An energy management of plug-in hybrid electric vehicles based on driver behavior and road information

Abstract

Plug-in hybrid electric vehicles (PHEVs) provide the flexibility to be used as pure electric vehicles in daily commuting, as well as common hybrid vehicles (HEVs) on longer trips. To appropriately distribute power between the engine and the energy buffer (normally battery) without sacrificing fuel economy, optimal control strategies of varies kinds have been proposed in recent years. Some strategies, however, cannot be applied in real time environment for the tightly independent on the future driving information. Some others could identify the different road condition but the identification accuracy is questionable, making their real-world effectiveness could suffer significantly. Moreover, the proposed energy management method seldom consider the effect from the driver behavior, which is an essential factor that can determine the vehicle energy consumption. For such reasons, this paper proposes a novel energy management strategy, the instantaneous optimization method—Equivalent Consumption Minimization Strategy (ECMS) is chosen as the foundation solution to the PHEV energy management problem. Then the paper studies the effect from different driver behavior and uses fuzzy logic method to identify different drive behavior. Combined with the pre-knowledge provide by Intelligent Transport System (ITS), the equivalent factor, which is the core of ECMS, is adjusted, making the real instantaneous application of ECMS. Then the simulation is performed to prove its viability and effectiveness.

Keywords

Equivalent Consumption Minimization Strategy (ECMS)Intelligent Transport System (ITS)equivalent factor fuzzy logic traffic flow theory

1 Introduction

In cities, plug-in hybrid electric vehicles (PHEVs) can be an emerging new mean of transportation, which combines the merits of pure electric vehicles (EVs) and hybrid electric vehicles (HEVs). PHEVs typically have long pure electric and combined driving mileage, good equivalent fuel economy, and most importantly, the ability to be used as pure electric vehicle (PEV) in daily commuting, and as HEV on occasional longer trips [1 –4].

In the research of PHEVs, there are a number of challenging issues. Among them, energy management is one of the key problems [5 –32]. The control strategy problem can be summarized as follows:

How to fulfill tractive power demand with the best possible fuel economy.

How to keep the battery state of charge (SOC) in a certain range during the charge sustaining (CS) stage, while at the same time ensuring that the PHEV has the ability to follow driving power requirements under different driving conditions.

To tackle these issues, great amounts of work has been done in the past 20 years by researchers from all over the world. Among the mass of research, three types of control method are favored: the control strategy based on heuristic knowledge, i.e. rule based control and fuzzy logic control; the control strategy based on numerical optimization, i.e. dynamic programming (DP); the control strategy based on analytical optimization, i.e. the Pontryagin’s minimum principle (PMP) [5 –20]. Regarding control strategies based on heuristic knowledge, some rule based control strategies have been proposed in [5 –7], and several fuzzy-logic control strategies are adopted in [8 –10]. The control strategies based on heuristic knowledge are relatively simple and typically feature fast real-time calculation. However, they mainly base on expert experience, which means they are not robust against diverse and changing driving conditions. Regarding control strategies based on numerical optimization, some optimal control methods like DP have seen application in [11 –14]. Although optimal control theories like DP provide global optimization, they require information in advance about the driving cycle, which means they are also less adapted to diverse and changing driving conditions. Regarding control strategies based on analytical optimization, some researchers chose PMP theory to perform the control with acceptable amounts of calculation and hypothesis [15 –20]. The PMP, nevertheless, is a global optimal solution as well, which is why its effectiveness is also largely affected by the actual driving conditions. To eliminate the limit by the requirement of driving condition, some researchers bring the Equivalent Consumption Minimization Strategy (ECMS) into application [21 –23]. The ECMS make the global optimization problem into instantaneous calculation. The ECMS, however, cannot fully break away from the dependence on the pre-knowledge about the future road condition for the equivalent factor in ECMS relies on the future driving behavior [24, 25]. Based on former research about the PHEV control strategies, some researchers try to eliminate unfavorable influence of actual road condition by taking the characteristic features of typical drive cycles into account when designing the control strategy for PHEV. In [26 –34], the road condition is appropriately considered by investigating the characteristic features of different driving cycles using statistic method. Then the vehicle control unit (VCU) identifies the actual road condition using the characteristic features of different driving cycle, and performs the corresponding optimal control with regards to that specific road condition. This method, in certain degree, can solve the influence on the energy management of PHEV from the changing future road condition. However, this method fully depends on the studying into the typical drive cycles which cannot reveal the actual road condition.

