Você está na página 1de 11

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO.

4, MAY 2011

1419

Optimal Design and Real-Time Control for Energy Management in Electric Vehicles
Lei Wang, Member, IEEE, Emmanuel G. Collins, Jr., Senior Member, IEEE, and Hui Li, Senior Member, IEEE
AbstractTo extend the lithium-ion (Li-ion) battery cycle life, an active combination of an ultracapacitor (UC) with an energydense Li-ion battery is shown as a promising approach for electric vehicles (EVs) and plug-in hybrid electric vehicles (PHEVs). In this paper, the problem of the sizing of the Li-ion battery and UC, as well as the degree of hybridization between the UC power and battery power, is approached from a new perspective, i.e., by solving an optimization problem to minimize fuel consumption. To implement this optimized power sharing in real time, a novel energy management strategy is proposed, which includes battery power reference generation, UC state-of-charge regulation, and forecast control based on the driver commands. Finally, simulations and experiments in which the ywheel + generator + controlled load is used to emulate the vehicle drivetrain are provided to verify the reduced stress on the battery current and the improved fuel economy achieved by the proposed method. Index TermsBidirectional dcdc converter, energy management, energy storage, plug-in hybrid electric vehicles (PHEVs).
TABLE I C OMPARISON OF E NERGY S TORAGE E LEMENTS [9]

I. I NTRODUCTION HE CONCEPT of an electric vehicle (EV) or a plug-in hybrid electric vehicle (PHEV) using a battery and ultracapacitor (UC) as a hybrid energy storage system (ESS) has already been widely accepted [1][7]. Usually, the battery provides the average power or steady-state power, and the UC provides the peak or transient power. The technical and economic advantages of combination technologies for future EV/PHEVs are demonstrated in [1]. However, previous research focused on the battery cycle life and did not address fuel economy. In addition, no systematic procedure has been developed for choosing the UC size and the degree of hybridization (DH). Various batteries and UCs are analyzed in [9]. Candidate energy storage components for EV/PHEV applications are shown in Table I, along with key performance metrics. It is shown
Manuscript received September 7, 2010; revised December 9, 2010; accepted January 26, 2011. Date of publication March 3, 2011; date of current version May 16, 2011. This work was supported in part by the National Science Foundation under Grant ECCS-0641972. The review of this paper was coordinated by Prof. M. E. Benbouzid. L. Wang was with the Department of Electrical and Computer Engineering, Florida A&M UniversityFlorida State University College of Engineering, Tallahassee, FL 32310 USA. He is now with the High-Frequency Power Electronics Laboratory, GE Global Research, Niskayuna, NY 12309-1027 USA (e-mail: wang@ge.com). E. G. Collins, Jr. is with the Department of Mechanical Engineering, Florida A&M UniversityFlorida State University College of Engineering, Tallahassee, FL 32310-6046 USA (e-mail: ecollins@eng.fsu.edu). H. Li is with the Department of Electrical and Computer Engineering, Florida A&M UniversityFlorida State University College of Engineering, Tallahassee, FL 32310 USA (e-mail: hli@caps.fsu.edu). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TVT.2011.2122272

that the high-energy lithium-ion (Li-ion) battery has the highest specic energy and energy density, whereas the UC has the highest specic power. In addition, Li-ion batteries have the best cycle life of batteries, whereas the UC cycle life is much higher than the only other power-dense component, i.e., the high-power Li-ion battery. Thus, Li-ion batteries and UCs are the preferred choices in the latest EV and PHEV applications. Several energy management strategies in EV/PHEV have been proposed in previous research [3][7], [11][19]. An energy management strategy that uses both a UC and a battery for an autonomous rail vehicle is introduced in [3]. Although it is implemented in real time, it uses both a PC and a dSPACE system to realize the control functions. Thus, it is costly to implement in passenger cars. Furthermore, the fuel economy obtained by using a UC in conjunction with a battery is not clearly analyzed. In addition, the sizing of the UC and battery and their effect on the total vehicle fuel economy have not been addressed. This paper rst introduces an optimization problem that maximizes fuel economy based on the choice of the following three parameters: 1) the total weight of the ESS; 2) the UC weight ratio nmass , which is the percentage of the total ESS weight due to the UC; and 3) the DH, which is the percentage of the power provided by the ESS due to the UC. By solving this problem, both the total ESS weight and the sizing of the UC can be determined based on vehicle size and driving cycle. In actual vehicle operations, the optimal DH varies based on the driving prole. To implement this optimized power sharing in real time, a novel energy management strategy is proposed. The key is vehicle load power estimation, because the load power prole is required to solve the optimization problem. The load power estimation is based on the past load power information and current system efciency. This energy management strategy includes battery power reference generation,

0018-9545/$26.00 2011 IEEE

1420

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 4, MAY 2011

Fig. 1. Proposed EV/parallel PHEV system conguration.

