A Network Modeling Systematics for Transition Paths Toward Climate Neutral Gas Networks—NeMoSys ['nεmɒsis] –
Abstract
Meaningful network modeling requires high spatial and temporal resolution and thus adds further complexity to the already complex energy systems analyses; unsecured and varying framework conditions, as well as their changes over time, have to be taken into consideration on the backdrop of path-dependencies for network development over time. This perspectives article provides a systematic framework for gas network modeling, starting with a morphology of modeling dimensions and elements and the hierarchy of influencing factors. It gives an overview on the necessary data and attributes, development steps, and assessment criteria; it showcases solutions for data management, fluid dynamic network simulation, and results assessment that have been applied and developed in the course of the TransHyDE project and other related research projects. Framework and solutions are merged into a consolidated network modeling systematics that shall be further developed and made available to researchers as well as to network planners for practical application and governmental bodies for assessments. The systematics presented here will be further developed, applied, and operationalized in a database and tool compilation with open-access for collaboration and further development.
1 Structure and Objectives
Energy systems analyses rely on tools and data and their useful combination. The accuracy and reliability of their results highly depend on the level of detail in the two main dimensions, time and space. Temporal differentiation to the hour has become standard.[1, 2] Spatial differentiation at regional and local levels is increasing for supply and demand analyses (e.g.[3, 4]). High granularity in both dimensions—temporal and spatial disaggregation—is essential when it comes to the analysis of network infrastructures, and particularly for their multisectoral integration. In addition, the third dimension comes into play, which is the resolution and accuracy of the network modeling itself; this dimension ranges in several layers from coarse and linearized regions/capacities modeling to fine-granular technophysical fluid dynamic modeling.
In the context of the (not only) German energy transition toward a fully climate-neutral energy system, we know that temporal and spatial framework conditions change frequently already in present reality; for the modeling tasks in energy systems analysis, more severe changes and several alternative developments must be considered in different future scenarios. The interplay of, e.g., the most favorable wind and solar electricity generation, the most efficient origination of green hydrogen, and the prime allocation of consumption may completely change the outlay of an adequate energy network integration, including electricity, natural gas (over a transitional phase only), and hydrogen (possibly with changing “colors” during the transition).
To date, there is no systematics available that allows for a swift modeling of these network designs and subsequent analyses of their important technoeconomic properties. The need for such a systematics is demonstrated, i.e., by the current intense discussions about reasonable development paths for the different energy networks.[5-8]
In the following, we present a comprehensive overview of the broad range in which network models are designed and applied, starting with how the network modeling systematics is embedded in general energy systems analyses (Section 1.1) and how the complexity of network modeling can be systematically categorized in the dimensions time, space, and network modeling accuracy, the spheres for data compilation and management, and the layers for the network resolution and technophysical precision (Section 1.2). Section 2 discusses the morphological scope of network model applications, dimensions, and detail design options (Section 2.1), approaches to narrow down reasonable combinations for their selection (Section 2.2), and prepares for a system of top-down and bottom-up relations to hierarchically combine the less and the more detailed analysis levels especially for the layers of network modeling resolution and accuracy (Section 2.3).
Section 3 then structures the process for network modeling analyses from input data over the core tasks in network modeling through to the results, their documentation, and interpretation. For each step, the main tasks, requirements, and tool concepts are presented, including prototype solutions that exist already and/or are under development and will be made available in the further continuation of TransHyDE.
In Section 4, we show some examples for the successful application of the main elements of the network modeling systematics. Section 5 compiles the conclusions and recommendations.
1.1 Embedding in Energy Systems Analyses
Energy systems analyses play an important role in understanding the complexities of modern energy infrastructure and facilitate informed decision-making. They can be classified into two categories: top-down models, which illustrate the interaction between energy systems and the broader economy based on macroeconomic indicators, and bottom-up models, further divided into optimization and simulation models.[9, 10] Optimization models utilize technoeconomic parameters of energy-related technologies as inputs, enabling the computation of optimal investments and scheduling based on a target function, typically focused on minimizing costs. Linear (LP) or mixed-integer linear programming (MILP) techniques are commonly employed to achieve this optimization process.[11] Early pioneers of energy system optimization models are the Model for Energy Supply System Alternatives and their General Environmental Impact (MESSAGE)[12] and the Market Allocation (MARKAL) models[13] that were further developed into The Integrated MARKAL Energy Flow Optimization Model (EFOM) System (TIMES) model[14] by combining it with EFOM(.
