Assessing the Impact of Natural Gas and Hydrogen Blending in Integrated Energy System Modeling

This paper assesses the impact of natural gas and hydrogen blending in integrated energy system modeling based on a novel blending transport problem (B-TP). In contrast to a standard transport problem which oversimplifies technical realities of pipeline gas flows in the context of blending, the B-TP (mixed-integer linear program) ensures compliance with the maximum hydrogen blending rate and that natural gas and hydrogen flow in the same direction in a pipeline. To assess the impact of the B-TP, we formulate an expansion planning optimization model of the integrated power, natural gas, and hydrogen sectors based on the objective of minimizing total system cost. Our case study shows that the gas flow formulation strongly influences investment decisions in the power and hydrogen sectors and that omitting the technical realities of blending can ultimately lead to suboptimal infrastructure planning (represented in the model as “regret’’ in the form of non-supplied hydrogen).

Abstract-This paper assesses the impact of natural gas and hydrogen blending in integrated energy system modeling based on a novel blending transport problem (B-TP).In contrast to a standard transport problem which oversimplifies technical realities of pipeline gas flows in the context of blending, the B-TP (mixed-integer linear program) ensures compliance with the maximum hydrogen blending rate and that natural gas and hydrogen flow in the same direction in a pipeline.To assess the impact of the B-TP, we formulate an expansion planning optimization model of the integrated power, natural gas, and hydrogen sectors based on the objective of minimizing total system cost.Our case study shows that the gas flow formulation strongly influences investment decisions in the power and hydrogen sectors and that omitting the technical realities of blending can ultimately lead to suboptimal infrastructure planning (represented in the model as "regret" in the form of non-supplied hydrogen).
Index Terms-energy system modeling, expansion planning, hydrogen, natural gas, blending, decarbonization NOMENCLATURE A. Sets

I. INTRODUCTION
Worldwide, many countries have embarked on the pathway to becoming low-carbon or carbon-neutral societies.The European Union (EU) is leading the way with the goal of reducing greenhouse gas emissions by 55% by 2030 (compared to 1990 levels) and achieving climate neutrality by 2050 [1].One of the most prominent options in this context is the energy vector hydrogen (H 2 ).In light of Russia's invasion of Ukraine, the impending threat of energy shortages and the massive price increases, the expansion of H 2 infrastructure is to be vastly accelerated [2].To date, however, there is no infrastructure for production, storage, and transmission of (renewable) H 2 , nor is there H 2 demand at large scale.Given this, blending H 2 with natural gas (NG) is considered a promising mode of transportation, particularly during the initial phase of establishing a H 2 economy (2020s and early 2030s) [3] when both H 2 production and demand are still limited.In previous work [4], we closely investigated state-of-the-art integrated power, NG and/or H 2 expansion planning literature with a particular focus on gas flow formulations.The review showed, that in holistic energy system models (ESMs) it is common practice to represent pipeline gas transmission as a transport problem (TP), which disregards the physical relation of gas flow and pressure in a pipeline.Nevertheless, a TP can give a sufficiently good approximation of NG and H 2 flows, e.g. when NG and H 2 flow in separate transmission pipelines (in the future) [5], [6] or when only H 2 is transported [7], [8].However, when it comes to modeling NG and H 2 blending, the TP framework is an oversimplification of reality.This is since (i) it allows NG and H 2 to flow in opposite directions within a pipeline; (ii) it cannot guarantee compliance with the maximum permitted H 2 blending rate; and (iii) the direction of the endogenously determined gas flows can change in each time step, which is not possible in reality.Therefore, the research questions we aim to answer in this paper is how omitting these technical realities impacts expansion planning decisions in integrate ESMs and how the gas flow formulation can be improve to adequately model NG and H 2 blending.
Motivated by this gap in the literature and as an original contribution, we present a modified transport problem based on a mixed-integer linear program (MILP) formulation permitting blending of NG and H 2 for joint pipeline transmission.To test the proposed formulation and as a second minor contribution, we formulate an integrated (partially) sector-coupled ESM of the power, natural gas, and hydrogen sectors based on the objective of minimizing total system costs (generation expansion and operation).The ESM includes, inter alia, a detailed model of the power sector [9]; formulation of electrolyzer (EL), steam-methane reforming (SMR), and (long-term) NG and H 2 storage units.Furthermore, we consider that a share of the system's NG demand can be covered by H 2 and co-firing of H 2 in combined-cycle gas turbines (CCGTs) and open-cycle gas turbines (OCGTs).Our model is formulated as a deterministic MILP with a highly flexible temporal structure.This permits modeling of a full chronological time series (hourly resolution) as well as representative periods, e.g.representative days, in the context of an investment problem.
The remainder of this paper is structured as follows: In section II we provide the model formulation.In section III we conduct a comprehensive case study based on a stylized ESM.Finally, section IV concludes the paper.

