Real World Operation of a Complex Plug-in Hybrid Electric Vehicle: Analysis of Its CO2 Emissions and Operating Costs

Plug-in hybrid electric vehicles (pHEVs) could represent the stepping stone to move towards a more sustainable mobility and combine the benefits of electric powertrains with the high range capability of conventional vehicles. Nevertheless, despite the huge potential in terms of CO 2 emissions reduction, the performance of such vehicles has to be deeply investigated in real world driving conditions considering also the CO 2 production related to battery recharge which, on the contrary, is currently only partially considered by the European regulation to foster the diffusion of pHEVs. Therefore, this paper aims to assess, through numerical simulation, the real performance of a test case pHEV, the energy management system (EMS) of which is targeted to the minimization of its overall CO 2 emissions. The paper highlights, at the same time, the relevance of the CO 2 production related to the battery recharge from the power grid. Different technologies mixes used to produce the electricity required for the battery recharge are also taken into account in order to assess the influence of this parameter on the vehicle CO 2 emissions. Finally, since the operating cost still represents the main driver in orienting the customer’s choice, an alternative approach for the EMS, targeted to the minimization of this variable, is also analyzed.


Introduction
In 2010, the United Nations Intergovernmental Panel on Climate Change (IPCC) concluded that a reduction of at least 50% in global CO 2 emissions, compared to 2000 levels, had to be achieved by 2050, in order to limit the long-term rise of the global average temperature due to the greenhouse gas effect [1]. Although this target has been set for all sources of CO 2 emissions, the transportation sector, which is responsible for 33% of carbon dioxide emissions [2], unlike most of the other sectors, has shown an increase in total greenhouse gas emissions, which have been predicted to grow further in the coming years [3], due to the expansion of the global vehicle fleet.
In this framework, electric vehicles (EVs) could pave the way toward a more sustainable mobility since they do not generate pollutant emissions locally and can potentially rely on energy from a selection of renewable sources.
Nevertheless, since the range capability and long recharging time still limit the market penetration of EVs [4], nowadays plug-in hybrid electric vehicles (pHEVs) seem to be the most promising solution to bridge the gap between the desirable features of an electric powertrain and the range of conventional vehicles. In a pHEV, the presence of an additional energy reservoir (i.e., an electric battery) shifts a portion of the emissions burden, during the automobile travel, from on-road fossil fuel combustion to electricity generation from stationary power plants improving the efficiency of the powertrain. Although the European regulation only partially takes into account the CO 2 production related to battery recharge in order to foster the diffusion of such vehicles, the impact of this shift on the overall CO 2 emissions depends on the average efficiency of the internal combustion engine (ICE), and on the amount of the electricity required from the grid. It has been shown from different impact studies (for example, [5]) that emissions caused by electrical power generation, used for charging EV batteries, may vary significantly according to the combination of several factors, among which the period of the day used for charging and the energy demand daily profiles [6,7], the generation portfolio, and the amount of installed -green‖ power. Hence possible further reductions of CO 2 emissions related to the charging from the grid phases strongly depend on factors on which governments have no direct control in a liberalized market framework. A possible way to avoid this uncertainty could be tailor-made contracts between vehicle's owners and energy suppliers, in order to drive the choices of the pHEVs owners towards lower CO 2 emissions recharging patterns.
Moreover, not only CO 2 emissions, but also the operating cost of the vehicle should be taken into account in the definition of the vehicle targets since it is one of the main drivers in orienting the customer's choice and it depends to a greater extent on both fuel and electricity costs.
Although similar approaches have already been applied in literature for relatively simple hybrid architectures [8], the optimization and the full exploitation of the potentials of complex plug-in hybrids such as the one analyzed in this work still represent a quite challenging issue, which is worth to be addressed more in depth. Moreover, the assessment of the CO 2 emissions reduction potential of complex plug-in hybrids along driving patterns representative of real world operation, including significant altitude variations, which have been traditionally representing the Achille's heel of energy management strategies of hybrid powertrains, is also of paramount importance. This paper therefore aims to assess, through numerical simulation, the real performance of a test case pHEV the energy management system (EMS) of which is targeted towards the minimization of its overall CO 2 emissions, and to highlight, at the same time, the main differences with respect to the regulatory European test procedure. Different technology mixes used to produce the electricity required for the battery recharge are also taken into account in order to assess the influence of this parameter on the vehicle CO 2 emissions. Finally, since the operating cost still represents the main driver orienting the customer's choice, an alternative approach for the EMS, targeted to the minimization of this variable, is also analyzed.
