Conversion of a Small-Size Passenger Car to Hydrogen Fueling: Evaluation of Boosting Potential and Peak Performance During Lean Operation

Abstract

Energy and mobility are currently powered by conventional fuels, and for the specific case of spark ignition (SI) engines, gasoline is dominant. Converting these power-units to hydrogen is an efficient and cost-effective choice for achieving zero-carbon emissions. The use of this alternative fuel can be combined with a circular-economy approach that gives new life to the existing fleet of engines and minimizes the need for added components. In this context, the current work scrutinizes specific aspects of converting a small-size passenger car to hydrogen fueling. The approach combined measurements performed with gasoline and predictive 0D/1D models for correctly including fuel chemistry effects; the experimental data were used for calibration purposes. One particular aspect of H₂ is that it results in lower volumetric efficiency compared to gasoline, and therefore boosting requirements can feature significant changes. The results of the 0D/1D simulations show that one of the main conclusions is that only stoichiometric operation would ensure the reference peak power level; lean fueling featured relative air–fuel ratios too low for ensuring the minimum value of 2 that would allow mitigating NO_x formation. Top speed could be instead feasible in lean conditions, with the same gearbox, but with an extension of the engine speed operating range to 7000 rpm compared to the 3700 rpm reference point with gasoline.

1. Introduction

Without attempting to provide an exhaustive overview of the future evolution of the energy and mobility sectors, it suffices to say that hydrogen is bound to play an essential role [1], even in a scenario oriented towards electric vehicles [2]. The starting point for the idea behind this article was to evaluate the conversion of small-size passenger cars from gasoline to hydrogen, and how this could be achieved while maintaining reference performance levels, essential for preserving intended vehicle dynamics.

Combustion properties of hydrogen are an important aspect to be considered, especially as they are quite different compared to conventional fuels [3]. One direct consequence of these properties is that high laminar flame speed results in improved stability [4] and ensures the possibility of lean operation with good potential for improved efficiency [5]; exhaust gas recirculation can also be implemented with high rates [6,7] and results in fuel economy benefits, as well as pressure rise rates closer to those for gasoline [8]. Other issues such as heat loss [9] or how modeling is approached [10], as well as NO_x emissions [11], are to be considered as specific issues correlated with the use of hydrogen. Several of these aspects were successfully handled in the form of a market-ready solution [12], proving that efficient zero-carbon mobility is possible.

One essential aspect that emerges is that volumetric efficiency plays a crucial role [13] and needs to be carefully considered, especially when undertaking the conversion of a gasoline-powered vehicle. Following this line of research, previous overall analysis of full load performance revealed that low end torque is the most affected rpm interval in terms of power loss [14]. Alternative propulsion techniques such as mild hybrid configurations could be used for mitigating this loss of power [15]. These aspects highlight the complexity of converting a small-size passenger car to H₂, even when considering only one effect such as volumetric efficiency.

The current paper uses a 0D/1D engine (port-fuel injection PFI, turbocharged, gasoline powered 40 kW unit) model for a small-size passenger car that was updated compared to previous work. Several sub-models were developed or adapted for closer correlation to real-world components and control strategies. This was intended to scrutinize the true boosting potential of the power unit and ensure results that give a better idea into the modifications required, with a focus on electronic control unit (ECU) remapping. The potential for lean fueling was also analyzed in direct correlation with reference peak power level, as well as the torque required for ensuring vehicle top speed.

2. Materials and Methods

Measurements were performed on a test-bench that featured a 40 kW, 3-cylinder, turbocharged PFI SI engine fueled with gasoline; this unit powers a small-size passenger car with 720 kg weight and 135 km/h top speed. Limited datasets could be recorded, given that the official equipment manufacturer (OEM) ECU quickly defaulted; this was due to the fact that the automatic gearbox control is integrated with the engine ECU and continuous reporting of errors for the transmission part resulted in critical faults. Nonetheless, the recorded data proved instrumental for model calibration.

Detailed description of the engine characteristics can be found in [8] (Figure 1 shows the schematic overview of the test-bench;

Table 1. Engine specifications.