These established models have the shared objective to illustrate probable developments in the energy system at different spatial levels, spanning decades.[9] While traditional energy system models can capture some dynamics, they may be insufficient in representing the complete complexity and interconnectedness of these systems. The interactions and interdependencies between system components significantly influence overall behavior. Linear optimization models, like MARKAL/TIMES and MESSAGE, offer valuable insights into energy system cost optimization. Nevertheless, they can be inadequate in capturing spatial and temporal dependencies, dynamic interactions, and feedback loops that are inherent in real-world energy networks. To overcome these limitations, network modeling has increasingly been adopted, providing a more comprehensive approach. This allows for researchers to capture the topology, structure, and dynamics of interconnected components within an energy system.[15]
Transitioning from energy system models to energy network models, it is important to recognize the role of network models in capturing the complex interconnections and dynamics within energy systems. Network models serve as tools for capturing the intricate interconnections and dynamics within the energy infrastructure. Nodes within the network represent key components such as energy sources, technologies, and demand sectors, while edges illustrate their relationships and flows.
Recent advancements in this field include PyPSA-Eur, which encompasses the European transmission grid while also providing the capability for country-specific analyses by utilizing openly available data.[16] In this context, the PyPSA-Eur model is particularly worthy of mention as it enables the modeling of hydrogen network capacities.[17]
There exist several ways in which network models are embedded in energy systems analyses to enhance system understanding, optimize operations, and support long-term planning. Although there is a wide range of approaches available for modeling power networks, such as power flow analysis, optimal power flow, capacity expansion planning, and distribution system analyses, the modeling of networks for gaseous energy carriers has not received the same level of attention and research.
1.2 Dimensions, Spheres, and Layers for Network Modeling Systematics
In addition to the dimensions of energy system models and the indispensable spatial disaggregation, network models can be further differentiated from several perspectives. Their purposes—from a bird's eyes’ view for political decision-making to short-term dynamic analyses for network operation and control—include political, technical, and economic aspects. Each of the dimensions can be varied in a wide range with a plethora of combinations. In this perspectives article, we give a comprehensive overview of the main dimensions, spheres, and layers in technical terms, including their respective advantages and disadvantages as well as the repercussions on actual data requirements and practical steps in network modeling. We synthesize a compilation of reasonable and recommendable combinations of the dimensions for a hierarchical set of applications. 1) The dimensions considered here are time, space, network modeling resolution, and accuracy; 2) The spheres are the data collections for the description of the a) network task; b) the infrastructure of the network itself; and c) the details of configurations and settings for the elements of the network in the simulation runs; and 3) The layers are the different levels of resolution and accuracy of the network simulation, starting from coarse capacities/regions models down to technically detailed models including active elements for fluid-dynamic network simulation.
The generic procedure for gas network modeling is shown in Figure 1.
It starts with two main data input categories which should clearly be distinguished into the spheres where the data originate—the scenario-dependent network usage cases on the one side and the network topology on the other. Even though these categories are in some gas flow simulation software tools (e.g., SIMONE[18]) technically combined into one data set that also is shortly referred to as “scenario,” the differentiation is highly recommendable. The reasons are: 1) The scenario-based network usage cases represent the task the network has to cope with; they change routinely over time: a network usage case for an hour of high demand and maximum storage withdrawal, e.g., is different from an hour with low demand and storage injection; and 2) In contrast, the network topology represents the network that does not change routinely, but only with major planning, investment, and construction decisions; the investment-related infrastructure data define the network's capabilities.
The network simulation itself is then carried out, based on the combinations of the network usage cases and the topology data, with potentially several layers of different network resolution and accuracy. These range from 1) a network simulation based on a combination of regions and network transfer capacities between these regions that determine the maximum gas flow over the regions’ borders; this approach implicitly linearizes the flow simulation and does not account for the nonlinear properties of the flow of compressible fluids through pipeline networks; versus; 2) a more detailed fluid dynamic gas flow simulation accounting for the nonlinear equations for flow description;[19] this requires more detailed data and network simulation tools. A comparison and combination of capacities/regions and fluid dynamic models is possible and at the center of the network modeling systematics presented here (see Section 2).