The driver behavior that can influence the vehicle driving safety and fuel consumption has been studied in recent years [35 –38]. The drivers with different behavior tend to drive the vehicle in different ways, causing the fuel consumption be various. To be specific, drivers who prefer to drive the vehicle in sport style may require the vehicle can response the driving requirement fast, leading to more fuel be consumed. The drivers in moderate driving, on the contrary, prefer to drive the vehicle in mild method do not require the vehicle fast response the driving requirement, saving the fuel in some degree. To the PHEVs that is driven by two power sources, the driver behavior can also be a vital factor that determines the fuel consumption. Hence, it is suggested that study different driver behavior and distribute the power between the engine and the battery in PHEV accordingly.

Together with the evolution of Global Position System (GPS), Geographic Information System (GIS), the Intelligent Transport System (ITS) becomes mature, making the future driving information can be provided in detail, and previous limit on the energy management by solved with novel methods. In this paper, the ECMS is applied as the solution to the PHEV energy management problem. According to the different future driving information acquisition level and different driving behavior, a multi-level equivalent factor adjustment method is proposed, contributing to the accurate ECMS implantation.

The remainder of the paper is organized as follows. The next section deals with the model of the PHEV. In Section 3, different driver behaviors are identified by fuzzy logic with the aid from multivariate statistical analysis method. In Section 4, the ECMS and multi-level adjustment method for equivalent factor is discussed. In Section 5, the effectiveness of the proposed adjustable ECMS is analyzed and evaluated through Matlab/Simulink simulations.

2 PHEV model

The PHEV studied in this paper is the parallel type. Schematic diagram of a typical parallel PHEV is shown in Fig. 1.

Fig.1

Schematic diagram of a typical PHEV.

In the parallel PHEV, both engine and motor can drive the vehicle alone. Meanwhile, the engine can drive the motor in generating mode to charge the battery. The engine is model by the Willans line approach: $P_{eng} = {eQ}_{lhv} {\dot{m}}_{f} - \frac{V_{d} P_{ml} w_{e}}{4 π}$ (1) $P_{ml} = {\begin{matrix} P_{ml 0} & P_{eng} \neq 0 \\ 0 & P_{eng} = 0 \end{matrix}$ (2) where e is the intrinsic energy conversion efficiency, P_ml is the mean friction pressure, P_eng, V_d, w_e, Q_lhv, ${\dot{m}}_{f}$ represents engine power, displacement, rotation speed, fuel lower heating value, and mass flow rate.

The motor power P_mot can be expressed by the battery power P_batt and the electric machine efficiency η_em, which is shown in Equation 3. $P_{mot} = {\begin{matrix} P_{batt} η_{em} & discharge \\ \frac{P_{batt}}{η_{em}} & charge \end{matrix}$ (3)

The battery is modeled as an open circuit voltage source. The battery power can be written as: $P_{batt} = U_{0} I - R_{i} I^{2}$ (4) where U₀ is the open circuit voltage, R_i is the internal resistance, I is the battery current. In Equation 5, the battery current can be express as: $I = \frac{U_{0} - \sqrt{U_{0}^{2} - 4 R_{i} P_{batt}}}{2 R_{i}}$ (5)

The battery state of charge (SOC) is an important parameter to reveal the battery state, which is expressed as: $S \dot{OC} = - \frac{I}{Q_{batt}}$ (6) where Q_batt stands for battery capacity.

3 Driver behavior identification

In previous research about the driver behavior, the self-reported questionnaire is an ideal tool to investigate driver’s behavior [35 –38]. In this paper, we also design a self-reported questionnaire to comprehend the different driving style, which is shown in Table 1. In this self-reported questionnaire, the maximum value and minimum value of each item is 5 and 1, respectively.

Table 1
The self-reported questionnaire

Item Content

1 Too close to the front vehicle

2 Driving in the city with the speed larger than permitted

3 Driving in the highway with the speed larger than permitted

4 Change lane frequently when driving the vehicle

5 Run the red light

6 Driving the vehicle in the speed closed to permitted value

7 Racing with other car for personal grudge

8 Whistling frequently to express the dissatisfied for the low speed of the front vehicle

9 Getting too close to the front vehicle when braking

10 Forced to accelerate by surround vehicles that the speed are more fast

Item	Content
1	Too close to the front vehicle
2	Driving in the city with the speed larger than permitted
3	Driving in the highway with the speed larger than permitted
4	Change lane frequently when driving the vehicle
5	Run the red light
6	Driving the vehicle in the speed closed to permitted value
7	Racing with other car for personal grudge
8	Whistling frequently to express the dissatisfied for the low speed of the front vehicle
9	Getting too close to the front vehicle when braking
10	Forced to accelerate by surround vehicles that the speed are more fast