Displacement	599 cm3
Number of cylinders	3
Rated power	40 kW @ 5250 rpm
Rated torque	80 Nm @ 2000–4400 rpm
Bore × Stroke	63.5 mm × 63.0 mm
Connecting rod length	114 mm
Compression ratio	9.5:1
Number of valves	2 per cylinder
Intake valves opening/closure	363/164 deg bTDC
Exhaust valves opening/closure	157/349 deg a/bTDC
Fuel system	port fuel injection (PFI) at 3.5 bar for gasoline and 5 bar for gas H2
Ignition	inductive discharge, 2 spark plugs per cylinder

contains the main specifications). Minimal modifications were performed to the engine, such as substituting the throttle pedal with a voltage module or fitting a spark plug pressure sensor by replacing one of the OEM plugs on cylinder 1. Several ECU parameters were recorded through an on-board diagnostic (OBD) interface. Figure 2 shows three of these parameters recorded at 2500 rpm; a sort of an over-run of the pedal signal was possible by augmenting the voltage level beyond those that are found with the OEM pedal. These measurements were performed with and without the waste-gate (WG) solenoid valve control (i.e., normal operation vs. disconnected solenoid valve) and were aimed at identifying the effect of the pneumatic actuation and related intervention of the ECU.

In-cylinder pressure measurements were performed with an accuracy of ±1% and resolution of 0.5 deg crank angle by using the piezo-electric sensor integrated in the spark plug; referencing was implemented by using the OEM manifold absolute pressure sensor. Figure 3 shows the in-cylinder pressure trace obtained at 2500 rpm and 0.9 bar intake pressure, with stoichiometric gasoline fueling.

Simulations were performed with a 0D/1D model (shown in Figure 4, only the engine part of the powertrain was modeled) built with the specific purpose of including fuel chemistry effects; i.e., the predictive version of the combustion sub-model was employed [16], with a dedicated laminar flame speed correlation implemented for hydrogen [8]. Conversion to H₂ was hypothesized in the same conditions as in [17] in terms of injector flow and fuel delivery settings so as to minimize the possibility of backfiring, and they are in line with experimental work on H₂ PFI found in the literature [18]. Without going into details such as the basic principle that specific components of the engine are modeled as volumes with simplified geometry [16], only the specific modifications that were implemented for this work will be described in detail.

In light of obtaining results that are closer related to the ECU remapping goal, a correlation between throttle pedal position, throttle valve opening and target intake pressure was established throughout the rpm-load map. In its previous versions, the model contained a simple approach in the form of a condition that if throttled operation is needed, no boost is applied (i.e., the WG is left completely open); once the load level requires boosting, the throttle would be kept wide open and the WG controller would progressively close the exhaust gas bypass to augment intake pressure. This is actually not the case in real-world applications: more to the point, most applications feature a partially closed throttle even when boosting is required (it should be noted that the pressure drop across the throttle valve is minimal in turbo-charged conditions). Basically, this means that there is a map of target intake pressure based on throttle pedal position (which the driver uses for setting torque demand) and engine speed. Evidently, remapping the ECU for using hydrogen means that a new map needs to be developed; this part will be detailed in the ensuing section.

Turbocharger control was another particular aspect directly related to this work. More to the point, given the main goal of evaluating boosting potential, the compressor map was updated with experimental data found in the literature [19] so as to ensure the best possible accuracy close to the surge line. Rather than building a dedicated sub-model (e.g., [20]), a simplified approach was implemented. Data were recorded at 2500 rpm, with different throttle pedal position settings and gasoline fueling. The command signal that controlled the WG solenoid valve was acquired by connecting an opto-isolator in parallel; this allowed the evaluation of the dwell percentage of the pulse width modulation (PWM) signal (data shown with round symbols in Figure 2). The measurements showed that without the solenoid valve, the mechanical actuation of the WG keeps a value of 0.56 bar pressure difference between down-stream and up-stream of the compressor. Assuming that the WG valve is completely closed at 0 or negative pressure difference, a linear response can be modeled for each rpm setting (Figure 5).

As expected, the WG is opened more at higher engine speeds for keeping the constant 0.56 bar pressure difference up-stream vs. down-stream of the compressor, while at lower rpm (and consequently reduced air flow), the WG is opened less. The end result in terms of actual application in the 0D/1D model is an rpm-pressure difference look-up table that substitutes the pneumatic actuation of the WG valve. The boost PID controller acts on the output value that comes from the WG-mech component (i.e., the WG diameter obtained from the look-up table based on rpm and pressure difference). An additional feature is that a condition was implemented in the form of a lower limit for the PID controller that “forces” the WG open if surge is detected (i.e., if the surge margin RLT parameter is below 0).