UC state-of-charge (SOC) regulation, and forecast control based on the driver commands. The target is not only fuel economy but also the extended cycle life for the Li-ion battery, which results from enabling the UC to provide power at times instead of constant reliance on the battery. The proposed approach can be applied to EVs and PHEVs in the all-electric range (AER). The proposed energy management strategy was veried using an EV simulation implemented with a real-time digital simulator (RTDS) and hardware experiments on an EV test bed. In the experiments, the vehicle electric drive was emulated by a 10-kW permanent-magnet synchronous motor (PMSM) dynamometer. This paper is organized as follows. Section II describes the EV system structure and the ofine optimization problem used to size the ESS. Section III describes the real-time energy management strategy used to achieve a specied DH, along with other key design targets. Section IV presents both simulation and experimental results that demonstrate the efcacy of the proposed approach. Finally, Section IV presents the conclusions. II. S YSTEM S TRUCTURE AND E NERGY S TORAGE S IZING D ESIGN The system conguration of the proposed EV (or parallel PHEV) is shown in Fig. 1. A ywheel is added between the motor and the generator to represent the vehicle inertia. A highenergy Li-ion battery is chosen as the main energy source, and a UC is connected to the inverter dc bus through a bidirectional dcdc converter. The AER is designed to be 20 mi for a midsize passenger car in this paper. The energy requirement of this AER can be estimated using the ADVISOR software [10] or E= = 1 1 Pd dt 1 ma + CR mg + mg sin + a CD AF v 2 v dt (1) 2
Fig. 2.

TABLE II PARAMETERS FOR THE V EHICLE

Vehicle speed prole.

where a and v are, respectively, the vehicle acceleration and velocity, is the total system efciency, including transmission and motor, and the remaining parameters are dened in Table II. Using the parameter values given in Table II and the vehicle velocity prole in Fig. 2 to determine a and v , it follows from

(1) that E = 4.4 kWh. If a battery SOC window of 60% is assumed, then the total battery capacity is 7.3 kWh. The weight of a state-of-the-art high-energy Li-ion battery that meets this energy requirement and has the characteristics listed in Table I is then calculated as 49 kg. If the cell balancing and package components are taken into account, its nal weight is 60 kg based on the Saft VL45E cell and Intensium Flex module parameters. The battery, in conjunction with a UC, can provide power and energy that meet most of the performance requirements during normal driving conditions, particularly urban driving. However, during fast acceleration or high-grade vehicle traversal, the peak power requirement may exceed the capability of the Li-ion battery and UC, in which case, the engine will provide some of the required power. The engine size and power requirements for a midsize PHEV car are, respectively, selected as 1.6 L and 50 kW. The vehicle power Pveh , which is the sum of the battery unit (BU) power PBU and the UC power PUC , can be represented as

WANG et al.: OPTIMAL DESIGN AND REAL-TIME CONTROL FOR ENERGY MANAGEMENT IN ELECTRIC VEHICLES

1421

Fig. 3.

BU power PBU and total vehicle power Pveh at different DHs. (a) DH = 1 D2 . (b) 0 < DH < 1 D2 . (c) DH = 0.

the pulse power with a varied duty ratio D2 , as shown in Fig. 3, which also shows PBU and PUC . To analyze the optimal vehicle performance, the DH and UC weight ratio nmass are formally dened as DH =

therefore, the average UC power has to be zero to maintain its SOC during a long period. Eloss can be written as
D1 1

PUC PBU + PUC

nmass =

MUC MESS

Eloss =
0

i2 BU RBU +
D1

i2 BU RBU
D1 1

(2)

where PUC is the UC power during the D2 period, PBU is the battery power during the D2 period, MUC is the weight of the UC, and MESS is the total weight of the hybrid ESS, including both the battery and UC. Vehicle fuel economy is dependent on MESS , nmass , and DH. Let FE denote the fuel economy (e.g., in mi/gal) such that 1/FE is a measure of the fuel expenditure per distance traveled. Then, the corresponding optimization problem is 1 MESS ,nmass ,DH FE min where Mveh FE = FEbase ESS Mveh + MESS (4) (3)

+
0

i2 UC RUC

+
D1

i2 UC RUC .