30 different drivers finish the self-reported questionnaire, and then we made the factor analysis and cluster analysis. According to the analysis result, the driving style can be divided into three kind: Sport, normal and gentle. Drivers with sport driving style tend to drive the vehicle as fast as possible, drivers with normal sport driving style tend to drive the vehicle with most common way without initiatively driving fast but easy to be affected by the environment, drivers with gentle method tend to drive the vehicle in most mild method. The gentle drivers do not volunteer to drive fast and is not easily be affected by the environment. After finishing the driving style analysis, we design the driving style recognizer by using the fuzzy logic method. The input of the fuzzy logic recognizer are the ratio that accelerating pedal larger than 50% in the sampling period (that is 10 seconds) and the ratio that decelerating pedal larger than 50% in sampling period (that is also 10 seconds). The output of the fuzzy logic recognizer are the three different driving styles. The fuzzy logic rule table is shown in Table 2.

Table 2

The fuzzy logic rules

	PS(A1)	PM(A1)	PB(A1)
PS(A2)	NORMAL	GENTLE	GENTLE
PM(A2)	NORMAL	SPORT	SPORT
PB(A2)	NORMAL	SPORT	SPORT

In Table 2, A1 and A2 represents the two inputs, respectively; SPORT, NORMAL, GENTLE are three driving styles. PS, PM, PB means the positive small, positive medium, positive big, respectively. The membership function of the fuzzy logic can be witten as: $μ (x) = e^{\frac{(x - c_{i})^{2}}{d_{i}^{2}}}$ (7) where c_i is the function value center, d_i is the function width. The value of c_i and d_i is chosen based on the studied about the ordinary driving habit of 30 drivers who finished self-report questionnaire. The Fuzzy logic deduction results is shown inFig. 2.

Fig.2

Fuzzy logic deduction results.

4 The equivalent consumption minimization strategy

How to distribute power between engine and battery is the main control problem for PHEV. There are many optimized objectives: fuel consumption, emission, drivability, etc. Mathematically, the problem can be formulated as follows with the fuel consumption taken as the optimized goal: $min {J = \sum_{k = 0}^{N - 1} L (x_{k}, u_{k})}$ (8)

Where J is the performance index, L (x_k, u_k) is the cost function, x_k is the state variable, u_k is the control variable. The objective is global, in the sense that the quantity to be minimized is the integral over the whole trip. To ensure the integrity of the control process, some integral constrains should be taken into consideration: $\begin{matrix} {SOC}_{cs_min} & < & {SOC}_{k} < {SOC}_{cs_max} \\ P_{e_min} & < & P_{e_k} < P_{e_max} \end{matrix}$ (9) $P_{m_min} < P_{m_k} < P_{m_max}$ where SOC_{cs_min}, SOC_{cs_max} are minimum and maximum value of battery SOC in the CS stage, P_{e_min}, P_{e_max} are minimum and maximum value of engine power, P_{m_min}, P_{m_max} are minimum and maximum value of motor power, respectively.

The Equivalent Consumption Minimization Strategy (ECMS) is seen as an ideal method which successfully bring the global optimization problem into the real-time application. In this section, the promising ECMS approach is presented. Factor that determines the ECMS application in large degree is discussed, proposing a multi-level ECMS solution to govern the energy flow with reducing the fuel consumption as much as possible.

The basic ECMS

In the ECMS, the performance index J can be expressed as: $J = {\dot{m}}_{ice} (P_{ice} (t)) + ξ (P_{em} (t))$ (10) where ${\dot{m}}_{ice} (P_{ice} (t))$ is the sum of fuel consumption by engine, ξ (P_em (t)) means the equivalent fuel consumption related to the use of motor. Engine and motor related fuel consumption can be written as: ${\dot{m}}_{ice} (P_{ice} (t)) = \frac{P_{ice} (t)}{η_{ice} (t) Q_{lhv}}$ (11) $ξ (P_{em} (t)) = \frac{s (t)}{Q_{lhv}} P_{em} (t)$ (12) where P_ice (t) and P_em (t) is the engine power and motor power, respectively; η_ice (t) is the engine efficiency, Q_lhv is the fuel lower heating value, s(t) is the equivalent factor.

Generally speaking, the principle underlying the ECMS is to assign a cost to electric energy, making the consumed electric energy equivalent to certain quantity of fuel by introducing the equivalent factor. The EMCS denotes that any consumption of electric energy should be charged in the future by the fuel. In real application, the future consuming fuel for charging is assumed to be current equivalent fuel consumption by using the equivalent factor to evaluate the power distribution. Meanwhile, when ECMS is applied in HEV the equivalent factor is set as constant value for the battery SOC change in HEV is quite small. In PHEV, however, the battery SOC can fluctuate significantly with more electric energy can be consumed, leading the equivalent cannot be constant value any longer. Therefore, the equivalent factor is the variable which is related to future energy consumption. To make accurate ECMS application in PHEV energy management, the equivalent factor value should be carefully adjusted.