II. MODEL FORMULATION
We formulate a cost minimization problem based on the objective function (1), where (i-vii, x, and xii) denote operating costs and (viii, ix, and xi) denote investment costs.The total system costs comprise: (i) cost of NG supply to the system from gas wells ch4w; (ii) operation and maintenance (OM) costs of gas-fired thermal units t; (iii) unit commitment costs of thermal units (except gas-fired thermal units); (iv) OM costs of renewable units r; (v) OM costs of storage units s (power system); (vi) energy non-supplied; (vii) H 2 and NG nonsupplied; (viii) investment costs for power generation units; (ix) investment costs for H 2 units; (x) OM costs for H 2 units; (xi) investment costs for NG units; and (xii) OM costs for NG units.The system-wide NG demand is met from gas wells (i).Thus, the unit commitment costs (except for OM costs (ii) of CCGTs and OCGTs), costs for NG consumption from SMR units, and costs of meeting NG demand other than for power generation are accounted for implicitly.Constraints ( 2)-(3) establish lower and upper bounds for power, H 2 , and NG non-supplied, while (4) limits the investments in generation, H 2 , and NG units and establishes non-negativity.

A. Power System
Since the focus of this paper is to investigate the impact of NG and H 2 blending on expansion planning decisions, we do not explicitly state the mathematical formulation of the power system here, but provide a brief description.The applied formulation can be found in the online appendix and is based on [9].The power system is governed by a power balance equation comprising power generation of all units t, r, and s, charging of power storage units s, auxiliary power required for charging of H 2 units h2s, power demand, and power nonsupplied.The power flow is formulated as DC-optimal power flow (DC-OPF).

B. Gas System
The formulation of the gas system comprises gas wells, gasfired thermal units (CCGTs, OCGTs), (long-term) NG and H 2 storage units, EL, and SMR units as well as blending constraints for NG and H 2 .For the sake of clarity, only the framework of units coupling the gas and power systems is given here.The detailed formulation of all components is provided in the online appendix.Constraints ( 5)-( 7) describe the gas consumption and power output of gas-fired generators Finally, ( 12)-( 14) govern the gas demand including blending of NG and H 2 ∀rp, k, m.
The NG and H 2 sectors are established by two separate balance equations, which are both very similar in their structure.The NG balance comprises: production from gas wells and storage units; the variable share of gas demand outside the power sector that can be supplied by NG; the NG consumption from gas-fired thermals, storage, SMR, and compressor units; and NG non-supplied.The H 2 balance comprises: production form EL, SMR, and storage units; the variable share of gas demand outside the power sector that can be supplied by H 2 as well as a dedicated H 2 demand; the H 2 consumption from gas-fired thermals, storage, and compressor units; and H 2 non-supplied.Furthermore, both balances include gas flows through pipelines and compressors.
The gas flow is governed by the novel blending transport problem (B-TP) given in (15)-(20) allowing for NG and H 2 blending as well as accounting for coherence of flow direction of both gases.To this end, the total gas flow in a pipeline (15) is represented by the two components NG flow f CH4 rp,k,m,n,l and H 2 flow f H2 rp,k,m,n,l , all of which are free variables.Thus, and for the fact that in reality gas flow direction is a direct result of the gas pressure gradient, coherence of flow direction has to be ensured ( 16)-( 17).We achieve this by introducing the binary variable α gas rp,m,n,l , which in addition restricts the gas flow direction to one decision per representative period.Constraints ( 18