The results of the network simulation runs then produce the output data which describe the gas flows over the network. This allows for a subsequent assessment of the adequacy of the network infrastructure with a more long-term view toward investments, and a more short-term view toward operational conditions of the network, depending on the time frames of the network simulation.
2 Morphology
2.1 Entirety of Network Models
In contrast to prominent energy system analysis models which basically neglect network aspects completely,[20-22] networks are considered on a first rather aggregate level in many European analyses when reference is made to different EU member states[6] and the available cross-border capacities between them.
A similar approach is chosen for reflecting the network limitations that exist between different market regions in European energy system analyses when integrated market and grid research is carried out.[23, 24]
The next intermediate level is reached by the national network development plans (e.g., refs. [25-27]) and in the current discussion about the German “Wasserstoff-Kernnetz”[28] in the making. More detailed network models are typically applied internally by gas as well as electricity transmission system operators (TSOs) and, currently only for the latter, also in a coordinated manner by the European TSOs. These network models are also used by the TSOs for fulfilling their transparency and publications obligations: 1) In a longer-term application, network models are applied for the assessment of the adequacy of existing network infrastructures for current and future tasks; when inadequacy is found as a result, network expansions and thus investment planning and realization are the consequence; 2) For a shorter term, TSOs use network models for the determination of transfer capacities that can and must be given to the markets to facilitate energy transactions between the adjacent regions. This is the case for the booking procedures that cover all capacities in the gas grid access regimes, and the capacity calculation for cross-border capacities in the electricity markets; and 3) In an even shorter-term application, network models are used for operational and technical safety and grid stability analyses, including the detection of short-term congestion in electricity networks and the ascertainment of suitable remedial actions to overcome them.
In a more detailed spatial resolution, network models are implemented for the definition, planning, and execution of network expansion projects for transmission as well as distribution networks for electricity and natural gas, for the latter also comprising the changeover to hydrogen networks on the basis of existing natural gas pipelines (“repurposing”) and/or newly built pipelines.
In terms of the temporal dimension, gas grids—like other networks as well—have to be designed and assessed for the transmission or distribution of the maximum flow. This means that time-dependent variations have to be considered and flow values that do not account for them are of very limited scientific value; a temporal disaggregation down to at least time steps of one day or preferably 1 h is necessary. This is the temporal granularity where all the applications mentioned above are operated for the purposes of capacity assessment, expansion planning, and commercial booking. In all of these cases, steady-state flow calculations are standard, i.e., transient effects resulting primarily from the inertia of the mass flows are not considered.
For detailed operational and technical safety analyses, the implementation of shorter time steps down to the minute range is necessary, as well as the inclusion of transient gas flow behavior in the network simulation. This applies, e.g., to the pressure shock when safety valves are shut and pressure increases as a consequence of mass flow into the final pipe section; or for the effects of extreme gradients when hydrogen re-electrification power plants need to quickly ramp up to compensate for low availability of volatile solar or wind power generation—and the pressure drops in the final pipe section, because it is largely emptied before sufficient inflow from upstream sections take places.
From a technical perspective, further variations of the network models exist with respect to the elements of the network and their properties: 1) Pipeline connections can be combined into bundles (being represented by standard capacities between regions’ borders or by a single pipeline in a fluid dynamic model) or as single items representing a number of parallel pipelines, potentially with different diameters and pressure profiles, in a pipeline corridor. The technical attributes for an accurate reproduction of the pipelines’ properties in the network model comprise the pipeline diameter, the pipeline length, the exact routing, and the (maximum) operational pressure. The pipe roughness, though it is a relevant parameter, generally can only be estimated as comprehensive measurements are impossible; and 2) Compressor stations and the compressor units can be reflected in a largely simplified way as just one compressor or including the actual number and internal interconnection of the compressor units inside the compressor stations. In the latter case, assumptions or experience-based information about the configuration of the compressor units in the station are the basis of the modeling of how compressors can be operational in the model in terms of serial or parallel interconnection and direction (reverse flow). For the relation between pressure ratio and volume flow, simplified, generic, or machine-specific characteristic curves can be utilized.[19]
The utilization of different characteristic curves also applies to gas storages; these can be modeled in a largely simplified manner with constant values for available pressure and/or flow values for storage injection and deliverability, or by more accurately reflecting the pressure dependency of the flow rates by considering the individual filling levels and characteristic curves of the storages. An interim approach consists of generic characteristic curves for types of storage (distinguishing between porous rock vs cavern storages).