The adjustment of equivalent factor

With the development of Global Position System (GPS), Geographic Information System (GIS) and sensor technology, Intelligent Transport System (ITS) plays more and more important role in traffic administration. As a result, vehicle’s future driving information can be known in advance, providing the possibility to guide the energy management to be more efficient. According to the different developing stage and the application limit of the upper technology, there are three different levels of acquisition the future information:

Level 1: No future information is available. This level information acquisition commonly happens when vehicle driving in faraway rural road, where map information and other future driving information cannot be access to.

Level 2: Parts future information is available. This level information acquisition commonly happens when vehicle driving in the road where travel distance, future terrain profile can be provided while the future driving velocity cannot by gained due to lack of traffic sensor and blocked communication.

Level 3: Full trip information is available. Thanks to the combined performance of the GPS, GIS, traffic sensor, and the wireless communication, the travel distance, knowledge of future terrain profile and the future velocity can be predicted.

Based on different future driving information acquisition level, the equivalent factor related to future energy consumption can be adjusted, respectively.

No future information

Table 3

The chosen driving cycle

Road condition type	Chosen driving cycle
Traffic jam	Artemis_Urban
	JC08
Urban drive	UDC
	UDDS
Highway	EUDC
	HWFET

Without future information, the equivalent factor cannot be adjusted quite accurately. In this condition, the road condition identification method is applied. Vehicle’s future energy consumption is considered to be connected with different road condition (for example, urban versus highway). Accordingly, it is can be reasonable that assigning a certain equivalent factor value for ECMS application when vehicle in urban driving while valuing another equivalent factor value when vehicle is driving on the highway. The core of this method is the road condition identification. In this paper, the Learning Vector Quantization (LVQ) network is applied to identify different road condition. Normally, an LVQ network consists of the input layer, the hidden layer (also is called the competitive layer), and the output layer (also is called the linear layer). The schematic of the LVQ network can be seen in Fig. 3. The successful road condition identification by LVQ is based on the training LVQ network with reasonable training matrix. To training the LVQ, the standard driving cycle data is used.

Fig.3

Schematic of LVQ.

Road conditions can be classified into three types: traffic jam, urban drive, highway drive. The traffic jam type mainly represents vehicles driving in downtown areas where the traffic situation is poor and requires vehicles to start and stop frequently. The vehicle speed in traffic jam is very low. The urban drive type mainly represents vehicles driving in the city arterial road. In this type, the vehicle speed is normal although there are some crossroads. The highway drive type mainly reflect vehicle driving in highways with a relatively high speed. To form the training matrix, standard driving cycles who can represent the upper three driving conditions are picked up, the characteristic parameters of the driving standard cycle are extracted. After some study, the following driving cycles in Table 1 are chosen to extract the characteristic parameters.

The speed curve of chosen driving cycles are shown in Fig. 4. In this study, 24 characteristic parameters are chosen. When calculating the characteristic parameter values, previous researches divided the driving cycles into segments based on equal time interval or distance interval. These methods may be easy to implement, but the data matrix can be small which can result in the driving condition identifier being less accurate. In this study, the driving cycles are divided into small segments combined with time interval method and distance interval method, which is illustrated in Fig. 5. Considering former researches [28 –34], the time interval is chosen to be one-eighths of total operating time (the red sections in Fig. 5), while the distance interval is chosen to be one-sixth of the total distance (the green sections in Fig. 5). When training the LVQ network, the training data acquired form a [24*6] matrix. In this way, the data used to train LVQ network can be relatively large which is favored in training.

Fig.4

Speed curve of chosen driving cycles.

Fig.5

Illustration of driving cycle dividing method.

The classification process inside the LVQ network mainly includes two stages. First, a certain number of vectors are put in and the hidden layer identifies the input vectors. Secondly, the output layer output the target classes. The identification process in the hidden layer can be described as follows: Neurons (represented by E in Fig. 3) in the hidden layer evaluate the Euclidean distance between the input vector a, and the possible subclass vector p, and finally the closest subclass to the input vector is valued 1, other subclasses 0.

After identifying different road condition, the equivalent factor can be assigned. In different road condition, the equivalent factor values are listed in Table 4.

Table 4

Equivalent factor in different road condition

Road condition type	s _dis	s _cha
Traffic jam	2.59	2.73
Urban	2.5	2.71
Highway	2.45	2.69

In Table 4, s_dis and s_cha is the equivalent factor for discharge and charge, respectively.

