Adaptive Control Using Map-Based ECMS for a PHEV

Martin Sivertsson

Linkping University Post Print

N.B.: When citing this work, cite the original article.

Original Publication: Martin Sivertsson, Adaptive Control Using Map-Based ECMS for a PHEV, 2012, ECOSM'12 -- IFAC Workshop on Engine and Powertrain Control, Simulation and Modeling. http://dx.doi.org/ Copyright: International Federation of Automatic Control Postprint available at: Linkping University Electronic Press http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-84961

Adaptive Control Using Map-Based ECMS for a PHEV

Martin Sivertsson

Vehicular Systems, Dept. of Electrical Engineering, Link oping University, SE-581 83 Link oping, Sweden, [email protected]

Abstract: A plug-in hybrid electric vehicle(PHEV) is a promising way of achieving the benets of the electric vehicle without being limited by the electric range. This paper develops an adaptive control strategy based on a map-based ECMS approach. The control is developed and implemented in a simulator provided by IFP Energies nouvelles for the PHEV benchmark. The implemented control strives to be as blended as possible, whilst still ensuring that all electric energy is used in the driving mission. The controller is adaptive to reduce the importance of correct initial values but since the initial values aect the consumption a method is developed to estimate the optimal initial value for the controller based on driving cycle information. This is seen to work well for most driving cycles with promising consumption results. The controller also fullls all requirements set by the PHEV Benchmark. Keywords: Optimal Control, Plug-In Hybrid Electric Vehicles, Equivalent Consumption Minimization Strategy 1. INTRODUCTION A hybrid electric vehicle(HEV) utilizes both electric energy and energy from fuel to meet the demands set by the driver. This may lead to a reduction in environmental impact and fuel consumption of the vehicle. A Plug-In HEV(PHEV) is a HEV with possibility to recharge the battery from the grid. This adds the potential of using the vehicle as an electric vehicle, without the range limitations in a pure electric vehicle. The supervisory control algorithm for these more complex powertrains play an important role in realizing the full potential of the powertrain. In order to evaluate dierent strategies the IFP Energies nouvelles(IFPEN) is organizing a benchmark for a PHEV, see PHEV Benchmark Rules (2012). This paper extends the adaptive map-based Equivalent Consumption Minimization Strategy(ECMS) developed in Sivertsson et al. (2011) to the PHEV problem and it is implemented for the simulator made available in the PHEV benchmark. The optimal torque distribution and generator speed is calculated oine and stored in tables. Then an online control is developed which fullls the requirements of the benchmark for reasonable initial values and nally a strategy for estimating the optimal initial value is developed that together with the adaptive control fullls the benchmark requirements. 2. IFPEN PHEV BENCHMARK In the IFPEN PHEV Benchmark a simulator is provided for which a supervisory control algorithm is to be designed. This simulator is a quasi-static model of the Chevrolet Volt with vehicle and battery dynamics and all energy converters modeled using stationary maps. The Chevrolet Volt has three energy converters, internal combustion engine(ENG), electric motor(EM), and generator(GEN), connected through a planetary gearset(GB). Both electric machines can work in both motoring and generating mode. The powertrain also incorporates three clutches that allows the vehicle to be driven in the following four modes: Mode 1: One motor pure electric vehicle. Only the EM is connected to the GB. Mode 2: Two motor pure electric vehicle. Both the EM and GEN are connected to the GB. Mode 3: Series HEV. Only the EM is connected to the GB. The ENG and GEN works as an auxilliary power unit, producing electric power. Mode 4: Power split HEV. All energy converters are connected to the GB. In the benchmark the controller should output desired torque from the ENG, EM, and mechanic brakes, the speed of the GEN, the position of the three clutches, and if the engine should be on or o. The inputs to the controller are the requested torque from the driver model, Treq , minimum allowed regenerative torque, battery state of charge(SOC), vehicle speed, average speed in the driving cycle, vavg , and approximate driving cycle length, Dtot . The aim of the benchmark is to minimize the criteria described in Table A.1 with a battery that is fully charged at the beginning of the driving cycle and may be depleted at the end of the driving cycle. There are also rules on how close the controller has to follow the desired velocity prole, see PHEV Benchmark Rules (2012).

2.1 Models The models implemented in the simulator are briey described below. For more details see PHEV Benchmark Rules (2012).