For the configurations of the network, in addition to the configuration of the compressor units in the compressor stations, the configuration of valves and penstocks must be defined in particularly detailed network models.
One additional and very specific attribute for gas network modeling elements in the context of the transition toward climate neutrality is introduced by the question to what extent the network elements can be repurposed for the transport or distribution of hydrogen (“repurposability”). Specific investments for repurposing are only a fraction of those for newly built lines; thus, according to this attribute, the system split between natural gas and hydrogen subnetworks and more or less favorable supply/demand distributions over time can be analyzed.
For the economic dimension, all network elements’ specific costs (per piece or dependent on technical parameters like length and diameter, pressure levels, and capacity) are employed; these can depend on further parameters, e.g., the “repurposability” mentioned above for brownfield network elements versus specific investments for newly built lines in greenfield network elements. Additionally, expenditures for operation and maintenance, including in particular the energy consumed in the compressor drives, are covered.
2.2 Heuristic Compilation
To keep the more aggregated and the more detailed variants in all dimensions in connection, top-down and bottom-up relations are useful and can be systematically applied. Top-down relations are applied in a widespread manner for the disaggregation of supply and demand data in terms of spatial distribution and rolling out over standard or specific profiles into time series;[29] bottom-up relations are in many applications even more straightforward by summing up time series and/or regions. These both well-established schemes for the aggregation and/or disaggregation of data in space and time are related to the scenario data sphere that depicts the development over several years, and the network usage cases which break down the data description of the network tasks into, e.g., hourly specific data sets that are utilized in the network simulation.
Figure 2 shows this along the “Time” and “Spatial” axes.
For the network-specific data and characteristics, like capacities and/or flow values, approaches for aggregation and disaggregation are less commonly applied and require further research and development. This is shown in Figure 2, starting from an aggregate network transfer capacity over the bundled pipelines and their capacities and the line-specific level into the detailed modeling that includes the specifics of compressors and their configuration (the compressors here are representative for all active elements).
Figure 2 shows the aggregation and/or disaggregation in the dimensions of time, space, and network resolution/accuracy. Increasing network resolution and accuracy is displayed from top left to bottom right by layers from the surface into the depth. The network simulation details become more and more specific, starting with a coarse representation of network properties in capacities/regions models where gas flows are purely linear and easy to estimate, but not highly accurate. On the bottom right, the most accurate network simulation level for fluid-dynamic simulations includes compressors, valves, and their configurations inside compressor stations which improve accuracy but also stress computing and expert efforts to the limitations of applicability for large networks and/or high numbers of scenarios/network usage cases.
The top-down process for temporal and spatial resolution toward the detailed allocation of supply and demand data in the scenario data sphere can be described straightforward. It delivers specific entry and exit values for every network usage case in a detailed network model topology, as shown in Figure 2. These can then be applied for fluid dynamic simulations and delivered the outputs in terms of specific information (flow, pressure data, etc.) or general assessment (adequacy of the network or existing bottlenecks, etc.).
For the other direction—to open the network modeling systematics for bidirectional usage—the reaggregation of output data is also straightforward for the data that correspond to the scenario data sphere, i.e., from the specific entry/exit values per node and time-step over temporal and spatial aggregation into, e.g., (sub-)regional, country-wide data for days, week, or years.
For the network simulation and topology-specific outputs, however, the reaggregation is more complex. The bidirectional application, i.e., the utilization of the results of detailed analyses in terms of flow rates, capacities, and overall adequacy of the network under consideration, back to more aggregated levels, requires stepwise comparisons from level to level and careful transfer and transposition. This is described in Section 2.3 with a focus on capacity and adequacy assessments.