Parts future information available

This level of future information acquisition lacks of the instantaneous future velocity. To solve the problem, local history traffic data is employed to estimate the future velocity and travel time. Once the future velocity, travel time, the future required energy can be acquired by the following equations: $E_{m} = \int_{0}^{t} T_{wh} (t) w_{wh} (t) dt$ (13) where t is the travel time, T_wh (t) is the torque required at the wheel, w_wh (t) is the wheel rotational speed. After this, the equivalent factor can be adjusted by: $s (t) = s_{e} (t) + \sqrt{1 - k^{2}} (s_{0} (t) - s_{e} (t))$ (14) $k = min (1, \frac{E_{B_{L}}}{β E_{m}})$ (15) where s_e (t) is the equivalent factor when PHEV in Charge Depleting (CD) stage, s₀ (t) is the equivalent factor in charge sustaining mode in which the remaining All Electric Range (AER) is zero. E_BL is the left stored energy battery can provide. β is the driving behavior factor. To the Sport, Normal and Gentle style, β value is 1.5, 1.2 and 1, respectively. The driving behavior factor is set according to the study about the 30 drivers’ ordinary driving by logging the fuel consumption by OBD (On board diagnostic). s_e (t) and s₀ (t) can be calculated by the following: $s_{e} (t) = \frac{η_{g}}{Q_{lhv}} δ$ (16) $s_{0} (t) = \frac{1}{{\bar{η}}_{eng}} \sqrt{\frac{{\bar{η}}_{c}}{{\bar{η}}_{d}}}$ (17) where η_g is the efficiency of charging the battery pack by the grid; δ is the comparative price that electricity to the fuel, which is 0.05 in this paper; ${\bar{η}}_{c}$ and ${\bar{η}}_{d}$ is the average electric circuit efficiencies for discharge and charge, respectively, respectively; ${\bar{η}}_{eng}$ is the average engine efficiency.

In the level 2, the reasonable adjustment of equivalent factor is determined by the estimation future velocity. In city public traffic system like WisTransPortal [39], the history data like traffic density, traffic flow, average speed of each main roads in one day’s different time is all stored. When we using the history data to obtain the future travel velocity profile, the procedure can be:

Defining the departure time and travel path by the help of GPS and GIS.

Obtaining the average travel speed for each road segment of the whole travel path.

Obtaining travel time in for each road segment based on the acquired average speed and road segment length provided by the GPS and GIS.

Full future information available

In this level, both the history data and instantaneous information can be provided, making it possible to predict the future velocity profile. To predict the travel velocity, the most mature macroscopic traffic flow model—Light-Whitham model is applied. The macroscopic model tries to describe the traffic flow by the macroscopic values like average density, average velocity. The macroscopic model reveals the overall characteristics of the traffic flow which sees the traffic flow as the continuum compressible fluid.

In the Light-Whitham model, the traffic density and traffic average velocity satisfy the continuum fluid mechanics equation: $\frac{\partial ρ}{\partial t} + \frac{\partial (ρ v)}{\partial x} = 0$ (18)

where ρ is the traffic density, v is the traffic average velocity, x is the sampling location. To the average traffic velocity, it is assumed that: $v (x, t) = v_{e} (ρ (x, t))$ (20)

Therefore, the Equation 18 can be rewritten as: $\frac{\partial ρ}{\partial t} + [v_{e} + ρ \frac{\partial v_{e}}{\partial ρ}] \frac{\partial ρ}{\partial x} = 0$ (21)

The Equation 20 reflects the spread of traffic wave. To make the distribution of traffic wave be continuum, the Equation 20 can be expressed as: $\frac{\partial ρ}{\partial t} + v_{e} \frac{\partial ρ}{\partial x} = - ρ \frac{\partial v_{e}}{\partial ρ} \frac{\partial ρ}{\partial x} + D \frac{\partial^{2}}{\partial x^{2}}$ (19)

where D is constant value. Based on the Equation 21, the free flow velocity v_e of the traffic flow can be expressed as: $v_{e} (ρ) = v_{max} (1 - \frac{\partial ρ (x, t)}{ρ_{max}})$ (22) where v_max is the maximum unimpeded velocity, ρ_max is the maximum unimpeded density.

After obtained the instantaneous future velocity prediction, the future required energy can be calculated by using the Equation 13. Similarly, the equivalent factor can be adjusted by Equations 14∼17.

Fig.6

Process of multi-level ECMS application.

Based on the multi-level ECMS, the energy management problem aims at saving fuel in PHEV can be solved more rational, the process of multi-level ECMS application can be illustrated in Fig. 6. In this paper, the control variable u is the power split ration between engine and battery.