(7)

It can be shown as
2 Eloss = IBU _avg

(1 DH)2 DH2 + D2 D1
2 + IBU _avg

RBU RUC (8)

DH2 DH2 + D1 D2

which is equivalent to
2 Eloss = IBU _avg

(1 DH)2 DH2 + D2 D1 DH2 DH2 + D1 D2

V 2 0.05 SBU (1 nmass ) MESS V 2 0.05 SUC nmass MESS (9)

and FEbase is the base fuel economy at base vehicle weight and 100% ESS efciency, ESS is the ESS efciency, given by ESS = Pavg Pavg + Eloss /T (5)

2 + IBU _avg

where Eloss is the ESS loss at unit time, T is the per unit pulse duration, given by T = D1 + D2 = 1 (6)

and Pavg is the average vehicle load power. Fig. 2 shows different battery power proles when the DH changes. The difference between the vehicle power and the battery power is provided by the UC. D1 and D2 are the nondimensional time intervals for the zero vehicle power and pulse vehicle power, respectively, and are normalized as described in (6). Note that the average BU power is always equal to the average vehicle power, because the UC has low energy density;

where IBU_avg is the average battery current of a power prole shown in Fig. 2, RBU is the equivalent dc impedance of the BU, RUC is the equivalent dc impedance of the UC, V is the dc bus voltage (300 V), and SBU and SUC are, respectively, the specic power of the battery and UC at 95% efciency. Note that the battery impedance is a variable of SOC and temperature. Because it is hard to consider all kinds of situations, e.g., driving cycles, battery power prole, SOC, and temperature, in advance, an estimated average temperature and battery SOC have to rst be picked to decide its impedance in the sizing algorithm. However, RBU can be considered a dynamic variable in the real-time control algorithm described in Section III, in which the DH can be derived in a real-time calculation. The solution of the optimization problem (3) changes if D1 of the power prole changes. It is assumed that the vehicle is driven in either urban conditions (D1 = 0.5) or highway

1422

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 4, MAY 2011

Fig. 4. Total energy loss of hybrid ESS versus DH and nmass for urban driving (D1 = 0.5) with MESS = 67.2 kg, which is its optimal value. Fig. 6. Vehicle fuel economy versus MESS for (nmass , DHU , DHH ) = (0.128, 0.247, 0.049), corresponding to the optimal parameter values.

take into account both urban and highway driving, we can solve the optimization problem as min MESS ,nmass ,DHU ,DHH 0.8 0.2 + FEU FEH (10)

Fig. 5. Total energy loss of hybrid ESS versus DH and nmass for highway driving (D1 = 0.1) with MESS = 79.7 kg, which is its optimal value.

conditions (D1 = 0.1). By using the MATLAB function fmincon to solve this problem, the optimal triple (DH, nmass , and MESS ) is found as (0.29, 0.17, and 67.2 kg) for urban driving and (0.02, 0.04, and 79.7 kg) for highway driving. To illustrate the benets of properly choosing the optimization parameters, Fig. 4 shows the energy loss as a function of DH and nmass for urban driving when MESS = 67.2 kg, which is its optimal value, and Fig. 5 shows the same information for highway driving when MESS = 79.7 kg, which is its optimal value. These gures reveal that, if DH or nmass are not properly chosen, the energy loss can be two times higher than when DH or nmass is optimized. The optimum values of nmass and DH tend to decrease when the variation in the vehicle speed narrows and the speed approaches highway speeds. Although DH can always be adjusted online to meet changing drive conditions, the UC and BU sizes cannot be changed, which requires a compromise between urban and highway optimization results. Based on the Environmental Protection Agency (EPA) combined driving cycle estimation in the AER of PHEVs, 80% of the distance driven is in urban environments and 20% on highways in this paper. Hence, to obtain MESS and nmass that

where DHU and DHH are, respectively, the DH for urban and highway driving, and FEU and FEH , respectively, represent the fuel economies, with DH = DHU and DH = DHH . The optimal value of (MESS , nmass , DHU , DHH ) is given by (69.8 kg, 0.128, 0.247, 0.049). To show the effect of MESS on the fuel economy, Fig. 6 displays the curve of FE versus MESS , with (nmass , DHU , DHH ) = (0.128, 0.247, 0.049), corresponding to the optimal parameter values. Note that, if MESS is increased past its optimum value of 69.8 kg, the fuel efciency steadily decreases. The optimization problems given by (7)(9) can be used to determine MESS and nmass , targeting the best fuel economy. However, if the sizing of the Li-ion battery is based on the AER requirement of 20 mi instead of the best fuel economy, the battery weight should be selected as 49 kg. Then, the optimization function targeting fuel economy reduces to min 0.8 0.2 + FEU FEH (11)

MUC ,DHU ,DHH

where the FE is given by (4) and (5), with Eloss given by


2 Eloss = IBU _avg

(1 DH)2 DH2 + D2 D1 DH2 DH2 + D1 D2

V 2 0.05 BUW/kg 49 (12)

2 + IBU _avg

V 2 0.05 UCW/kg MUC

where MUC is the UC weight, and V is the dc bus voltage (300 V). Equation (12) is obtained from (9) by substituting (1 nmass ) MESS with 49 kg and nmass MESS with MUC . By using the MATLAB function fmincon, the optimal solution of (11) is given by (MUC , DHU , DHH ) = (15 kg, 0.26, 0.051).