After a brief description of the methodology (Section 2), the paper presents the main features of the case study hybrid architecture (Section 3) and the reference performance achieved with a control strategy focused on the minimization of the overall CO 2 emissions (Section 4.1). The main findings of the sensitivity analysis, performed on the technology mixes used to produce electricity to recharge batteries are then presented (Section 4.2). Finally, the main differences between CO 2 minimization and cost minimization strategies are pointed out (Section 4.3).

Methodology
The most noticeable feature of using a hybrid electric vehicle (HEV) is the additional degrees of freedom that can be obtained due to the presence of an additional energy reservoir-the electric battery-besides the fuel tank [9]. Indeed at each instant of time the power requested by the driver can be provided by either one of these sources or by a combination of the two. The choice among all the available powersplit strongly depends on the objective of the specific application: in most cases, it tends to minimize the fuel consumption of the vehicle, but it could also include the minimization of pollutant emissions, the maximization of power delivery, or a compromise among all these goals [10].
In general, the design of the EMS of a hybrid vehicle is an implementation of optimal control [11] and it can usually be addressed through several methodologies which can differ in performance, computational requirements and computational efforts [12,13]. Since the main focus of this paper is not the definition of the EMS, a global optimization algorithm, the dynamic programming (DP) [14,15], has been used to set the ideal performance of the case study hybrid architecture and to highlight the effects of some parameters on vehicle performance. The DP generates a numerical solution for an optimal control problem and it gives sufficient conditions for the global optimality. It is based on Bellman's principle of optimality [11] and is able to manage dynamic models of the system; since DP is commonly used to solve time-continuous control problems, the model has to be discretized in a sequence of time steps for which DP is capable of determining the optimal control laws. Even though the need for a backward procedure means that the solution can be obtained only offline, for a driving cycle known a priori, and therefore it is not implementable on a real vehicle, the optimal control law can be used to gather information for the development of simpler and implementable strategies and to benchmark their performance [16,17].
All the analyses presented in this paper were carried out through numerical simulation performed on a vehicle model developed in Matlab environment. It relies on a kinematic approach [18], based on a backward methodology where the input variables are the speed of the vehicle and the grade angle of the road. From these variables, the powertrain output speed and the traction force that should be provided to the wheels can be easily determined [19]. Both the ICE and the electric machines are represented through performance maps that were experimentally measured under steady state operating conditions. Despite the simple approach, other works proved good agreement [20] in the calculation of the instantaneous fuel consumption over the most common regulatory driving cycles.
The consistent number of control/state variables, coupled with long mission profiles, requires high computational power and leads to huge amount of processing data. Therefore, in order to avoid unacceptable computational time, the high performance computer (HPC) provided by the Department of Control and Computer Engineering of the Politecnico di Torino was exploited [21].

Case Study
The case study selected for this study is a pHEV featuring a complex architecture integrated in a mid-size European passenger car, the main specifications of which are represented in Table 1. As depicted in Figure 1, the vehicle is equipped with a 6 kW•h lithium-ion battery connected to a 60 kW electric motor (EM) which ensure an all-electric range of approximately 30 km in city driving. It also features another electric machine, which mainly works as a generator, and a 2.0-L, four-cylinder engine, paired with a six-gear manual transmission.
The two electric machines are connected to the crankshaft with a fixed gear of appropriate ratio to fully exploit their operating range. A clutch allows disconnecting the ICE and the generator from the EM and from the wheels, so that several operating modes can be selected from the EMS: series mode, parallel mode and all electric modes. Moreover, opening the clutch, the engine is not dragged during regenerative braking or all electric modes, with significant fuel savings. Although most of the powertrain control strategies for HEVs aims to reduce the fuel consumption, such an approach may not be suitable for plug-in architectures since it neglects the energy consumption related to the battery, which is not an energy buffer, like in a charge sustaining HEV [9], but an additional energy source that has to be recharged from the power grid.
Moreover, in order to prove the enhanced emissions performance of HEVs, complete life cycle analysis is also often considered in literatures [22,23]. However, the production process of these new vehicles cannot be easily compared with processes of conventional cars which have been optimized through several decades. Therefore, for this reason and for a fair comparison with the current legislation, in the present work equivalent -tank-to-wheel‖ emissions will be calculated.