5 Simulation and analysis

To prove the effectiveness of the proposed control strategy, Matlab/Simulink simulation is performed. The simulation study can be divided into two parts, one is the simulation analysis about ECMS application with the level 1 equivalent factor adjustment, the other part is about the ECMS application with the level 2 and level 3 equivalent factor adjustment. The parameters of the vehicle is shown in Table 5. In the simulation part 1, the simulation is operated in a specific driving cycle formed by combining UDDS and HWFET cycle. In simulation part 2, the simulation is operated in the driving cycle provided by ITS. As the level 2 and level 3 equivalent factor adjustment method is the same in origin, so the effect of the level 2 equivalent factor adjustment method is simulated in this paper.

Table 5
Parameters of the PHEV

Vehicle mass, 1315 Battery peak 147

M_veh (kg) 1315 power, kW

Drag coefficient, 0.3 Battery total energy kWh 22.6

Frontal area, m² 2 Engine nominal power, kW 45

Rolling resistance 0.0015 Drive motor 80

coefficient, power, kW

Tire radius, m 0.298

Figure 7 shows the speed curve of combined driving cycle, in which 0∼1369 s simulates the urban road condition and 1370 s∼2135 s simulates the highway driving. According to the SOC curve in Fig. 8, the battery SOC changes in CS stage by the ECMS with equivalent factor adjustment would be keep in more stable range than that by the ECMS without equivalent factor adjustment. Figure 9, the engine operating points by the ECMS with equivalent factor adjustment would be more concentrated around the BSFC line than that by the ECMS without equivalent factor adjustment, meaning that the ECMS with equivalent factor can make the engine operate in high efficiency fields as much as possible which is good for saving fuel. Figure 10 illustrates the predicted speed curve which is acquired by the history date stored in the local ITS. Figure 11 illustrates the SOC change by the ECMS with and without equivalent factor adjustment. The SOC with equivalent factor adjustment is kept stable in urban drive while decrease in highway justifying that the battery has to output energy to drive the vehicle together with the engine in CS stage when the requited tractive power is quite high and normally be kept in sustaining stage in CS stage. Figure 12 reveals the engine operating points by the ECMS with and without equivalent factor adjustment. It is clearly that engine operating points are also more concentrated around the BSFC curve, making the engine operate in the high efficiency fields more than that by ECMS without equivalent factor adjustment. Table 6 provides the fuel consumption by ECMS with and without equivalent factor adjustment. According to the simulation results, PHEV with adjustable ECMS would possessed lower fuel consumption than that with un-adjustable ECMS. Moreover, the equivalent factor adjustment effect by the future driving behavior prediction is better than that by road condition identifying. From these results, we can tell that the ECMS with equivalent factor adjustment would bring more reasonable management effect, saving more fuel.

Fig.7

The specific driving cycle in simulation part 1.

Fig.8

The SOC curve in simulation part 1.

Fig.9

The engine operating points in simulation part 1.

Fig.10

The speed curve provide by ITS in simulation part 2.

Fig.11

The SOC curve in simulation part 2.

Fig.12

The engine operating point in simulation part 2.

Table 6

The comparison of fuel consumption

Simulation Part 1
	With adjustment	Without adjustment
Equivalence fuel consumption (L/100 km)	5.31	5.73
Simulation Part 2
	With adjustment	Without adjustment
Equivalence fuel consumption (L/100 km)	5.13	5.73

6 Conclusion

This paper proposes using the ECMS to be the solution of the PHEV energy management problem. And according to different level of information acquisition, the equivalent factor in ECMS can be adjusted accordingly.

The ECMS is seen as an ideal method to deal with the PHEV energy management problem for its instantaneous application feature. In ECMS, the equivalent factor is the core whose value can change frequently. Based on the analysis, the equivalent factor value is related to the future vehicle behavior. Therefore, the equivalent factor should be adjusted accordingly. Thanks to the development of ITS, the acquisition difficulty of the future travel information is decreasing. According to the different level of information, the equivalent factor is adjusted.

The simulation results show that the vehicle with adjustable ECMS holds more ideal performance than that without adjustable equivalent factor.

As an indication for future work, the future information acquisition would be researched deeply, the elements that can influence the equivalent factor value would also be studied.

References

Rahman

, Jurkovic

, Stancu

, et al., Design and performance of electrical propulsion system of extended range electric vehicle (EREV) Chevrolet Voltec[C], Energy Conversion Congress and Exposition (ECCE), 2012 IEEE, 2012, pp. 4152–4159.

Liu

, Wang

, Zhou

, et al., Extended range electric vehicle control strategy design and muti-objective optimization by genetic algorithm[C], Chinese Automation Congress (CAC), 2013 IEEE, 2013, pp. 11–16.