The most complex sub-model was the combustion part; the predictive version contained in the code was used, with a custom laminar flame speed implemented for hydrogen. Calibration was performed for one operating point with gasoline fueling (the measured and simulated in-cylinder pressure traces are shown in Figure 3). It was decided to act only on the Taylor length scale (TLS) multiplier; all the other calibration parameters were left at their default value. Very good accuracy was obtained for a TLS multiplier value of 2.4. For hydrogen simulations, the calibrated model was used with the custom laminar flame speed correlation [21]; the main hypothesis was that the fuel chemistry part was correctly included, especially as applications of the aforementioned correlation ensured accurate results in the AFR_rel range of 1–2 [22]. Including corrections for specific effects of H₂ flame instabilities [23,24] could improve accuracy in lean conditions at high load, but for the scope of the current work, it was decided to keep the approach compatible with the used correlation, especially as the focus is on boosting characteristics.

Another modification compared to previous versions of the model was to introduce a peak pressure limitation in the component that controls ignition settings. In addition to the knock limited spark advance (KLSA) module, a peak pressure limited spark advance (PLSA) routine was implemented. A threshold of 95 bar was chosen as the upper limit for peak pressure. This value was taken as representative, given that the full load simulations predicted maximum pressure slightly over 93 bar throughout the rpm range; it was also found to be compatible with limits available in the literature [25,26].

3. Results and Discussion

Once the model was calibrated, a detailed analysis could be implemented in terms of evaluating operating conditions, with the focus on boosting potential. Two main lines of scrutiny were outlined based on air–fuel ratio settings: stoichiometric and lean fueling.

3.1. Stoichiometric Operation and Turbocharger Control Implications

One major issue with H₂ fueling is that volumetric efficiency is significantly lower compared to gasoline. For this reason, engines converted to run on hydrogen are de-rated or need relatively high boosting levels [27]. In this initial phase, only stoichiometric (i.e., relative air–fuel ratio AFR_rel or lambda value of 1) operation was considered, so as to ensure the highest possible power density. This may induce several issues, with abnormal combustion being a key factor [28,29], but overall running an H₂ engine with an AFR_rel of 1 is feasible, even with an increased compression ratio [30].

Given the essential change in boosting requirements, one of the first aspects that was under scrutiny was the throttle pedal position–target intake pressure correlation. Figure 6 shows the related pedal signal voltage and its position, as well as the target intake pressure for the two fuels (the data for gasoline were recorded, while that for hydrogen are the results of simulations). The modified correlation is essential for ensuring the same vehicle dynamics that the driver expects, i.e., the throttle pedal position directly defines the torque demand. One interesting observation is that the OEM pedal signal reaches a maximum of only 72%, a percentage read by the ECU at 2.4 V voltage input; by “boosting” the voltage above this threshold, the intake pressure target increased from 1.57 bar_abs to 1.88 bar_abs at 100% pedal. So even with a simple change like manipulating the throttle pedal signal, a gain of close to 20% in intake pressure can be obtained. Hydrogen fueling requires around a 30% increase (from around 1.5 to slightly over 2.0 bar_abs) in intake pressure, and so, more extensive modifications in terms of remapping are required compared to gasoline.

Taking the aforementioned target intake pressure of over 2 bar calculated for H₂ as reference, simulations were performed with gasoline at higher full load output (i.e., 52 kW peak power compared to the 40 kW rating; Figure 7 shows the full load brake power traces). A value of 52 kW peak power target is completely compatible with the engine (practically the same hardware can deliver this level of output with minimal modifications in terms of control settings [31]); for achieving the 30% increase in brake torque between 2500 and 5500 rpm, the results indicate an intake pressure in the range of 1.92–2.14 bar, compared to 1.48–1.52 bar for the 40 kW version.

For attaining the 40 kW reference power at 5250 rpm, hydrogen fueling requires an intake pressure range similar to the aforementioned 1.9–2.1 bar. This can be seen as an initial confirmation that the required boosting levels can be achieved without hardware modifications. Therefore, one of the main conclusions was that peak power can be ensured with stoichiometric hydrogen fueling at the same level as the gasoline reference; this result features increased confidence compared to previous evaluations [14], given that more details were included in the dedicated sub-models. Only the engine speed range below 3000 rpm featured significant power loss, due to compressor surge limitations.

