Decarbonizing energy generation is today a main issue, as related to the ONU Agenda 2030 as well as to fulfill the commitments assumed at a global level regarding the release of greenhouse gases (GHGs) within tolerable limits, so as to mitigate negative effects of climate changes. Currently considered measures to reduce the environmental impact deriving from the exploitation of fossil fuels in reciprocating internal combustion engines (ICEs), for both energy direct use and transportation purposes, include various alternatives.

Natural gas’ (NG) chemical properties and its easy use due to diffused accessibility to national distribution networks and/or to storage facilities make this fuel a largely used option for steady cogeneration purposes (Sandalci et al., 2019). NG is mainly composed by methane (CH₄) and has a low carbon-to-hydrogen (C/H) ratio, so as to provide relatively low specific CO and CO₂ emissions deriving from its combustion in spark ignition (SI) engines, combined with an intrinsically high knocking resistance that allows increasing the ICE compression ratio with consequent better overall efficiency (Çeper, 2012).

Issues related to the environmental impact of fossil-fuel–powered energy systems have recently brought increased attention towards the use of hydrogen (H₂), due to the intrinsic possibility of this energy carrier to produce only water vapor from its combustion. This fuel is a valuable candidate for energy storage (Liu et al., 2020), especially for the so-called seasonal storage, to exploit possible electricity surpluses during peak power production periods to efficiently generate energy during shortage ones (Uchman et al., 2020) and to also avoid selling of electricity surpluses during low market price intervals. The problem, indeed, regards today many already installed cogeneration plants, especially in Italy, having up to the recent past profited of incentives to power generation and delivery to the national grid, but that, due to a massive energy production, currently undergo economic losses due to a scenario of overall lower energy selling prices.

In specific, the conversion of surplus of electric energy into chemical energy is the core element of the so-called power-to-gas concept and is performed through the process of water electrolysis, where electrical energy by only renewables can be used to derive both hydrogen and oxygen from water (Schiebahn et al., 2015). The hydrogen reconversion to electrical energy may occur through fuel cells, although its direct use in SI cogeneration engines is often a preferable route, especially to avoid investment costs in further components beside electrolyzers. The interest towards hydrogen use for combustion in SI engines, indeed, is gaining a renewed attention by energy saving companies (ESCos), as a way to reach an overall higher efficiency of cogenerative systems over a continuous yearly basis.

When dealing with hydrogen combustion in SI ICEs, its larger flammability limits compared to NG and gasoline and a knocking more resistant behavior must be considered: its oxidation process greatly differs from that of conventional fuels and would need proper redesign of the combustion chamber to face the higher propagation speed and the resulting higher pressure gradients and temperatures within the combustion chamber. Blending hydrogen with NG is an effective solution to profit of this energy vector without modifying too much the existing facilities and to simultaneously achieve greater efficiency in power production under electrical energy surplus occurrence. Hydrogen–NG blends also allow engine operation with lean mixtures with consequent positive effects on operating costs.

A summary of the main properties of NG, hydrogen, hydrogen-enriched compressed natural gas (HCNG, 10% of hydrogen in volume), and gasoline is reported in Table 1. Blending hydrogen with NG is not straightforward and always convenient as the hydrogen fraction needs to be optimized according to the specific engine operating condition. As the level of hydrogen increases at a fixed engine operative point, the in-cylinder temperature increases too, while the combustion duration reduces. The overall effect can be summarized in lower unburned hydrocarbons (UHC) and CO emissions, but in augmented nitrogen oxide (NOx) emissions, due to the higher combustion temperature and enhanced heat losses (Yan et al., 2018).

Many laboratory experiments were performed on hydrogen–NG or hydrogen–methane combustion to assess the advantages of these blends up to the release of a commercialized gas in the United States with the registered trademark of Hythane, a mixture of 15% H₂ and 85% of NG in volume by Kavathekar et al. (2007). The achieved results are described into detail in the next section of this article. The benefits of hydrogen addition in reducing the overall noxious emissions and in improving thermal efficiencies are also demonstrated in the study by Pede et al. (2007).