Vehicle Model The vehicle motion equation is implemented as (1) where Twh is the torque from the powertrain at the wheels and Tb is the torque applied by the brakes. rwh dv 2 (1) dt = Jveh Twh Tb rwh (mg sin + c0 + c1 v + c2 v ) Battery Model The battery model is of equivalent circuit type and implemented as: Ib = Uoc 2Rc
2 4R P Uoc c b 2 4Rc

ECMS see Paganelli et al. (2002); Sciarretta et al. (2004); Musardo and Rizzoni (2005). The problem is formulated as: min (H = Pf + Pech ) (14) [TEM , TEN G , GEN , M ode] = argmin(H ) (15) Subject to: Twh = Treq Tmin ( ) T Tmax ( ) (16) 0 max Pb,min (SOC ) Pb Pb,max (SOC ) Where the torque and speed limits are applied to each individual energy converter. 4. OFFLINE OPTIMIZATION Since the kinematic relations change with the actuation of the clutches the optimization problem to be solved diers between the modes. Due to the dicult nature of the problem the optimal solution is not calculated online. Instead in order to nd which mode to use when, the minimum cost for each mode is calculated oine and stored in tables for a given set of Treq , wh , and . The SOC is found to only have minor eects on the optimal solution, therefore that eect is ignored. To ensure that (16) are all fullled, or in the case of Treq = Twh , the produced torque is as close to the requested as possible for that mode, the cost function in (14) is augmented so that the closest point, that fullls all the inequalities, is selected. In order to nd which mode is optimal for each combination of Treq , wh , and , the optimal torque and speed setpoints also has to be found. But instead of just storing all the control variables in tables a few insights can be gained from the kinematic relations in (5)-(11) to reduce the amount of memory used: TEM can be calculated from Treq in all modes. Therefore Mode 1 requires no tables. Mode 2: Only GEN has to be stored and since Pf = 0 it is independent of Mode 3: TEM , EM , and therefore PEM are given by Treq and wh . The optimal output power from the generator should be on the optimal operating line of the engine-generator combination(GENSET). Therefore only the optimal output power for each PEM , combination has to be stored together with the optimal operating line of the GENSET. This results in 7-tables to be stored, shown in Fig. 1. That is: Mode-selection (3-D) Mode 2: GEN (2-D) Mode 3: PGEN SET (2-D), optline (1-D), and Toptline (1-D) Mode 4: GEN (3-D), and TGEN (3-D) The optimization is performed for a rather dense grid in Treq , wh , and . In order to minimize the amount of memory used the Treq , wh , or resulting in the smallest error in the interpolation scheme used in the online implementation if removed, is removed in an iterative manner. This is performed for all tables, so each table has its own discretization. To simplify the implementation and

(2) (3) (4)

Pech = Ib Uoc Ib SOC = dt Q0

Transmission The transmission is a planetary gear set with three clutches, c1 , c2 , and c3 . The kinematic relations between the energy converters and the wheels are: EN G = GEN c3 (5) EM = wh f d (1 + rs ) GEN c2 (1 c1 )rs (6) Ts = TEM GB
sgn(TEM )

(7) (8) (9) c2 (TGEN + c3 TEN G ) rs (10) (11)

Tr = TGEN = Tr c3 TEN G Tc = (1 + rs )

sgn(T ) Ts rs c2 GB s

(1 c2 )Ts + Twh = Tc GB
sgn(Tc )

f d

Consumption There are two consumptions provided in the simulator and used in this paper, fuel consumption and a fuel equivalent of the electricity consumption. They are dened as: m f mf = (12) f Dreal Pech mf,equiv = (13) avg qLHV f Dreal Where m f is the fuel ow, f the density of the fuel, Dreal is the distance travelled, Pech the electrochemical power, avg the average eciency from fuel to electricity, and qLHV is the lower heating value of the fuel. 3. PROBLEM FORMULATION Looking at the scoring metrics and CO2 data in Table A.1A.2 the problem can be reformulated as delivering the torque requested by the driver, or as close as possible if the requested torque is infeasible, in a fuel and computationally ecient way. Even though the CO2 emissions are higher for electricity production, the higher eciencies of the electric energy converters compared to the eciency of the combustion engine, results in that the minimization of the well-to-wheel CO2 emissions can be interpreted as fuel consumption minimization. So the aim is to minimize the energy use, with emphasis on the fuel consumption, while fullling the driver requests. This problem is well represented by the equivalent consumption minimization strategy(ECMS) where the sum of fuel and battery power is minimized. However, battery and fuel power arent directly comparable and therefore an equivalence factor relating the two is needed, for more information on

Mode selection(=2.15)

Mode 2: GEN

[Distance]

Distance T_mech_brake

4 Mode [] 3 2 GEN [rad/s]

600 400

[D_tot]

D_tot Lambda Lambda Eng_on T_ENG

[v_avg]

v_avg

[SOC]

SOC

Lambda T_EM

200
Lambda
Ignited? Mode Mode

w_GEN c1

1 5000 200 0 T
req

0 5000 200 0 T [rad/s]

req

[w_wh]

Req

100 5000 0P Mode 3: GENSET

wh

100 5000 0 Mode 3: optline

Mode controller
wh

c2 c3 Ignition?