This issue is better illustrated in Figure 8. Basically, it shows that PFI hydrogen fueling shifts the operating conditions into less favorable regions of the compressor map, with lower air flow and a higher pressure ratio (thus closer to the surge limit). Turbo-matching would improve boosting capabilities (e.g., fitting a “smaller” turbocharger would reduce surge limit effects and most likely improve response in the low rpm region), but given the focus of converting an existing power unit, this approach was not considered. Even if intake pressure is lower compared to the levels found with gasoline fueling and 52 kW target power, there is a significant difference in terms of air flow. The main reason for this is that hydrogen displaces air when injected (around 30% at AFR_rel 1 compared to less than 2% for gasoline); even though for the same energy input hydrogen requires less air for complete oxidation (around 17% less), the reduction in volumetric efficiency is still more prominent. One positive aspect is that the simulation results suggest that turbine inlet temperature would increase by around 50 °C when using hydrogen (from 850 to 900 °C), completely comparable with the increase predicted when considering the 52 kW peak power situation.

Another aspect that was considered was the possibility of failure of the WG solenoid valve. As shown in Figure 9, the effects are minimal for gasoline; the slightly augmented throttle pedal opening completely compensates for the missing intervention of the ECU (more to the point, if the solenoid valve fails, only the mechanical–pneumatic control of the WG is present). For hydrogen instead, intake pressure is far from the target that would ensure intended full load engine output. This is due to the fact that WG regulation is governed only by the pressure difference and the mechanical action of the spring calibrated for maintaining a compressor upstream–downstream pressure difference of 0.56 bar. Even if the throttle pedal is at 100% throughout the rpm range, the lack of an additional control margin provided by the solenoid valve results in much lower intake pressure compared to the target.

3.2. Lean Operation Potential

Once the simulations confirmed the potential of attaining 40 kW peak power with H₂ stoichiometric fueling and complete functionality, the analysis was extended to lean operation. The idea was to increase the relative air–fuel ratio up to the value of 2.2, at which NO_x emissions are low enough to ensure compliance with regulations, without the need for dedicated exhaust gas after-treatment [32]. It should be noted that the vehicle considered for the simulations is equipped with a three-way catalytic converter that should ensure correct conversion of NO_x during stoichiometric fueling. In a previous evaluation, exhaust gas recirculation (EGR) was found to be limited by boosting potential to levels below 30% [8]. Indeed, roughly the same limit is suggested by the results shown in Figure 10; more to the point, at lambda values over 1.4, the peak power starts to drop. Another possible issue is that the boost pressure needs to be augmented over 2.5 bar_abs, which is more than 25% compared to the value predicted for the gasoline 52 kW version; this may go too close to the physical limits of the intake system. An interesting observation is that the efficiency rating shows a non-monotonic trend, meaning that it tends to increase at a higher rate between AFR_rel 1.4 and 1.6 compared to the interval between 1.2 and 1.4. This is directly linked with the fact that the peak pressure limitation resulted in retarded spark timing compared to the optimal setting, and thus caused reduced benefits in terms of efficiency. Just to put things into perspective, at peak power, spark timing was predicted at 16 deg bTDC for gasoline, while for H₂, this parameter was at 2.5 deg aTDC with stoichiometric fueling and at 6.8 deg bTDC for lambda 2; peak pressure limitations resulted in retarded CA50 by over 5 deg for e.g., AFR_rel 1.4. It should be noted that the PLSA aspect was not considered in the previous study on EGR potential [8], and further emphasizes the complexity of converting a gasoline powered vehicle to H₂ fueling.

Simulated knocking probability (in the form of the knock time integral) features a peak for AFR_rel 1.4, therefore suggesting more probable abnormal combustion phenomena compared to stoichiometric fueling. The pressure rise rate was predicted to be less of a concern, with a continuous decrease as the mixture was leaned out; levels close to those calculated for gasoline were obtained around lambda 1.7. As an overall conclusion, the numerical results suggest that mixture leaning is less favorable compared to, e.g., stoichiometric operation or EGR; this is due to the fact that the maximum AFR that ensures the intended level of peak power is too low for successfully reducing NO_x emissions, and may even result in worsened knock propensity. Gains in efficiency were also less evident than expected, given the peak pressure limitation. It should be noted that simulations throughout the engine speed–load map (not included here for the sake of brevity) suggest a limit of 40–50% of full load torque request (depending on engine speed, with a higher percentage at increased rpm) when operating the engine at lambda 2; interestingly enough, in this low load condition, boosting levels were quite low, with peak values close to those calculated at full load for gasoline (i.e., below 1.5 bar_abs intake pressure). A positive aspect is that the simulations predicted a significant increase in brake efficiency, even over 25% compared to stoichiometric fueling. This effect was most prominent at low rpm; more contained gains of around 10–15% were noted at higher rpm. Overall, the results suggest that leveraging a mixed operating strategy is most interesting, with lean operation at low torque request and stoichiometric fueling at high load.