WANG et al.: OPTIMAL DESIGN AND REAL-TIME CONTROL FOR ENERGY MANAGEMENT IN ELECTRIC VEHICLES

1423

Fig. 7.

Control diagram of the proposed energy management in EV.

A UC with the optimal weight of 15 kg and the characteristics of the UC in Table I has an energy storage capacity of 67 Wh. III. R EAL -T IME E NERGY M ANAGEMENT S TRATEGY The focus of the previous section was determining the optimal values of MESS and nmass , which are parameters that are xed and do not vary during real driving. However, the previous section also showed that, to achieve the best fuel economy, the DH should vary with the driving conditions. The parameters that affect the optimal DH include D1 , RBU , and RUC . Note that the resistances RBU and RUC are not constants and can be varied with SOC and temperature; a real-time calculation of DH is therefore required and more accurate than a lookup table. In the following discussion, we describe a novel real-time energy management strategy to achieve the following objectives: 1) Find the optimal or near-optimal DH in real time for the best fuel economy and maintain proper power sharing between the BU and UC to achieve the desired DH; 2) control the UC SOC, because its voltage needs to be maintained in a specic range; and 3) minimize dc bus voltage overshoot through acceleration and deceleration forecast control. The EV system and control diagram are shown in Fig. 7. Among these objectives, only the rst objective is targeted for DH and fuel economy, whereas the second and third objectives are system requirements. The second objective requires the UC SOC to meet the following two requirements: 1) Keep the SOC high enough to provide the required peak power when the acceleration event starts, and 2) keep the SOC low enough to recuperate braking energy when a deceleration event begins. All these objectives are achieved by controlling the dcdc converter in real time. In this system, the Li-ion battery is connected to the inverter dc bus through a dc boost converter, and the UC is connected to the dc bus through a bidirectional dcdc converter. This bidirectional dcdc converter is a phase-shift-controlled softswitching converter. Because the converter between the battery (BU) and the dc bus is unregulated and operates as an ideal

dc transformer, the battery can be treated as the one directly connected to the dc bus. The only control variable is the phaseshift angle of the dcdc converter interfacing UC, which is utilized to control the UC power. Because IBU = Iload IUC and Iload is derived from the driving input conditions, as long as IUC is controlled through , IBU is indirectly regulated to a safe region. The bidirectional dcdc converter topology is shown in Fig. 8. Compared with the voltage-fed dual-active-bridge (DAB) converter, this current-fed DAB converter is implemented with a dual half bridge (DHB) instead of a dual full bridge, resulting in a reduced number of devices and, therefore, reduced gate drivers and auxiliary power supplies to lower the cost and increase the power density. In addition, it injects and draws smooth current with a reduced current ripple from the source, which is preferred for ESS applications. Furthermore, to reduce the switching loss of this dcdc converter, its switches are in the soft-switching condition, which is obtained without adding additional devices. The operation principle is illustrated in Fig. 8(b) and (c) with the primary-referred simplied circuit model of the converter and the transformer voltages. The bidirectional power ow is controlled by the phase-shift angle of the DAB and the transformer leakage inductance Lr . The details of the control of the phase-shift angle in this dcdc converter are shown in Fig. 9. First, it needs to calculate the optimal DH and then make the battery power follow its reference P . The analysis in Section II and in Figs. 4 and 5 shows that the optimal DH depends on D1 , RBU , and RUC , i.e., the driving cycle and power information; therefore, one of the keys is to estimate D1 , which is achieved by collecting the past vehicle drive prole (speed) and current information (Iload ) and using this value to estimate the current vehicle peak speed, average speed, and stop interval D1 . This process is repeated every 1 s. The controller (DSP F2812) then uses the driving-cycle estimates to conduct a real-time search for DH that minimizes Eloss given by (8). In this calculation, RBU can dynamically be adjusted based on the current battery SOC (IBU integration method), In the future, it can further be adjusted by taking into account temperature. DH is initialized as DH(0) using Table III. As illustrated in Fig. 10, it is changed by a xed step size (DH) for a given number of steps N to yield DH(1), DH(2), . . ., DH(N). Note that N = 3 in Fig. 10 and the reported experiments. The optimal DH is chosen as DH(i ), where i is the solution to
i{1,...,N }

min

Eloss (DH(i)) .