However, despite the high number of conducted works, there are still some open challenges, for example, the choice of the right amount of hydrogen fraction allowable to avoid rapid pressure rises and, more importantly, defining the optimal combination that exists between hydrogen addition, the excess air ratio, and spark timing (Alrazen and Ahmad, 2018; Visciglio, 2019). This is especially relevant to cogeneration purposes, where SI engines, also of very large size, are powered extremely lean and with high supercharging pressure, so as to reach extremely interesting thermal efficiencies for this kind of combustion concept.

The present study intends to give a contribution in this direction. In this perspective, a numerical simulation approach is followed, as it can provide useful tips for a deeper understanding of hydrogen–NG blend combustion characteristics with reduced costs as compared to experiments. Zero or monodimensional (0–1D) models are typically employed to conduct fast parametric analyses to investigate the optimal tradeoff between power output, fuel consumption, and emissions (Yan et al., 2018). A model of this kind is developed here within the GT-Power environment, with reference to a real cogenerative engine designed to operate under NG fueling, in order to virtually characterize this retrofitting option for already installed systems.

A more comprehensive summary of the main experimental and numerical studies conducted for hydrogen–natural gas blend fueling of ICEs is first introduced in the subsequent section, with emphasis on SI reciprocating engine applications. This is then followed by a description of the numerical model proposed here, as validated on available technical data. The main novelty of the present study relies on the characterization of the engine performances when fueled with different hydrogen–NG blends by virtue of an original predictive approach able to correctly reproduce the blend combustion speed within the engine combustion chamber, according to the actual volumetric fractions of the components of the mixture. Engine performances are analyzed with the aim of assessing the suitability of several hydrogen–NG blends by varying the spark advance, the air-to-fuel ratio, and the hydrogen fraction. Lastly, an optimization study is conducted in order to identify the optimal condition in which the engine is characterized by the highest power production with the minimum fuel consumption and related environmental impact.

The achieved results provide guidelines for optimization according to the user’s energetic demand, giving particular attention to the environmental impact of the engine in terms of noxious emissions and to the structural limits of the system, these lasts identified by keeping the derivative of pressure within coherent limits with those assumed by the original manufacturer.

This section gives a look over the main experimental and numerical studies focused on the analysis of the performances of hydrogen–NG-fueled SI engines.

Several already existing studies in the literature give comprehensive reviews dealing with hydrogen–NG blends in SI engines. The first efforts are attributed to Nagalingam et al. (1983), where experiments on the performance and emission characteristics of a research engine fueled by various blends with hydrogen content were conducted. Later, Swain et al. (1993) tested a 20% H₂–80% NG mixture over several SI research engines operating at different engine speeds and loads, registering an increase in the NOx and a reduction in the CO and UHC emissions, with a general increase in the flammability limits.

Higher percentages of hydrogen volume fraction were also tested by Sierens and Rosseel (1999), suggesting the need to operate with an upper limit of the 20% fraction of hydrogen to guarantee a moderate increase in the NOx emissions below the imposed limits, while Bauer and Forest (2001) provided an optimum operative condition for a fuel mixture with 60% of hydrogen in volume.

It is also worth mentioning the work conducted by Mehra et al. (2017), where an overview over the technical approaches followed for hydrogen–NG combustion optimization (such as lean-burn combustion, exhaust gas recirculation (EGR), and direct injection (DI) systems) is presented, together with numerical procedures as the quasidimensional models.

The several processes that can lead to NG and hydrogen production are summarized from both a technical and economical point of view in the review study of Alrazen and Ahmad (2018), where the effects of hydrogen–NG blends in SI engines are described with a closer look to the noxious emissions. The environmental impact of hydrogen–NG-fueled engines is also largely discussed by Yan et al. (2018), together with a wide description of the influence of the hydrogen fraction over the individual thermochemical properties (LHV, autoignition temperature, flammability limits, and ignition delay) of the resulting blends.