[Nm]
4

[rad/s]

[Nm]

[rad/s]

x 10 0 [W]

[T_wh]

Req

400 200 5 4 3 2 1 x 10 0
4

[T_regen_lim] Modes

GENSET

GEN

4 [Nm] 100 50 0

Mode 3: Toptline

5 [] 10 2

0 PEM [W]

1 x 10
5

ENG

6 0

3 2 PGENSET [W]

1 x 10

0
4

Fig. 2. Structure of the controller. The controller consists of three main subsystems, one where the equivalence factor is calculated, one where the mode is selected, and one where the torque and speed setpoints are calculated. is guaranteed to make use of the stored electric energy and does not need information about the future driving mission. The second strategy is to mix usage of fuel and electricity throughout the driving cycle, a strategy known as blended strategy. It is well established in the litterature that a blended strategy may result in lower fuel consumption than CDCS, see for instance Larsson et al. (2010). However, in order for a blended strategy to make use of all the energy in the battery the length of the driving cycle has to be known. In the driving cycles provided by the PHEV Benhmark organizers only the approximate distance as well as the mean speed is known. In the provided driving cycles this approximate distance can deviate from the actual distance of the driving cycle by up to almost 10%. 6.1 Equivalence factor adaptation In order to make use of all the stored energy in the battery, a mix between the blended and CDCS strategies is implemented. The strategy is to undershoot the approximate distance by 10%, and use that as a horizon for the blended strategy. If the actual distance is longer than that used for the blended strategy, the control goes over into charge sustaining mode. This is achieved by setting a SOC reference, SOCc that is linear in ratio of travelled distance vs. expected distance, see Fig. 3, a method also used in Tulpule et al. (2009). The minimum SOCc is set to 0.315 in order to ensure that SOC (end) 0.3. The strategy used in Sivertsson et al. (2011) is then extended to t the PHEV problem. The strategy is to adapt the equivalence factor according to a tangent function in SOC . The idea is that as long as the SOC is near the desired SOC the control should remain rather constant. But when the SOC approaches the limits the control needs to adapt. In Sivertsson et al. (2011) this is used in a HEV where the aim is to maintain the SOC around a constant level. Here, since it is a PHEV, it is desirable to use the energy stored in the battery, therefore the center of the tan-function is SOCc . The used SOC -window is also decreased linearly with distance traveled. This is to allow larger deviations early in the driving mission, and then make the control follow the SOCc narrower towards the end of the driving cycle. The -adaptation is given by (17) where l1 and ls are

Mode 4: GEN(=2.2)

Mode 4: TGEN(=2.2)

600 [rad/s] 400 200 TGEN [Nm] 200 100 0 0 wh [rad/s]

200 100 0 100 4000 2000 Treq [Nm] 2000 Treq [Nm] 0 0 200 100 wh [rad/s]

Fig. 1. Structure of the stored data to reduce the memory consumption Mode 4 is only used when Treq > 0. 5. CONTROLLER The outline of the controller is shown in Fig. 2. The controller consists of three main subsystems. The rst subsystem calculates the value of the equivalence factor, , using the SOC and driving cycle data. The second subsystem controls which mode to engage and the third calculates the torque and speed setpoints for the energy converters. The modes block consists of ve subsystems, one for each mode and one for engine start. The mode controller outputs which mode to activate and if the engine should be started or not. In order to avoid too frequent engine starts/stops two thresholds are used, ton and tof f . The controller has to try to turn the engine on/o for a duration longer than ton /tof f before it is turned on/o. The torques and speed are then calculated using the tables calculated oine and the kinematic relations for that mode dened in (5)-(11). Care is also taken not to exceed any of the constraints in (16). 6. ENERGY MANAGEMENT The energy management of a PHEV can be divided into two categories. The rst is to make use of all the stored energy in the battery, that is run as an electric vehicle until the SOC is under a certain limit, and then operate as a hybrid in charge sustaining mode. This strategy is commonly denoted charge deplete-charge sustain strategy (CDCS). The main advantage of this strategy is that it