There, we follow a heuristic approach based on successful network simulation runs which can be achieved in highly detailed model setups for limited network sizes and/or network usage cases without exceeding computing and expert resources. The approach accepts suboptimal solutions whose feasibility, however, is guaranteed. The optimization-based validation of gas flow nominations for large networks in high technical resolution has been researched in one of the largest applied research projects in mathematical optimization, see refs. [30, 31] with further references. The gas flow nominations correspond to a network usage case, and their validation has been modeled in ref. 30 as a stochastic mixed-integer nonlinear optimization problem. The software implementing the method developed in ref. 30 was delivered to the participating industry partner; its running time for the validation of single nomination scenario can exceed several days or encounter problems of solvability,[31] thus posing problems for other applications. Research on mixed-integer nonlinear optimization for gas networks continues.[32]
2.3 Hierarchical Iterations
Bidirectional application typically integrates several iterations that consist of a combination of top-down and bottom-up approaches. We differentiate: 1) The changes in the network tasks, i.e., the different demand/supply, entry/exit values, etc., coming from outside the network, that are stemming from changes in the scenarios (over years or more) and/or the network usage cases (subsequent from the scenario changes or due to the temporal breakdown into hourly time series), are managed in the scenario data sphere; 2) The changes of the network infrastructure, i.e., the topology and/or the network elements and equipment, like pipelines and compressors, which in real life would mean investments and construction activities, thus influencing capital expenditures (capex), are managed in the network topology sphere; 3) Changes inside the given network which correspond to operational decisions, configurations, and settings (possibly changing operational expenditures—opex—but not capex); these decisions concern most notably the active elements, in particular the compressors and the configuration of compressor stations (Active elements represent potential degrees of freedom like operation of different compressors at different intensities and steering the pressure values in different network sections in different manners. Depending on the network usage case, there are often wide ranges of possible settings and configurations which all lead to a fulfillment of the network task, but their determination can be very time consuming in setting up the network simulation for successful runs. Unlike in electricity networks, active elements (for electricity this could be phase-shifters) are not a rare exception but are present in the transmission networks in high numbers.). These changes need to be effectively managed in network modeling analyses, even without any external of the network task (1) or infrastructural changes of the network (2). This takes place in the network simulation sphere.
The iterations on this third level correspond to the experience-based decision-making procedures in the network control centers of transmission system operators. Decisions about compressor configurations, valves, and penstock settings are necessary because there exist many degrees of freedom which generate some of the mathematical optimization problems addressed in ref. 32.
By accepting suboptimal solutions, the heuristic approach relaxes the required computational and expert time and effort and provides feasible network simulation results for predefined or newly found configuration settings for active elements. This iteration-based analysis is carried out only on the deepest network simulation level, defined by the dimensional combination of “Compressors,” “Nodes,” and the time steps of “Minute” or “Hour” (For steady-state network simulations, the time step does not represent a categorical distinction, but only differs in terms of the time that is meant to be represented adequately. The minute is recommended for transient network simulations.), as shown in Figure 3.
Results and findings from this level of the highest resolution are then transferred to less detailed levels and can be compared to results of network simulation results with a more limited resolution. This is illustrated in Figure 3 for the network simulation level without detailed analyses of the compressor configuration settings, carried out, e.g., in the time steps “Minute” or “Hour” for a more aggregated spatial resolution.
For every pair of two blocks inside the complete cube in Figure 3, we see one direct or several indirect interfaces where the transfer of results can be realized and compared for both sides. In case of inconsistencies between the two sides, iterations may become necessary to keep the inputs—from the network task sphere as well as from the network topology sphere—and the outputs resulting from the network simulation sphere coherent.
Comparisons can also show discrepancies between the direct results on different levels and thus determine limitations for the direct results on the more aggregate levels. Depending on the degree of accuracy that is required in a wider context, conclusions can be drawn for the applicability of low-resolution network models, or, in contrast, when more detailed network simulations have to be applied, e.g.: 1) for network usage cases with low utilization of maximum capacity values, the regions/capacities models may be accurate enough for coarse estimates of wide-area gas flow analyses;[33] or 2) for very high gradients in demand/supply values from power plant and/or storage operation, steady-state network simulation results need to be verified by transient network analyses.[34]
With this sequence of comparisons from layer to layer, the spatial aggregation can proceed, e.g., in the time step “Hour” and, in a subsequent step, with a simplification from line-wise capacities between the regions to bundled capacities and finally to interconnection capacities between aggregate regions on the least detailed level.
3 Databases, Tool Elements, and Interfaces
As seen in the previous chapters, gas network modeling requires comprehensive databases and accurate data management in separate spheres to effectively analyze and simulate the design and behavior of the network. In this section, we will explore the three spheres in more detail.