III. CASE STUDY
The purpose of this case study is to investigate the effects of considering (or neglecting) detailed H 2 blending in an expansion planning framework.To this end, we compare planning results from the proposed B-TP to a standard transport problem (S-TP) (21)-( 22) ∀rp, k, a, which comprises a linear program (LP).For the S-TP, we assume that 10% of a pipeline's capacity is pre-reserved for H 2 transmission.Despite this assumption, the S-TP constitutes a more flexible system than the B-TP since (i) it allows NG and H 2 to flow in opposite directions within a pipeline, which is not possible in reality due to the pressure gradient; (ii) it cannot guarantee compliance with the maximum permitted H 2 blending rate, e.g.10% of the actual NG flow; and (iii) the direction of gas flows can change in each time step, which is not possible in reality.

A. Test System
The studied energy system is based on a modified version of an integrated 24-bus IEEE Reliability Test System and a 12node gas system presented in [10].Fig. 1 depicts the integrated NG and H 2 system and system data is provided in the online appendix.The case study is based on a brownfield approach where the only existing units are three gas wells and two long-term NG storage units.This allows to capture long-term, e.g.seasonal effects, in a representative period framework.Thermal candidate units comprise two 200 MW CCGTs, two 120 MW OCGTs, and a 152 MW coal-fired power plant.In order to reflect the ongoing transition towards renewabledominated energy systems, we consider 500 MW of variable renewable energy (VRE) sources (wind and solar) and battery energy storage systems (BESS) at each bus of the power system.H 2 candidates comprise: 200 MW EL, 1.02 MSm 3 /h SMR, 113,500 Sm 3 /h short-term storage units (per node), and a 14,160 Sm 3 /h H 2 cavern for long-term storage.All investment variables are continuous 1 , except for thermal units 1 Relaxing integrality for, e.g.VRE units, BESSs, EL units, etc., gives a good approximation as we are planning a GW-scale energy system.
978-1-6654-6441-3/23/$31.00 ©2023 IEEE Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.and long-term H 2 storage units, which are binary.The power flow is governed by a DC-OPF.The temporal framework is based on seven representative days with hourly temporal resolution.Representatives are determined via a k-medoids clustering procedure.Time series for demand comprise the Austrian power and NG demand data of 2020 (no spatial resolution), scaled down to the test system and distributed on the busses and nodes in accordance with [10].We consider an emerging dedicated H 2 demand, e.g. for fuel cell electric vehicles or direct reduction of iron for steel production.This is an exogenous parameter and (for simplicity) based on the same time series as the NG demand.In addition to this dedicated H 2 demand, up to 10% of NG demand can be substituted by H 2 .The same applies to NG consumption of CCGTs and OCGTs.
The case study is executed on a notebook with a 2.80 GHz 11 th Generation Intel Core i7-1165G7 (4 cores) and 32 GB RAM using GAMS 37.1.0and Gurobi 9.5.0.The number of continuous and discrete variables are: S-TP (76310 / 6061) and B-TP (76380 / 6131).The relative MIP gap is set at 0.1%.