GEN

0 4000

SOCc=0.8075 5 4 SOC [] 0.7 0.6 0.5 0.4 0.7 0.8 0.9 SOC [] 0.3 0 0.2 0.4 0.6 distance [] 0.8 1 [] 3 2 1 0 0.9 0.8 SOCc SOCc+dSOC SOCcdSOC [] 4 3 2 1 0 0.3 5

SOCc=0.315

0.9 0.8 SOC [] 0.7 0.6 0.5 0.4 0.3 0 1000 2000 3000 4000

SOC( =2)
c

SOC(c=3) SOCc+dSOC SOCc SOC(

c, opt

=2.85)

SOCcdSOC

5000

6000

0.32 SOC []
6 (c=2) (c=3) (c, opt=2.85)

Fig. 3. The outline of the basic control shape

[]

5 4

Table 1. The change in consumptions compared to c,opt for dierent values of c . All controls ensure SOC (end) 0.3.
Cycle-info 10xFUDS Dtot = 119.9km Dreal = 119.9km 10xNEDC Dtot = 119.9km Dreal = 110.1km 10xUS06 Dtot = 119.9km Dreal = 128.9km c 3 c,opt =2.65 2 3 c,opt =2.63 2 3 c,opt =2.85 2 mf [%] -0.65 10.35 27.02 0.80 -0.03 5.30 mf,equiv [%] -0.19 2.32 15.87 0.56 -0.71 2.20 SOC (end) 0.3099 0.3076 0.3066 0.3420 0.3197 0.3162 0.3147 0.3103 0.3078

3 2 1 0 0 1000 2000 3000 time [s] 4000 5000 6000

Fig. 4. The SOC and -trajectories for dierent c values on the US06 driving cycle. A poor c may lead to a switching characteristic of the control. Table 2. The change in consumption with adaptive c for dierent c,init , compared to c,opt .
Cycle-info 10xFUDS Dtot = 119.9km Dreal = 119.9km 10xNEDC Dtot = 119.9km Dreal = 110.1km 10xUS06 Dtot = 119.9km Dreal = 128.9km c,init 3 c,init,opt =2.65 2 3 c,init,opt =2.55 2 3 c,init,opt =2.85 2 mf [%] 0.10 -0.57 -0.87 15.61 -0.16 1.69 0.07 0.03 0.39 mf,equiv [%] -0.29 -0.05 -0.16 0.45 0.08 0.45 -0.41 -0.18 -0.05 SOC (end) 0.3110 0.3075 0.3081 0.3306 0.3188 0.3184 0.3143 0.3109 0.3101

constants that control the slope and range of the tangent function, and dSOC is the allowed deviation from SOCc . ls (SOC SOCc )) (17) 2dSOC A benet with this formulation is that the smaller dSOC is, the steeper the slope around SOCc becomes and the faster the control reacts to deviations. In Fig. 3 the shape of the control is shown for the case when the approximate distance is correct. That is the SOCc undershoots the distance traveled, and thus results in the control going over to charge sustaining mode. Also shown is that the allowed SOC deviation gets smaller with distance. The variable c still has to be decided. In Table 1 the change in consumptions compared to the consumptions with optimal c and end SOC are shown for dierent values of c and dierent driving cycles. A c is considered optimal if the trajectory follows SOCc . But since SOCc is based on undershooting the approximate driving cycle length, there are c values that result in lower consumptions, this is however hard to predict. It is seen that the optimal value changes with the driving cycle. The control ensures SOC (end) 0.3 for all c but it might come with a substantial increase in consumption if the c value is wrong. = c l1 tan( In Fig. 4 the and SOC trajectories for the dierent values of c on the US06 driving cycle is shown. Due to the driving mission length provided only being approximate, the control undershoots the length in order to make sure all electric energy is used. The US06 cycle is however 7% longer than the length provided, resulting in an undershoot of roughly 16% for the controller. It is seen that the control for c = c,opt does not follow the SOCc , instead it follows SOCc dSOC . For c = 2 this results in a control that switches rapidly between 2.5 and 5, something that comes with a large consumption penalty.