A possible way of taking both the electrical and fuel contributions into account, is to minimize the overall CO 2 emissions of the vehicle. As shown in Equation (1), besides considering the CO 2 produced by the engine, a second term related to the battery discharge and to the technology mix used to produce the electricity supplied by the grid is also considered: (1) where J is the cost-to-go function; µ CO2 and µ fuel are the molar mass of CO 2 and fuel, respectively; ṁ f is the instantaneous fuel consumption of the engine; u(t) is the vector of the control variables; T is the duration of the vehicle mission; η chg is the average battery charging efficiency; η grid is the transmission and distribution efficiency of a typical grid; CO 2,pr is the average CO 2 emission related to the production of the electrical energy that is supplied by the grid to recharge the battery; ΔSOC is the variation of the state of charge (SOC) from the beginning to the end of the vehicle mission; and E batt is the total electrical energy that can be stored in the battery.
As far as batteries charging and grid efficiencies were concerned, according to the data reported in literatures [24,25] grid transmission and distribution losses were estimated to be equal to 6% of the generated electrical power, while, for the lithium batteries considered in this work, a charging efficiency equal to 86% was considered [26].
Moreover, the minimization of the cost function should ideally consider the entire life cycle of the vehicle. Nevertheless, in practical cases, the optimization horizon is finite and usually coincides with a short trip: according to several studies [27], about 70% of the daily driving distances in Europe do not exceed 50 km, and could therefore be covered by pure EVs. In this study a reference trip length of about 70 km was considered with the aim of satisfying most of the customers' requirements (about 90% in Europe).
Finally since the real performance of pHEVs could significantly differ from the one recorded through the type approval procedure, the following analysis were performed in different driving conditions considering not only typical regulatory driving cycles (such as New European Driving Cycle (NEDC), Worldwide harmonized Light vehicles Test Procedures (WLTP) [28], US06, etc.), but also taking into account two real world driving schedules, experimentally measured, as representative of typical urban driving conditions (Aachen cycle- Figure 2a) and of extra-urban driving conditions with significant altitude variations (Arco-Merano cycle- Figure 2b) [29,30].

CO 2 Minimization: Reference Case
The aim of this analysis was to define the optimal control strategy which minimizes the total CO 2 produced by the vehicle, for several driving schedules, which were obtained through the repetition of the considered driving cycles until the target trip distance of about 70 km is reached. These tests were performed in charge depleting mode, starting from a battery SOC of 66%. This value was arbitrarily chosen in place of 100% only to avoid excessive simulation time to reach the minimum SOC of about 20%.
The simulations were performed assuming a value of 326 g/kW•h for the CO 2 , pr (Equation (1)) as representative of the average for the European countries belonging to the Organization for Economic Cooperation and Development (OECD) for the year of 2009 [31]. Taking into account the η grid and η chg values previously specified, an overall CO 2 production rate of about 400 g/kW•h is obtained. Figure 3, where the operating mode selected by the DP is represented, shows that, despite of the relatively low energy capacity of the battery, the vehicle mainly operates in EV mode and the thermal engine only supports the propulsion during high demanding phases (i.e., vehicle accelerations or the extra-urban drive). Furthermore the DP does not exploit the series mode because of its low efficiency, which is the result of multiple energy conversions between the ICE and the electric machines. On the contrary the parallel mode represents the best way to transmit power to the wheels. Moreover, the analysis of the powersplit represented in Figure 4 points out the limited contribution of the generator due to the efficiency maps of the electric machines and to the kinematic constraints imposed by the architecture. Indeed the DP uses the EM for regenerative braking and as an electric booster, while the generator (GEN) assists the EM for the electrical generation when the ICE is on (Figure 4). Globally such a control law mainly intends to keep the engine on its best efficiency region ( Figure 5a) and to achieve a linear discharge of the battery (Figure 5b) which represents the best exploitation of the energy stored in the battery as already highlighted by scientific literature for different hybrid architectures [32,33]. As a matter of fact, the CO 2 specific emission of the engine (about 650 g/kW•h on average) is significantly higher in comparison with the average CO 2 production rate CO 2,pr of industrialized countries (Section 2); therefore the DP exploits the battery as much as possible, trying to reach the minimum SOC at the end of the trip. The same behavior can be observed on the Aachen cycle (Figure 6a), which represents typical real world urban driving conditions, but not on the Arco-Merano cycle (Figure 6b). Its particular altitude pattern, with two strong downhills, leads to a SOC evolution which is no more linear with the travelled distance. In this case the DP prefers to collect all the CO 2 saving coming from the usage of the battery during the first half of the cycle (when the powertrain provides power to the wheels) and then to recover the kinematic energy of the vehicle during the downhill. Moreover, it is eye-catching that, at the end of the driving cycle, the battery is no more fully depleted. Since a strong downhill occurs just before the end of the trip, the only chance to reach the lower limit of the SOC at the end of the mission would have been to push the discharge of the battery below the lower SOC limit before the last downhill, but, since the violation of the minimum SOC limit is prohibited, the best the DP can do is to reach the minimum SOC immediately before the last downhill.