As an overall guide to balance emissions and performances at different operative conditions, Ma et al. (2010) recommended a mixture with 20% in volume of hydrogen while Moreno et al. (2012) suggested using a 30% blend. These percentages are also advised by Flekiewicz et al. (2012) to avoid abnormal phenomena as knocking in SI engines. Several authors (Ma et al., 2010; Hoekstra, Collier et al., 1995; Huang, Wang et al., 2007) investigated on hydrogen–NG fueled engines operating with lean mixtures, achieving a significant extension of flammability limits and lower NOx emissions (40–50%) with respect to natural gas.

As already said, among the numerical simulations performed in this field, quasidimensional models are the most adopted ones due to their intrinsic flexibility and simplicity to predict the overall engine combustion and emission performances, despite the fact that no information can be achieved regarding the in-cylinder flow-field and thermochemical property evolution (Ma et al., 2008; Verhelst and Sheppard, 2009; Djouadi and Bentahar, 2016). On the other hand, a high detailed characterization of the internal flow-field may derive from high-computational approaches such as computational fluid dynamics (CFD) models, although heavy burdening of the computational time is required (Zaker et al., 2015; Costa and Piazzullo, 2018).

A good compromise solution between these two approaches is represented by monodimensional (1D) models, where the Navier–Stokes equations are solved in a monodimensional discretization and the combustion and pollutant emission formation is predicted through suitable and validated submodels based on empirical correlations. Among these approaches, it is worth mentioning the work of Mariani et al. (2012) that studied the performances and emissions of a SI ICE fueled by NG–hydrogen blends under a stoichiometric condition in order to assure an efficient exhaust after treatment adopting a three-way catalyst, while Kamil and Rahman (2015) studied the effects of hydrogen–NG blends on a single-cylinder port injection engine.

In general, the exploitation of hydrogen–NG blends in ICEs has proven valuable to improve thermal efficiency and to produce less noxious emissions. The challenge of parameter optimization still needs to be faced, especially for ultralean mixture conditions. The developed formulation based on a 0–1D approach is here proposed to find a relationship helping to calibrate the engine as based on the hydrogen fraction in the blend, this last being a mandatory aspect poorly discussed in the literature if cogenerative applications have to be considered (Alrazen and Ahman, 2018).

The proposed numerical model is developed with reference to the Jenbacher JMS 420 engine, of the type-4 class, employed for lean-burn cogenerative purposes under NG fueling and characterized by a high power density and outstanding efficiency. The engine is equipped with an ABB TPS 52-F32 turbocharger, while a bypass circuit serves, downstream of the compressor, to the regulation of the output power and as a protection device (Chvatal et al., 1997). The main geometrical characteristics and operative conditions are reported in Table 2.

The engine performances under NG and hydrogen–NG fueling are evaluated through a 0–1D numerical model developed within the GT-Power™ (Gtisoft Website, 2021) environment. The model solves the Navier–Stokes equations in a one-dimensional framework; thus, all quantities are averaged on pipe sections and are assumed as changing only along the flow. Pipes and flow splits are discretized into subvolumes, to better approximate every bend or restriction. In each volume, scalar variables as pressure and temperature are assumed to be uniform, while vector variables are calculated at each boundary. All the engine components are modeled, from the intake to the exhaust ducts, the throttle, the valves, and the cylinders, whose geometrical properties are attained from an available CAD model of the engine. The engine model representation is shown in Figure 1, where the presence of a wastegate valve is also clearly visible. This makes for a correct turbomatching between the compressor and the turbine and serves to regulate the maximum attainable boost pressure under each operative condition.

The air compressor is modeled by relying on the similar ABB TPS 57-D compressor whose map is available in the literature (ABB Website, 2021), while the turbine performances are simulated by relying on the operative map available in the software after a proper tuning of the relevant mass flow rates.

The combustion phase is modeled through the predictive model named “EngCylCombSITurb” (Hernandez et al., 2005). This approach is preferred to the classical Wiebe function (Heywood, 1988), where the fuel burned mass fraction is defined through a sigmoid-like function, as this last is validated for traditional fuels but it fails for nontraditional ones as for hydrogen–NG blends, whose chemical characteristics and combustion behavior depend upon the volumetric fractions of the components (Caputo et al., 2018; Caputo et al., 2019).