6.2 Adaptive control of c In order to avoid the switching nature of the -control seen in Fig. 4 the idea is to adapt c if the SOC deviates too much from SOCc . This is done with a PI-controller according to: c = c,init + Kp (SOCc SOC ) + Ki (SOCc SOC )dt (18) The values of Kp and Ki control how fast the controller adapts, but a faster controller comes with a slight consumption penalty. In Table 2 the consumption change compared to c,opt without adaptive c is shown for dierent driving cycles and c,init . A c,init is considered optimal if it roughly produces a SOC trajectory that follows the desired trajectory without c deviating too far from c,init . It is seen that the adaptive c performs as well as c,opt , better in some cases, worse in some cases. But most of all it reduces the eect of poor initial values. This is also conrmed in Fig. 5 where the SOC and -trajectories are shown for the US06 driving cycle. The switching nature is almost completely removed, resulting in a near constant value during the entire blended phase. 6.3 Estimating c,init Even if the developed control has been seen to perform well for all reasonable initial c , the consumption is still aected by it. Therefore it is desirable to achieve an estimate as close as possible to the optimal c . In Fig. 6,

0.9 0.8 SOC [] 0.7 0.6 0.5 0.4

SOC(

c, init

=2)
3.2

SOC(c, init=3) SOCc+dSOC SOCc SOC(

c, init, opt

=2.85)

NEDC US06 FUDS FHDS Artemis urban Artemis extra urban Artemis Highway

SOCcdSOC
2.8

0.3

1000

2000
c

3000 ( (
c, init

4000

5000

6000
c, opt 2.6

=2)

c, init

=2) =3)

6 5 4 [] 3 2 1 0 0 1000 2000

(
c

c, init

c, init

=3)
2.4

c(c, opt, init=2.85) (c, opt, init=2.85)

2.2

3000 time [s]

4000

5000

6000

2 50

100

150 Distance [km]

200

250

300

Fig. 5. The SOC and -trajectories with adaptive c for dierent values on c,init on the US06 driving cycle. The adaptive c reduces the eect of poor initial values. the optimal c,init is plotted against approximate distance for the driving cycles used. It is seen that the shape of the proles is similar for all driving cycles. The all electric range, that is the distance for c,init > 2, diers up to almost 100% for the dierent driving cycles. In Fig. 7 the approximate distance required to exceed the all electric range is plotted against mean speed. Even if the mean speed isnt enough to describe the driving cycle, since neither slope nor how transient it is is captured by the mean speed, the all electric range is approximated by a linear function, shown in Fig. 7. Artemis Urban is plotted in magenta to mark that it is considered an outlier and is not included when the line is tted. Since the losses in the vehicle motion equation (1) are quadratic in speed, a straight forward assumption would be that the allelectric range decreases with mean speed, an assumption that is also used here. The approximate distance is then corrected with the proposed linear correction, in order to compensate for the dierent all-electric ranges. The result is shown in Fig. 8. It is seen that the correction shifts the points to the same region, a trend that is well captured by an exponential function. The nal scheme to estimate c,init is of the form: Dcorr = Dtot (k1 vavg + k2 ) c,init = k3 (1 exp(k4 Dcorr + k5 )) (19) (20)

Fig. 6. The optimal c,init as a function of approximate distance for dierent driving cycles. The general shape of the curves are similar, however the all-electric range diers between the cycles.
140 120 Dstart [km] 100 80 60 40 Cycles used Artemis Urban Model

10

15 vavg [m/s]

20

25

30

Fig. 7. The approximate distance required to exceed the all-electric range for dierent driving cycles and a linear model to capture the behaviour. Artemis Urban is considered an outlier and is not included when the line is tted.
3.2 3 2.8 c 2.6 2.4 2.2 2 50 0 50 100 Dscaled 150 Cycles used Model Artemis Urban 200 250

In Table 3 the results for the full controller with c,mod is compared to the results for c,opt . Even if the estimated c,init is not too far from the optimal, the consumption can dier substantially. Interesting to note is that the driving cycle with the largest c,init error shows the best result. Looking at Fig. 9 this appears to be due to that the c,mod -control has a higher value when entering charge sustaining-mode which results in a less switching behaviour and lower consumption. Another important property of the c,init estimation is that it should be such that it avoids unnecessary engine starts if the driving mission is within the all electric range. This is achieved for all tested driving cycles except

Fig. 8. c,init,opt vs. corrected approximate distance and how it is modeled. Artemis Urban is considered an outlier and is not included when the curve is tted. Table 3. The change in consumption with modeled c,init , compared to c,opt .
Cycle-info 10xFUDS(c,init,opt =2.65) 10xNEDC(c,init,opt =2.55) 10xUS06(c,init,opt =2.85) 20xArtemis Urban(c,init,opt =2.84) c,init,mod 2.518 2.533 2.741 2 mf [%] 13.62 0.80 5.30 -6.69 mf,equiv [%] 2.32 0.56 2.20 -1.41 SOC (end) 0.3066 0.3162 0.3078 0.3104

FHDS (1 unnecessary start) and Artemis Extra-Urban (2 unnecessary starts), which is deemed acceptable.