An even more detailed analysis, directly related to the application, was performed for evaluating the boosting potential at the top speed of the considered vehicle. The transmission features overall reduction ratios of 14.2, 10.3, 7.4, 5.6, 4.1, and 2.9, for the 1st, 2nd, 3rd, 4th, 5th, and 6th gear, respectively; combined with 15-inch rims and 175/55 tires, the 135 km/h top speed rating can be reached at 3700, 5100, and 7000 rpm in 6th, 5th, or 4th gear. The required engine power value was calculated at 27 kW by considering a frontal area of 1.9 m², C_x 0.37, and 90% transmission efficiency; evidently, lower torque demand would be set at higher engine rpm. Indeed, the results shown in Figure 11 confirm that higher AFR_rel can be employed at top speed if selecting lower gears. Unfortunately, the only setting that ensures the required brake power at a high enough air–fuel ratio (i.e., lambda 2 for low NO_x emissions) is that with the 4th gear selected and 7000 rpm engine speed. Apart from NVH issues, this operating point is right at the limit of the acceptable mean piston speed (i.e., 14.7 m/s compared to a maximum of 15 m/s [33]). Furthermore, predicted efficiency was around 10% lower compared to stoichiometric operation at 3700 rpm. One positive aspect is that the model predicted that NO_x mitigation is feasible with lean fueling at lambda 2. More to the point, simulated raw emissions were 22 g/kWh for gasoline at top speed, while for H₂, at 3700 rpm with stoichiometric fueling, NO_x emissions were 28 g/kWh and at 7000 rpm with an AFR_rel of 2, the value was 0.2 g/kWh; if considering a catalytic conversion efficiency of 95% [33], this suggests significant margins in terms of emissions compliance with lean fueling. These results further highlight the benefits of approaching engine control with the mixed AFR target mentioned earlier, rendered possible by the specific properties of hydrogen.

4. Conclusions

The overall conclusion of the analysis reveals that the true boosting potential limits actual applicability of lean fueling to low load in the 0–50% range of torque demand. This requires careful, multi-objective evaluation of ECU remapping when converting a gasoline powered vehicle to H₂. Other specific conclusions can be listed as follows:

• Intake pressure target values need to be augmented by around 30% for stoichiometric H₂ fueling compared to gasoline, and boosting margins should be enough for most power units that equip passenger cars; another consequence is that the related pedal position–intake pressure correlation needs to be specifically adapted for hydrogen;
• The specific properties of hydrogen that result in reduced volumetric efficiency tend to shift turbocharger operation in a less favorable part of the compressor map, with lower air flow and higher pressure ratios; this shift moves the entire operating range closer to the surge line and causes significant power loss of even up to 40%, especially at low engine rpm; re-engineering the turbo-matching part is most likely to result in minor gains and diminishes the benefits of cost competitiveness of converting gasoline powered vehicles;
• At full load, lean fueling cannot be ensured at high enough air–fuel ratio to mitigate NO_x emissions (e.g., reference peak power cannot be ensured for lambda values over 1.4) and can even result in unexpected effects in terms of knock propensity; peak pressure limitations can also hinder obtaining maximum gains in terms of efficiency when leaning the mixture;
• Vehicle top speed could be ensured with lean fueling when combined with updated gearshift control, but with around 10% lower efficiency compared to stoichiometric operation.

All these aspects highlight the specific needs to be considered when undertaking conversion from gasoline to hydrogen fueling. Changes in volumetric efficiency play a determining role, and choosing a mixed lean–stoichiometric approach in terms of target air–fuel ratio is the strategy most likely to maximize gains in efficiency while maintaining peak power values at the level that can be obtained with gasoline.