(13)

If the optimal DH is obtained, the optimal battery power reference P can be derived from the load power Pload (1 DH), as shown in Fig. 9. The power reference is then converted . to the battery current reference IBU Second, the UC SOC needs to be controlled to follow its reference UC _SOC . This reference is not a constant value but varies, depending on the vehicle speed to prepare for oncoming acceleration or deceleration events, as shown in Fig. 9. Third, fast acceleration or deceleration often causes voltage transients on the inverter dc bus, or even worse, they might cause the battery current to greatly increase. This happens

1424

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 4, MAY 2011

Fig. 8. (a) Developed two-port current-fed DHB converter (C1 = C2 , C3 = C4 ). (b) Primary-referred simplied AC model of the converter. (c) Transformer primary and secondary voltages.

Fig. 9. Energy management and control of the bidirectional dcdc converter. TABLE III DH(0) L OOKUP TABLE [BASED ON (3)]

because the UC converter response is based on the load power measurement; therefore, it always lags the transient. The forecast (feedforward) control makes the UC converter

respond to not only present conditions but future conditions as well. For example, when the driver steps on the gas pedal, it will trigger the UC converter forecast control to increase its output power. Although the load power has not yet changed, it will become larger in the future; therefore, this advanced UC power increase will be simultaneous with the load power. Then, the overshoot/undershoot on the inverter dc bus will be minimized.

WANG et al.: OPTIMAL DESIGN AND REAL-TIME CONTROL FOR ENERGY MANAGEMENT IN ELECTRIC VEHICLES

1425

A. Simulation Results The simulation models are shown in Fig. 12. The UC and BU dynamic models are shown in Fig. 12(a) and (c), respectively [21]. The parameter values in the models in Fig. 12(a) and (c) were derived from the physical device measurement in the laboratory. The bidirectional DHB converter average model is shown in Fig. 12(b). The electric motor and inverter are represented by traditional mathematical models. The bus converter interfacing the BU is represented by an ideal dc transformer. The simulation results using only a battery are rst compared with the results obtained from an optimally designed hybrid ESS. The results are shown in Figs. 1315. Furthermore, the DH for this hybrid ESS is based on real-time optimization. It is shown in Fig. 13 that, when the only power source was a Li-ion battery, the battery peak power was 1.5 kW. For a 40-V battery, the peak current is 37 A. In contrast, Fig. 13 shows that the hybrid ESS approach yielded a peak power of only 900 W, corresponding to a reduced peak current of 22.5 A. Furthermore, the battery current slope in the latter case was slower compared with the battery-only method, particularly during the braking period. Hence, the hybrid ESS approach led to an increase in the Li-ion battery cycle life. In addition, Fig. 14 shows that the total energy in the proposed hybrid ESS is 66 kJ, whereas the total energy when only the Li-ion battery was used is 67 kJ under the ECE40 cycle. Therefore, the proposed method can achieve improved fuel economy for EVs or PHEVs in the AER, which results in 2% energy savings. The improvement in fuel economy in the simulation is relatively low, because an average model of the dcdc converter is used in the simulation, which does not enable the dcdc converter loss, particularly the switching loss, to be emulated. In a real system, the active resistance introduced by the dcdc converter would affect the fuel economy, and higher fuel economy improvement can be achieved, as shown in the experimental results. Fig. 15 also shows the energy loss for both the hybrid optimal approach (1.8 kJ) and when only the battery is used (2.8 kJ). Figs. 1315 correspond to the optimal DH given by (13), whereas Figs. 1618 correspond to a nonoptimal DH that was chosen to use the UC as much as possible while also limiting the BU current slope and maintaining the BU peak current to around the average load current. By comparing Fig. 13 with Fig. 16, it is shown that the nonoptimal DH is much bigger than the optimal DH; thus, the peak UC power in the nonoptimal case is much bigger. Although the peak BU power and the average BU power are similar in both cases, the loss in the UC overwhelms the energy savings achieved due to the lower peak BU power. By comparing Fig. 14 with Fig. 17, it is shown that, although the Energy provided by the battery (EBU) that corresponds to the optimal DH in Fig. 14 (62 kJ) is higher than the nonoptimal DH control in Fig. 17 (53 kJ), the optimal DH reduces the total energy consumption from 69 kJ to 66 kJ, which results in 6% energy savings. This case is because the goal of optimal power allocation between the BU and the UC is to separately achieve the minimum total system loss instead of minimizing the Energy provided by the ultracapacitor (EUC) or the EBU. By comparing Fig. 15 with Fig. 18, it

Fig. 10. Real-time iteration of the DH.

