, the amount of possible configurations alone is nChargingStations

  nBusStops = (9) k k=0           26 26 26 26 26 = + + + + = 17902. 4 3 2 1 0 X

However, with increasing amount of varying parameters, it is relatively clear that the amount of required simulation cycles resulting from simply creating all possible combinations can easily outgrow the available computing powers to determine the optimal placement of charging stations. Due to the time- and space-discrete properties and nonlinearity of traffic models, future work will include the implementation of an optimization framework based on genetic algorithms to cover the manifold of variables and to find the appropriate parameter set for optimal system operation. Simulation results can also include the evaluation of occupancy rates at different positions within the road network. This data can be used for the design and alignment of inductive charging stations that maximize their duty cycle and thus the operation efficiency. By implementing different scenarios for the amount of compatible vehicles in the future, it will also be possible to determine saturation points for the amount of participating vehicles, where operational interferences and obstructions among public transportation and between public and private vehicles can be expected. ACKNOWLEDGEMENTS The project emil is funded by the German Federal Ministry of Transport, Building and Urban Development (Bundesministerium f¨ ur Verkehr, Bau und Stadtentwicklung - BMVBS). It is jointly carried out by Braunschweiger Verkehrs-AG, Bombardier Transportation GmbH, BS|Energy, and Technische Universit¨ at Braunschweig and coordinated by the NOW GmbH (National Organisation Hydrogen and Fuel Cell Technology). We hereby thank all our project partners for their continuous and kind cooperation.

REFERENCES M. Winter. Elektromobilit¨ at mit Lithium-IonenTechnologie: Chancen, Herausforderungen, Alternativen. Proceedings of the HEV 2012, Hybrid and Electric Vehicles, Braunschweig, 2012. 42