0.9 0.8 SOC [] 0.7 0.6 0.5 0.4 0 2000 4000 6000 8000 10000 12000

SOC(c, init, opt=2.84) SOCc+dSOC SOCc SOC(c, initmod) SOCcdSOC

Table A.2. Data for CO2 emissions

Gasoline well-to-tank emissions Gasoline combustion Electricity production(Europe average) 12.5gCO2 /MJ of fuel 73.4gCO2 /MJ of fuel 94.7gCO2 /MJ of electric energy

Appendix A. BENCHMARK DATA

14000 16000 18000

REFERENCES Larsson, V., Johannesson, L., and Egardt, B. (2010). Impact of trip length uncertainty on optimal discharging strategies for phevs. In IFAC Symposium on Advances in Automotive Control. Munich, Germany. Musardo, C. and Rizzoni, G. (2005). A-ECMS: An adaptive algorithm for hybrid electric vehicle energy management. In IEEE Conference on Decision and Control and the European Control Conference, 44, 1816 1823. Paganelli, G., Delprat, S., Guerra, T., Rimaux, J., and Santin, J. (2002). Equivalent consumption minimization strategy for parallel hybrid powertrains. In IEEE Conference on Vehicular Technology, 55, 20762081. PHEV Benchmark Rules (2012). http://www.ecosm12. org/sites/ecosm12.org/files/PHEV\%20benchmark\ %20rules.pdf. Downloaded 2012-07-18. Sciarretta, A., Back, M., and Guzzella, L. (2004). Optimal control of parallel hybrid electric vehicles. IEEE Transactions on Control Systems Technology, 12(3), 352363. Sivertsson, M., Sundstr om, C., and Eriksson, L. (2011). Adaptive control of a hybrid powertrain with map-based ECMS. In IFAC World Congress. Milano, Italy. Tulpule, P., Marano, V., and Rizzoni, G. (2009). Eects of dierent PHEV control strategies on vehicle performance. In American Control Conference. St. Louis,USA.

4 c(c, init, opt=2.84) 3.5 3 2.5 2 1.5 ( (

c, init, opt

=2.84)

c(c, initmod)
c, initmod

[]

2000

4000

6000

8000

10000 time [s]

12000

14000

16000

18000

Fig. 9. SOC and trajectories for the full controller, with c,init,mod , compared to the c,init,opt . The modeled c,init is quite far from the optimal but still SOC (end) 0.3.

Table A.1. The scoring metrics used in the benchmark

Metric Acceleration 0-100km/h [s] Acceleration 70-120km/h [s] Acceleration 0-1000m on 4% slope [s] Braking Distance from 100km/h [m] Total energy use (fuel+electricity) [MJ/km] Fuel consumption [MJ/km] Well-to-wheel CO2 emissions [kg/km] Processor use [simulation time] Memory use [MB] Weigth 7.5% 7.5% 7.5% 7.5% 15% 20% 15% 10% 10%

Performance(30%)

Energy and Economy(50%) Computational performance (20%)

7. CONCLUSIONS An adaptive map-based implementation of ECMS is developed and implemented for the IFPEN PHEV benchmark problem. The control strives to be as blended as possible, but still ensuring that all electric energy is used. The control tries to follow a SOC reference that is linear in traveled distance, but to ensure that all electric energy is used this distance is underestimated. The equivalence factor is adapted according to a function in SOC , a function whose center adapts according to how well the SOC reference is followed. Finally a method for estimating the initial value for the controller from driving cycle data is developed. The resulting controller fullls all the requirements of the benchmark. SOC (end) 0.3 for all tested cycles, and the velocity prole error is within the tolerances prescribed by the benchmark. The control is also seen to adapt despite an initial estimate dierent from the optimal value. It also avoids unnecessary starts for most tested driving cycles that are within the all electric range. The controller provides a fuel economy that is within [-6, 13]% of what is considered optimal and is implemented in such a way that it only needs approximate driving mission length and average velocity.