The adopted model calculates the laminar speed through the following expression: S L = [ B m + B φ ( φ − φ m ) 2 ] ( T u T r e f ) α ( p p r e f ) β , (1) where

• B m is the maximum laminar flame speed at reference conditions (T_ref = 300 K and P_ref = 1 bar)

These parameters are already tuned and listed for different traditional fuels as methane in the GT-Power™ library, but not for hydrogen–methane mixtures. Therefore, these parameters are here calculated for each mixture within the Chemkin™ environment by relying on the detailed kinetic mechanism Gri-Mech 3.0 (Smith et al., 1999), so as to build a more customized combustion model as the hydrogen fraction varies in the fuel mixture. The validation of this approach was previously performed by Caputo et al., 2019, with reference to a ternary mixture of H₂–CH₄–CO with respect to experimental measurements under reference conditions. Figure 2 reports the calculated laminar flame speed as a function of the equivalence ratio for NG, hydrogen, and three blends of hydrogen–NG (hydrogen volume fraction at 10, 20, and 30%) at standard pressure and temperature and by varying the fuel–air equivalence ratio. Increasing the hydrogen percentage in the fuel blend obviously leads to higher laminar flame speeds and moves the maximum value of the bell-shaped curve toward fuel-richer mixtures. Lastly, in the engine model, the wall heat transfer phenomenon is described through the work of Woschni (1967), while Zel’dovich’s chemical mechanism Zeldovich et al. (1947) is adopted for the prediction of the NO_x emissions.

Model validation is performed under NG fueling at the steady-state condition under full and 75% load by comparing the datasheet brake power, brake efficiency, fuel and the mixture mass flow rate, brake specific fuel consumption (BSFC), and brake mean effective pressure (BMEP) of the real engine under study with the results achieved by the model. The simulations under both loads are conducted after a proper regulation of the throttle valve angle aimed at achieving the desired power. The imposed initial and boundary conditions are shown in Table 3.

The comparison between the datasheet and calculated results under both loads is reported in Table 4. The brake efficiency is evaluated as the ratio of the brake power and the primary power, this last obtained by multiplying the fuel mass flow rate by the lower heating value of NG. The results can be said to have good agreement, as a maximum absolute error slightly below the 2% can be noticed for the brake efficiency at the 75% load case, as derived from a numerical overprediction of the total mixture mass flow rate.

The brake efficiency of the considered engine is very high due to the lean operation and supercharging as for the new-generation cogenerative engines by the main current original equipment manufacturers (OEMs).

The feasibility of using hydrogen–NG blends is evaluated by replacing NG with increasing percentages of hydrogen, in a range between 10 and 30% by volume. Properties of each blend are shown in Table 5 in terms of density, stoichiometric air-to-fuel, and LHV. A parametric analysis is performed for each blend by varying the spark timing and air-to-fuel ratio in order to study the influence of these parameters over the engine efficiency, the combustion characteristics, and the formation of noxious emissions.

A numerical model of a real turbocharged SI engine for cogenerative applications, developed in GT-Power environment, is presented here. The engine model is first validated under NG fueling, and then, its performances are numerically evaluated by substituting a certain percentage of natural gas with hydrogen. A parametric analysis highlights the quantitative influence of the spark timing and the air-to-fuel ratio over the engine performances and noxious emission formation as the hydrogen fraction is varied.

The main results show that

• higher amounts of hydrogen fractions in the hydrogen–NG blend speed up the combustion propagation, determining higher and more anticipated pressure peaks. The combustion efficiencies and fuel consumptions enhance, but the formation of the NO_x pollutants grows exponentially with the hydrogen fraction. A 20% of hydrogen fraction in the blend guarantees the highest brake power regardless of the air-to-fuel ratios and spark timing, as a further increase in the hydrogen fraction leads to reduced volumetric efficiencies.

Lastly, an optimization procedure is carried out, aimed at individuating the best condition in which the engine operates with the maximum power output and the minimum BSFC, NOx, and UHC emissions without burdening the system with excessive solicitations. In summary, shifting to leaner air-to-fuel ratios is generally preferred to avoid rapid pressure rises, especially if the blend is characterized by a hydrogen fraction higher than 15%.