The CO 2 emissions results obtained over the repetition of the selected driving cycles are shown in Table 2: it can be noticed that the full exploitation of the hybrid powertrain potential in terms of CO 2 emissions reduction achieved by the DP is quite impressive, since, even considering real world driving cycles, and taking into account the contribution of the battery recharge, remarkably low CO 2 emissions are reached, all below 100 g/km. From the analysis of Table 2, it is also worth pointing out that, if the contribution of the battery recharge phases is taken into account as in the proposed methodology, the overall CO 2 emissions of the proposed pHEV are significantly higher than the values calculated through the European regulation for plug-in hybrid vehicles, which only partially takes into account the CO 2 produced by the engine. As a matter of fact the European Union (EU) procedure requires two tests [34]: • Condition A: which is carried out with a fully charged electrical energy storage device; • Condition B: which is carried out with an electrical energy storage device in the minimum SOC (maximum discharge of battery capacity).
Consequently the actual CO 2 emissions of the vehicle are represented by the weighted average of the data recorded in the previous tests and the weights are the vehicle electric range and the average distance between two battery recharges (Equation (2)): (2) where M 1 and M 2 (g/km) are the CO 2 emissions recorded in Conditions A and B, respectively; D e is the electric range of the vehicle (the distance covered in EV mode on the considered cycle or its multiple); and D av is the average distance between two battery recharges (which is assumed to be about 25 km).
As depicted in Table 3, applying the above-mentioned procedure to the tested pHEV, the benefits of the hybridization are emphasized. As a matter of fact, during Condition A, when the battery is fully charged, there is no CO 2 production since the DP only exploits the EV mode (only on the WLTP cycle the capacity of the battery is not enough to perform the entire test, so a small amount of emissions is recorded). On the contrary, if the CO 2 emissions due to the battery recharge phases are taken into account as in the proposed methodology, the pHEV CO 2 emissions result to be significantly higher, although still remarkably lower than those of a conventional vehicle.

Energy Mix Influence: A Parametric Analysis of CO 2 Rate
As it was already pointed out in the previous sections, different technology mixes to produce the electricity required for the battery recharge may hugely affect the decisions taken by the DP since they change the cost of the electric energy (Equation (1)).
Statistics data collected by the International Energy Agency (IEA) revealed that the CO 2 specific emissions related to electricity production depend to a great extent on the considered geographical region [31]: the values can vary from 90 g/kW•h (e.g., for France, where a significant fraction of the electrical energy is produced through nuclear power plants), up to about 1000 g/kW•h in emerging countries (Table 4). Therefore in order to emphasize the effects of this parameter on the control law defined by the DP and on the vehicle performance, a sensitivity analysis was performed in the range from 75 g/kW•h to 1000 g/kW•h. Despite some works in literatures (for instance, [1]) report that using the average CO 2 intensity of electricity generation to estimate the emissions caused by charging EVs could lead to imprecise results, the range of emissions here analyzed is so wide to bypass this issue. Table 4. CO 2 production rate of the main countries in recent years. US: United States; and EU: European Union.  World  495  500  500  503  508  504  500  504  US  571  571  570  542  549  535  508  531  Japan  444  427  429  418  452  438  415  435  France  81  79  93  87  90  87  90  89  Germany  434  436  406  404  468  441  430  447  Italy  511  459  449  468  440  421  386  416  The United Kingdom  478  486  485  507  499  490  450  480  EU 27  374  366  358  362  373  355  339  356  China  776  804  787  787  758  744  743  748  India  892  931  923  921  943  954  951  950 The simulation results showed that, as expected, the higher the CO 2 production rate is, the higher the energy produced by the ICE will be ( Table 5). Nevertheless significant changes in the control strategy can be appreciated only if the CO 2,pr is about 600 g/kW•h (the energy provided by the ICE figures in Table 5), as shown in Figure 7. Below this threshold the CO 2 production rate related to the battery recharge is lower than the average specific emission of the engine (about 650 g/kW •h), so the DP exploits the energy stored in the battery, while above 600 g/kW•h the use of the engine is more convenient.  Figure 7. SOC trends obtained by changing the CO 2 production rate.