Colzi

and de

, Nigris, A. Corti, et al., Extended range electric vehicles components preliminary sizing based on real mission profiles[C], Electric Vehicle Symposium and Exhibition (EVS27), 2013 World IEEE, 2013, pp. 1–8.

Yang

, Markert

and Heinkel

, Fuzzy logic based energy management algorithm of a hybrid electric vehicle with range-extender[C], Multi-Conference on Systems, Signals & Devices (SSD), 2014 11th International IEEE, 2014, pp. 1–5.

Banvait

, Anwar

and Chen

, A rule-based energy management strategy for plug-in hybrid electric vehicle (PHEV)[C], American Control Conference, 2009 ACC’09 IEEE, 2009, pp. 3938–3943.

Bianchi

, Rolando

, Serrao

, et al., A rule-based strategy for a series/parallel hybrid electric vehicle: An approach based on dynamic programming[C], ASME 2010 Dynamic Systems and Control Conference American Society of Mechanical Engineers, 2010, pp. 507–514.

Anbaran

S.A.

, Idris

N.R.N.

, Jannati

, et al., Rule-based supervisory control of split-parallel hybrid electric vehicle[C], Energy Conversion (CENCON), 2014 IEEE Conference on IEEE, 2014, pp. 7–12.

Khoucha

, Benbouzid

M.E.H.

and Kheloui

, An optimal fuzzy logic power sharing strategy for parallel hybrid electric vehicles[C], Vehicle Power and Propulsion Conference (VPPC), 2010 IEEE, 2010, pp. 1–5.

Yang

, Markert

and Heinkel

, Fuzzy logic based energy management algorithm of a hybrid electric vehicle with range-extender[C], Multi-Conference on Systems, Signals & Devices (SSD), 11th International IEEE, 2014, pp. 1–5.

10.

and Heping

, Fuzzy logic torque control strategy for parallel hybrid electric vehicles[C], Fuzzy Systems and Knowledge Discovery (FSKD), 2011 Eighth International Conference on IEEE, 1, 2011, pp. 640–644.

11.

Wang

, He

, Sun

, et al., Application study on the dynamic programming algorithm for energy management of plug-in hybrid electric vehicles[J], Energies8(4) (2015), 3225–3244.

12.

Yuan

, Teng

, Fengchun

, et al., Comparative study of dynamic programming and Pontryagin’s minimum principle on energy management for a parallel hybrid electric vehicle[J], Energies6(4) (2013), 2305–2318.

13.

Cipek

, Corić, M., Škugor

, et al., Dynamic programming-based optimization of control variables of an extended range electric vehicle[R], SAE Technical Paper, 2013.

14.

Bin

, Reama

, Cela

, et al., On fast dynamic programming for power splitting control of plug-in hybrid electric vehicles[C], ASME 2009 Dynamic Systems and Control Conference American Society of Mechanical Engineers, 2009, pp. 229–236.

15.

Sharma

O.P.

, Onori

and Guezennec

, Analysis of Pontryagin’s Minimum Principle-Based Energy Management Strategy for PHEV Applications[C], ASME 2012 5th Annual Dynamic Systems and Control Conference Joint with the JSME 2012 11th Motion and Vibration Conference American Society of Mechanical Engineers, 2012, pp. 145–150.

16.

Hou

, Ouyang

, Xu

, et al., Approximate Pontryagin’s minimum principle applied to the energy management of plug-in hybrid electric vehicles[J], Applied Energy115 (2014), 174–189.

17.

Yuan

, Teng

, Fengchun

, et al., Comparative study of dynamic programming and Pontryagin’s minimum principle on energy management for a parallel hybrid electric vehicle[J], Energies6(4) (2013), 2305–2318.

18.

Kim

, Cha

and Peng

, Optimal control of hybrid electric vehicles based on Pontryagin’s minimum principle[J], Control Systems Technology, IEEE Transactions on19(5) (2011), 1279–1287.

19.

Serrao

, Onori

and Rizzoni

, A comparative analysis of energy management strategies for hybrid electric vehicles[J], Journal of Dynamic Systems, Measurement, and Control133(3) (2011), 031012.

20.

Chen

, Mi

C.C.

, Xia

, et al., Energy management of power-split plug-in hybrid electric vehicles based on simulated annealing and Pontryagin’s minimum principle[J], Journal of Power Sources272 (2014), 160–168.

21.

Sivertsson

and Eriksson

, Design and evaluation of energy management using map-based ECMS for the PHEV benchmark[J], Oil & Gas Science and Technology–Revue d’IFP Energies Nouvelles70(1) (2015), 195–211.

22.