Fig. 11. Control loop for the BU current.

The control-loop design for the BU current is shown in Fig. 10. Because the converter is nonlinear, its small-signal plant function is derived as V12 ( 2) 1 iBU 4Lr n (14)

where n is the transformer turn ratio between v3 and v1 , = 2f , and f is the switching frequency. The compensation function in Fig. 11 is Gc (s) = K 1 + where K = 80, and T = 0.1 ms. IV. S IMULATION AND E XPERIMENTAL R ESULTS The proposed EV system was simulated using an RTDS. The parallel calculation capability of RTDS results in fast and accurate simulation results. The driving prole is based on the European Urban ECE40. Although the dynamics of ECE40 is low, it is quite representative of urban driving and, hence, yields useful results for the case in which an EV or a PHEV in the AER is primarily used for commuting. The vehicle power in simulation is downscaled to enable comparison with experimental results with a similar power rating. The downscaling factor is mostly based on the vehicle mass, i.e., all the inertial and load torque are derived based on a much smaller vehicle weight. As long as the same speed prole is used for either the real vehicle or the downscaled one in the simulation/experiment, they will have different levels (in kilowatts per second) in acceleration. However, the key for an emulated vehicle is the power peakto-average ratio at the same speed prole, which is shown as 2 in both the simulation and experimental results, which is quite close to the normal ECE40 estimation. The UC and BU sizes are chosen based on the analysis in Section II, and they are also proportionally reduced. 1 Ts (15)

1426

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 4, MAY 2011

Fig. 12. Simulation models in RTDS. (a) UC dynamic model. (b) DHB converter average model. (c) BU dynamic model.

Fig. 14. RTDS simulation results. Energy requirements for the optimal DH hybrid ESS system and the ESS with only a Li-ion battery.

economy than the pure-battery solution since cycle life. On the other hand, the optimal DH solution shows better fuel economy than the pure-battery solution. B. Experimental Results The EV drivetrain was built in the laboratory, as shown in Fig. 19. U-Charge RTs from Valence were the selected Li-ion batteries, and three are in series with the nominal voltage 40 V. A bidirectional dcdc bus converter was built in the laboratory to boost the battery voltage to 320 V. The UC is a BMOD0430E16 (16 V/430 F) from Maxwell. A 5-kW bidirectional dcdc converter was used to interface the UC to the inverter DC bus. The energy management strategy was implemented using a TMS320F2812 DSP. The generator load was controlled to emulate real road load torque. Because the simulation results shows that the fuel economy of a nonoptimal DH method is less than the fuel economy of a

Fig. 13. RTDS simulation results. Power requirements for the optimal DH control hybrid ESS system and the ESS with only a Li-ion battery.

is shown that the optimal DH reduces the energy loss of the hybrid ESS from 4.7 kJ to 1.8 kJ. It also veries that the choice of DH may signicantly affect the system loss. A conclusion can therefore be made from the aforementioned results that the nonoptimal DH solution selected in this objective of the former case is focused on improving battery power, which has less fuel

WANG et al.: OPTIMAL DESIGN AND REAL-TIME CONTROL FOR ENERGY MANAGEMENT IN ELECTRIC VEHICLES

1427

Fig. 17. RTDS simulation results. Energy requirements for the hybrid ESS system in the nonoptimal DH control.

Fig. 15. RTDS simulation results. Joule losses of the optimal DH hybrid ESS and the ESS with only the Li-ion battery.

Fig. 18. RTDS simulation results. Energy loss of the hybrid ESS in the nonoptimal DH control.

Fig. 16. RTDS simulations. Power requirements for the hybrid ESS system in the nonoptimal DH control. Fig. 19. Experimental test bed.

pure-battery solution, the optimal DH and pure-battery methods are selected in the experimental verications for comparison. Fig. 20 shows the system performance with only a Li-ion battery, where the battery peak current is 32 A. Fig. 21 shows the system performance with a Li-ion battery and optimized

UC, where the DH is calculated as the optimal value and is implemented in real time. The battery peak current is 15 A. Furthermore, the battery current slope is smaller than the batteryonly method, which leads to an improved battery cycle life.