Globally if the CO 2 production rate is higher, also the overall CO 2 emissions increase since the reduction of the battery contribution is not able to balance the increase of the engine usage.

Customer Perspective: The Operating Cost Approach
Although nowadays customers are more and more aware of environmental issues, from their perspective HEVs primarily represent a way to reduce fuel expenditure. Therefore, the performance of the EMS, described in the previous sections, will be compared with an alternative methodology aiming to minimize the energy cost (Equation (3)): (3) As for the CO 2 emissions, the cost function is composed of two parts: the first is related to the fuel cost, while the second is related to the electrical energy cost. C fuel and C elec represent fuel and electricity costs, respectively, and they were assumed to be equal to the 2011 European average [35,36].
In order to highlight possible differences in the effects of this new EMS, a comparison between the operating modes chosen in this case with the operating modes of the reference case (targeted to the CO 2 minimization) can be analyzed as shown for instance in Figure 8. The analysis of Figure 8 does not reveal any significant differences in the control law defined by the DP if the optimization is targeted to the minimization of the energy cost instead of the CO 2 emissions. As a result, these approaches almost attain the same performance, as depicted in Table 6. This lack of difference is a consequence of the huge difference between the specific cost of the fuel (0.42 €/kW•h) and the specific cost of the electricity (0.184 €/kW•h), as in the CO 2 optimization. Hence, since the ratio between these costs (2.27) is quite close to the ratio between the specific CO 2 emissions of the engine and the CO 2 production rate (1.6), also the path defined by the DP to minimize the cost function is almost the same.
Therefore, it can be stated that an EMS aiming to minimize the CO 2 emissions of the vehicle will also achieve the minimum of the energy expenditure. Nevertheless, since these results could significantly change depending on the future trends of fuel and electricity costs, additional scenarios for these parameters were analyzed in order to point out their effects on the powertrain control strategy. Two parameter sweeps, for electricity and fuel costs respectively, were performed keeping fixed all the other parameter values. The fuel cost was varied in the range 1.25-2 €/L while the electricity cost in the range 0.1-0.25 €/kW•h.
The comparison of the SOC profiles, represented in Figure 9, points out that the control law defined by the DP is not affected by these sensitivity sweeps since, in each scenario, the cost of the fuel is significantly higher than the electricity cost in each considered case. The battery discharge is always the preferred option of the EMS, even if the total cost of the trip will increase.

Conclusions
This work analyzed, through a global optimization algorithm, the real benefit of a plug-in hybrid architecture integrated into a mid-size European passenger car. The optimal powertrain control strategy was primarily obtained in different driving conditions with the goal of minimizing the overall CO 2 emissions of the vehicle. The analysis of the DP results showed that a linear discharge of the battery over the travelled distance represents the optimal strategy only for vehicle missions without altitude variations. The relevance of the CO 2 related to battery recharge from the grid was also highlighted comparing the real emission of the vehicle with the values attained through the European regulation which neglects the contribution of the battery recharge.
Furthermore, the proposed methodology based on the minimization of the CO 2 emissions was also proved to be able to minimize the energy cost which is the most important parameter from the customer perspective. In both cases, the use of the engine was in fact shown to be significantly more expensive than the battery discharge. This finding was also strengthened by the sensitivity analysis on the CO 2 production rate, which showed negligible effects on the control law defined by DP when typical data from industrialized countries such as the US, EU and Japan, were used (appreciable effects could only be observed for very high CO 2 production rates, which are typical of China and India). architecture and performed all the simulations. Federico Millo was the project supervisor and played a key role in the analysis of the simulation results.