Schacht

E.J.

, Bezaire

, Cooley

, et al., Addressing drivability in an extended range electric vehicle running an equivalent consumption minimization strategy (ECMS)[R], SAE Technical Paper, 2011.

23.

Skugor

, Ranogajec

and Deur

, On smoothing HEV/EREV supervisory control action using an extended ECMS approach[C], Electric Vehicle Symposium and Exhibition (EVS27), 2013 World IEEE, 2013, pp. 1–10.

24.

Sun

, He

and Sun

, The role of velocity forecasting in adaptive-ECMS for hybrid electric vehicles[J], Energy Procedia75 (2015), 1907–1912.

25.

Musardo

, Rizzoni

, Guezennec

, et al., A-ECMS: An adaptive algorithm for hybrid electric vehicle energy management[J], European Journal of Control.

26.

Tate

E.D.

and Boyd

S.P.

, Finding ultimate limits of performance for hybrid electric vehicles[R], SAE Technical Paper, 2000.

27.

Jeon

, Jo

, Park

, et al., Multi-mode driving control of a parallel hybrid electric vehicle using driving pattern recognition[J], Journal of Dynamic Systems, Measurement, and Control124(1) (2002), 141–149.

28.

Ericsson

, Independent driving pattern factors and their influence on fuel-use and exhaust emission factors[J], Transportation Research Part D: Transport and Environment6(5) (2001), 325–345.

29.

Bata

, Yacoub

, Wang

, et al., Heavy duty testing cycles: Survey and comparison[R], SAE Technical Paper, 1994.

30.

Ericsson

, Variability in urban driving patterns[J], Transportation Research Part D: Transport and Environment5(5) (2000), 337–354.

31.

Crauser

J.P.

, Maurin

and Joumard

, Representative kinematic sequences for the road traffic in France[R], SAE Technical Paper, 1989.

32.

Kuhler

and Karstens

, Improved driving cycle for testing automotive exhaust emissions[R], SAE Technical Paper, 1978.

33.

Carlson

T.R.

and Austin

R.C.

, Development of speed correction cycles[J]. Sierra Research, Inc., Sacramento, CA, Report SR97-04-01, 1997.

34.

Langari

and Won

J.S.

, Intelligent energy management agent for a parallel hybrid vehicle-part I: System architecture and design of the driving situation identification process[J], Vehicular Technology, IEEE Transactions on54(3) (2005), 925–934.

35.

Yang

, Yang

J.H.

, Feron

, et al., Development of a performance-based approach for a rear-end collision warning and avoidance system for automobiles[C], Intelligent Vehicles Symposium, 2003 Proceedings IEEE, 2003, pp. 316–321.

36.

Aberg

and Rimmo

P.A.

, Dimensions of aberrant driver behaviour[J], Ergonomics41(1) (1998), 39–56.

37.

Özkan

, Lajunen

and Summala

, Driver Behaviour Questionnaire: A follow-up study[J], Accident Analysis & Prevention38(2) (2006), 386–395.

38.

Polychronopoulos

, Tsogas

, Amditis

, et al., Dynamic situation and threat assessment for collision warning systems: The EUCLIDE approach[C], Intelligent Vehicles Symposium, 2004 IEEE, 2004, pp. 636–641.

39.

Gong

, Li

and Peng

Z.R.

, Trip based optimal power management of plug-in hybrid electric vehicles using gas-kinetic traffic flow model[C], 2008 American Control Conference IEEE (2008), 3225–3230.

Vehicle mass,	1315	Battery peak	147
M_veh (kg)	1315	power, kW
Drag coefficient,	0.3	Battery total energy kWh	22.6
Frontal area, m²	2	Engine nominal power, kW	45
Rolling resistance	0.0015	Drive motor	80
coefficient,		power, kW
Tire radius, m	0.298

An energy management of plug-in hybrid electric vehicles based on driver behavior and road information

Abstract

Keywords

1 Introduction

2 PHEV model

Table 5 Parameters of the PHEV Vehicle mass, 1315 Battery peak 147 M veh (kg) 1315 power, kW Drag coefficient, 0.3 Battery total energy kWh 22.6 Frontal area, m2 2 Engine nominal power, kW 45 Rolling resistance 0.0015 Drive motor 80 coefficient, power, kW Tire radius, m 0.298

References

Table 5
Parameters of the PHEV

Vehicle mass, 1315 Battery peak 147

M_veh (kg) 1315 power, kW

Drag coefficient, 0.3 Battery total energy kWh 22.6

Frontal area, m² 2 Engine nominal power, kW 45

Rolling resistance 0.0015 Drive motor 80

coefficient, power, kW

Tire radius, m 0.298