1428

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 4, MAY 2011

Fig. 22. System efciency and ESS energy provided by the BU-only solution (method 1) and optimal DH solution (method 2) during a 200-s ECE40 cycle.

is 55.9 kJ. This result shows that the optimal DH method has better fuel economy. The load energy is 50.8 kJ in both cases. It can be derived from (16) that the system efciency in the purebattery method is 85.5%, whereas the efciency for the optimal DH method is 91%, which is also shown in Fig. 22.
Fig. 20. method. Experimental results. Pload and IBU of the Li-ion-battery-only

V. C ONCLUSION This paper has developed a new hybrid ESS design method for EVs, which can also be applied to PHEVs in the AER operation. Traditional methods focused on improving battery cycle life when designing the hybrid ESS. The goal of the proposed method is not only to increase ESS cycle life but to reduce the total ESS loss as well, leading to improved equivalent fuel economy. This objective has been achieved by the following three steps. Develop and solve an optimization problem to improve the fuel economy based on driving-cycle estimation, where the ESS weight and the UC weight ratio can accordingly be found. The traditional method chose the size of the battery and the UC based on the peak power requirement and allocation. Propose the DH as a variable, and calculate it in real time to achieve the minimum system loss. The traditional method treated it as a constant or selected it based on a lookup table. Integrate UC SOC control and forecast response according to the driver command into the system control strategy. Simulation and experimental results showed that an optimally chosen Li-ion battery and UC can reduce Li-ion battery peak and root-mean-square (RMS) current stress, thus improving its cycle life. Furthermore, improved fuel economy is achieved in an optimal DH-control-based hybrid ESS compared with the fuel economy achieved when only a Li-ion battery is used. R EFERENCES
[1] J. M. Miller, U. Deshpande, T. J. Dougherty, and T. Bohn, Powerelectronic-enabled active hybrid energy storage system and its economic viability, in Proc. IEEE Appl. Power Electron. Conf., 2009, pp. 190198. [2] T. Markel and A. Simpson, Plug-in hybrid electric vehicle energy storage system design, presented at the Advanced Automotive Battery Conf., Baltimore, MD, May, 2006. [3] C. Romaus, J. Bocker, K. Witting, A. Seifried, and O. Znamenshchykov, Optimal energy management for a hybrid energy storage system combining batteries and double-layer capacitors, in Proc. IEEE Energy Convers. Congr. Expo., 2009, pp. 16401647. [4] J. Cao and A. Emadi, A new battery/ultracapacitor hybrid energy storage system for electric, hybrid and plug-in hybrid electric vehicles, in Proc. IEEE Veh. Power Propulsion Conf., 2009, pp. 941946.

Fig. 21. Experimental results. Pload , IUC , IBU , and VUC at the optimal DH control hybrid ESS method.

All of the peak power during the transient is provided by the UC, particularly during braking periods. The UC recuperates most of the braking energy during the driving cycle, because it has high power capability and high charging efciency. The UC SOC is also controlled, and it changes from 12.4 V to 12 V after a 200-s driving cycle. Fig. 22 compared the total provided energy and system efciency between these two methods. The system efciency sys and energy Estorage provided by the ESS are dened as sys = ESS converter inverter = Estorage = EBU , EBU + EUC , EMotor Estorage (16)

(method 1, BU only) (method 2, BU + UC).

In the pure-battery method, EBU is 59.8 kJ during a 200-s ECE40 cycle. In the optimal DH control hybrid ESS method EBU is 54 kJ, and EUC is 1.9 kJ; therefore, the total ESS energy