1) Investment decisions:
Under the S-TP framework, total system costs amount to 1,265 MC.Investments in the power sector include the four gas-fired candidate units (640 MW) and a total of 5,446 MW wind, 1,632 MW solar PV, and 884 MW BESSs.In the B-TP case, total costs increase by 6 MC (+0.5%), which, a priori, does not seem vast.However, the impact of suboptimal planning can be much higher when it comes to operation of the system, which we assess in section III-B3.Taking a closer look at the investment decisions in the power sector (Table I), we find that there is a substantial shift in expansion planning decisions.This results in the net relocation of 2,113 MW of power generation units, which corresponds to 23.7% of the total installed capacity (based on the B-TP case).In particular, the model no longer invests in the CCGT connecting bus 16 and node 7. The situation is similar in the hydrogen sector (Table II).Here the net capacity  It is also notable that most of the additional H 2 investment in the B-TP case is in SMR (+52.0%)rather than EL units (+2.0%).This tendency is reflected in H 2 production from SMR (+72.2%) and EL, which even decreases (-4.8%).We conclude that SMR is favourable in order to compensate for the reduced flexibility of the B-TP formulation.The key takeaway from this paragraph is that although the impact on total system costs may be small, modeling detailed blending can strongly influence siting decisions in the power and H 2 sectors.
2) Operating conditions: In addition to expansion planning decisions, the resulting operating conditions can be of particular interest to the system operator.To this end, we investigate pipeline gas flows during representative period 5 (RP5), which corresponds to the representative day with the highest absolute load as well as total load in both the power and NG sectors.Looking at the combined NG and H 2 flows (Fig. 2), we find that they are not representative at all in the S-TP case.For example, the gas flow in the pipeline connecting nodes 4 and 7 changes direction six times, thereof five times within an eighthour period, which is not possible in reality.As intended, this does not happen in the B-TP case.
In the power system, we find that the annual generation from OCGTs decreases significantly (-16.3%).This is despite the fact that the B-TP formulation allows for higher NG flows (≤ 100% of the pipeline capacity) than the S-TP (≤ 90%).However, it is evident that this cannot compensate the reduced flexibility in the B-TP case, but results in a more constrained  gas system (CCGT is not built, see III-B1).In addition, we note that the number of hours in which loading of power lines exceeds 70% increases in the B-TP case (66 versus 113 hours).This is consistent with the increased investment in VRE (and BESS) resulting in higher curtailment in the B-TP case.We conclude that the B-TP formulation can positively affect the significance of operating results in the gas and in the power system.
3) Quality of planning decisions: In order to get an idea about the quality of expansion planning decisions, we evaluate results from the S-TP against the more realistic B-TP framework.To this end, we fix all investment decisions from the S-TP case and run the B-TP case as an operational problem (by omitting additional investments).As depicted in Fig. 3, investments from the S-TP case are not sufficient to provide H 2 demand, although total H 2 demand decreases by 30.0 MSm 3 (12.1%).This is mostly due to less H 2 utilization in the NG sector.Consequently, we end up with 7.3 MSm 3 of H 2 nonsupplied (2.6% of total H 2 demand), with an associated cost of 10 C/Sm 3 .Ultimately, this leads to an increase in total system cost of 98 MC (15.2%) (26 MC (4.0%) if cost of H 2 non-supplied is excluded) due to reduced NG substitution and altered operating decisions in the power sector.As a takeaway, oversimplified modeling of NG and H 2 blending can result in suboptimal expansion planning.

IV. CONCLUSIONS
In this paper, we proposed a novel MILP gas transport problem allowing for NG and H 2 blending in an expansion planning framework.To this end, we formulated an integrated power, NG, and H 2 ESM and benchmarked our formulation against a standard transport problem (LP).Our results clearly show that considering the technical details of blending can have a significant impact on expansion planning results -and siting in particular -not only in the NG and H 2 sectors, but also in the power sector.The higher level of technical detail of the proposed formulation also increases the significance of operational model results, e.g.gas flow directions, which can be of particular interest to the system operator.That said, the increased accuracy comes at the cost of additional computational effort, which can be challenging when modelling large-scale sector-coupled energy systems, e.g. in combination with unit commitment.Nevertheless, omitting these technical details can ultimately result in suboptimal infrastructure planning, which could potentially lead to non-supplied H 2 .

Fig. 2 .
Fig. 2. Combined NG and H 2 flows per pipeline under the S-TP (top) versus the B-TP framework (bottom) during representative period 5.

Fig. 3 .
Fig. 3. Quality of planning decision for the S-TP versus the B-TP framework.
)-(19) comprise the bounds on H 2 flow based on the maximum H 2 blend rate B H2 .Finally, (20) represents the bounds for gas flows.The constraints apply ∀rp, k, a.

flow in % of pipeline capacity Hours of representative period 5
978-1-6654-6441-3/23/$31.00 ©2023 IEEE Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.