WANG et al.: OPTIMAL DESIGN AND REAL-TIME CONTROL FOR ENERGY MANAGEMENT IN ELECTRIC VEHICLES

1429

[5] J. Moreno, M. E. Ortuzar, and J. W. Dixon, Energy-management system for a hybrid electric vehicle, using ultracapacitors and neural networks, IEEE Trans. Ind. Electron., vol. 53, no. 2, pp. 614623, Apr. 2006. [6] R. M. Schupach, J. C. Balda, M. Zolot, and B. Kramer, Design methodology of a combined batteryultracapacitor energy storage unit for vehicle energy management, in Proc. IEEE Power Electron. Spec. Conf., 2003, vol. 1, pp. 8893. [7] X. Li and S. Williamson, Efciency and suitability analyses of varied drivetrain architectures for PHEV applications, in Proc. IEEE Veh. Power Propulsion Conf., 2008, pp. 16. [8] Y. Gao and M. Ehsani, Design and control methodology of plug-in hybrid electric vehicles, in Proc. IEEE Veh. Power Propulsion Conf., 2008, pp. 16. [9] A. F. Burke, Batteries and ultracapacitors for electric, hybrid, and fuel cell vehicles, Proc. IEEE, vol. 95, no. 4, pp. 806820, Apr. 2007. [10] Advisor (Advanced Vehicle Simulator), ver. 2002, Golden, CO, 2002. [11] P. Corbo, F. E. Corcione, F. Migliardini, and O. Veneri, Experimental assessment of energy management strategies in fuel cell propulsion systems, J. Power Sources, vol. 157, no. 2, pp. 799808, Jul. 2006. [12] S. A. Rahman, N. Zhang, and J. Zhu, A comparison on fuel economy and emissions for conventional hybrid electric vehicles and the UTS plug-in hybrid electric vehicle, in Proc. IEEE Int. Conf. Comput. Autom. Eng., 2010, pp. 2025. [13] Q. Gong, Y. Li, and Z. Peng, Trip-based optimal power management of plug-in hybrid electric vehicles, IEEE Trans. Veh. Technol., vol. 57, no. 6, pp. 33933401, Nov. 2008. [14] E. Tara, S. Shahidinejad, S. Filizadeh, and E. Bibeau, Battery storage sizing in a retrotted plug-in hybrid electric vehicle, IEEE Trans. Veh. Technol., vol. 59, no. 6, pp. 27862794, Jul. 2010. [15] A. Emadi, Y. Lee, and K. Rajashekara, Power electronics and motor drives in electric, hybrid electric, and plug-in hybrid electric vehicles, IEEE Trans. Ind. Electron., vol. 55, no. 6, pp. 22372245, Jun. 2008. [16] A. Khaligh and Z. Li, Battery, ultracapacitor, fuel cell, and hybrid energy storage systems for electric, hybrid electric, fuel cell, and plug-in hybrid electric vehicles: State of the art, IEEE Trans. Veh. Technol., vol. 59, no. 6, pp. 28062814, Jul. 2010. [17] Y. Lee, A. Khaligh, and A. Emadi, Advanced integrated bidirectional AC/DC and DC/DC converter for plug-in hybrid electric vehicles, IEEE Trans. Veh. Technol., vol. 58, no. 8, pp. 39703980, Oct. 2009. [18] Z. Amjadi and S. Williamson, Power-electronics-based solutions for plug-in hybrid electric vehicle energy storage and management systems, IEEE Trans. Ind. Electron., vol. 57, no. 2, pp. 608616, Feb. 2010. [19] J. Kessels, M. Koot, P. van den Bosch, and D. Kok, Online energy management for hybrid electric vehicles, IEEE Trans. Veh. Technol., vol. 57, no. 6, pp. 34283440, Nov. 2008. [20] J. Lai and D. Nelson, Energy management power converters in hybrid electric and fuel cell vehicles, Proc. IEEE, vol. 95, no. 4, pp. 766777, Apr. 2007. [21] P. L. Moss, G. Au, E. J. Plichta, and J. P. Zheng, Investigation of solid electrolyte interfacial layer development during continuous cycling using ac impedance spectra and microstructural analysis, J. Power Sources, vol. 189, no. 1, pp. 6671, Apr. 2009.

Lei Wang (S09M11) received the B.S. degree in electrical engineering from Shanghai Jiaotong University, Shanghai, China, in 1996, the M.S. degree in electrical engineering from Zhejiang University, Hangzhou, China, in 2002, and the Ph.D. degree in electrical engineering from Florida State University, Tallahassee, in 2010. He is currently with the High-Frequency Power Electronics Laboratory, GE Global Research, Niskayuna, NY. His research interests include the modeling and control of bidirectional dcdc converters, fuel cell hybrid vehicles, and plug-in hybrid electric vehicles.

Emmanuel G. Collins, Jr. (M83SM99) received the Ph.D. degree in aeronautics and astronautics from Purdue University, West Lafayette, IN, in 1987. For seven years, he was with the Controls Technology Group, Harris Corporation, Melbourne, FL, before joining the Department of Mechanical Engineering, Florida A&M UniversityFlorida State University College of Engineering, Tallahassee, where he is currently the John H. Seely Professor and the Director of the Center for Intelligent Systems, Control and Robotics. His research interests include the navigation and control of autonomous vehicles (ground, water, and air) in extreme environments and situations, multirobot and humanrobot cooperation, energy-efcient control of hybrid vehicles, and the control of aeropropulsion systems.

Hui Li (S97M00SM01) received the B.S. and M.S. degrees in electrical engineering from Huazhong University of Science and Technology, Wuhan, China, in 1992 and 1995, respectively, and the Ph.D. degree in electrical engineering from the University of Tennessee, Knoxville, in 2000. She is currently an Associate Professor with the Department of Electrical and Computer Engineering, Florida A&M UniversityFlorida State University College of Engineering, Tallahassee. Her research interests include bidirectional dcdc converters, cascaded multilevel inverters, and power electronics applications in hybrid electric vehicles.

Você também pode gostar