3.1 Scenario Data Sphere
Scenario data refer to the inputs related to gas production, consumption, storage injection, and withdrawal, and sometimes also transfer capacity values. Production data include gas production sources, volumes, and their profiles over time. Understanding the characteristics and temporal patterns of gas production helps in accurately representing the supply side of the network. Consumption data capture the demand for gas, including gas consumption patterns, peak demands, and variations over time. Accurate consumption data are essential for evaluating the network's capacity requirements and assessing its ability to meet demand. Storage injection and withdrawal data pertain to storage facilities and their injection and withdrawal capacities as well as their filling levels. These data reflect the capability of the network including the storages for balancing supply and demand fluctuations and guaranteeing security of supply. Capacity and transfer data involve calculating cross-border capacities and determining transfer values for energy transactions between adjacent regions. These data facilitate the efficient utilization of the network and support market interfaces.
Acquiring scenario data involves collaboration with gas market operators, energy agencies, and industry stakeholders. Reliable data sources, such as energy statistics, market reports, and forecasts, are used to capture realistic scenarios for modeling purposes. Scenario data in gas network modeling play a crucial role in capturing the uncertainties associated with the development of prices, consumption, and production. These uncertainties arise due to various factors such as market dynamics, policy changes, technological advancements, and external influences. To address these uncertainties, scenario data often rely on scientific models that incorporate a multitude of parameters and energy–economic interrelationships.
Scenario data typically can be stored in structured query language (SQL)-type databases. These are relational database management systems that utilize SQL to store, manage, and retrieve data in a tabular format with defined relationships between tables.
3.2 Network Topology Sphere
Network topology data provide essential information about the physical infrastructure and components of the gas network. These data include details about pipeline infrastructure, compressor stations, storage facilities, valves, and penstocks. The accuracy and completeness of network topology data are important for capturing the network's characteristics and ensuring reliable modeling results.
Pipeline infrastructure data encompass parameters such as pipeline diameter, length, routing, and operational pressure. Additionally, estimating pipe roughness, a relevant but challenging parameter, requires careful consideration as comprehensive measurements are often impractical.
Compressor station data involve representing compressors and their internal interconnections. Characteristic curves, such as pressure ratio versus volume flow, are utilized to establish the relation between compressor performance and operational conditions.
Storage facility data encompass available pressure and flow values, individual filling levels, and characteristic curves. Accurately reflecting the pressure dependency of flow rates and considering the type of storage (e.g., porous rock vs cavern storage) contributes to a more realistic representation in the network model.
Detailed configuration and specifications of valves and penstocks are necessary to accurately simulate network behavior, particularly during critical scenarios such as emergencies or maintenance activities. Acquiring network topology data involves accessing various sources, including industry databases, surveys, historical records, and collaboration with gas transmission and distribution system operators. However, challenges such as data consistency, completeness, and the availability of proprietary information must be addressed to ensure reliable data inputs.
Managing and integrating network topology data can typically implemented in NoSQL databases, e.g., of the MongoDB type which is a document-oriented NoSQL database that provides a flexible and scalable solution for storing and managing data in a JSON-like format. An effective tool for achieving this is DAVE[35] a data fusion tool designed for energy infrastructures.
3.3 Network Simulation Sphere
The interfaces between the scenario data sphere and the network topology sphere, on the one hand side, and the network simulation sphere, on the other side, reflect their granularity. That means that the interfaces have the same levels of aggregation and/or disaggregation in the dimensions of time and space. In addition, inside the network topology sphere, the layers of network modeling resolution and accuracy are managed. They range from the simplified linear modeling based on capacities between regions (in the following Section 3.3.1) to the fluid-dynamic modeling (Section 3.3.2), with intermediate levels, as shown in Figure 2. For both the basic modeling approaches, there are several software tools available. The most relevant software tools that have been applied in the course of TransHyDE and produced the case analyses’ results in Section 4 are presented here. Their overall characteristics are summarized at the end of Section 3.3.2, with more details in the appendix (Table A1); their application within the network modeling systematics as well as the interchange between the spheres are briefly laid out in Section 3.4, also with more details in the appendix Section A1.
3.3.1 Regions/Capacities Modeling
In the applications in Section 4, we will present an example of the abovementioned hierarchical network modeling approach. On the higher level (gas distribution in Europe) only transfer capacities are considered. The physical and technical aspects of transport besides the capacity limits can be neglected at this level. Consequently, it is much easier to perform optimization tasks on this scale, since the problem size is highly reduced.
Capacity modeling involves assessing the adequacy and reliability of energy system resources, considering generation capacity, transmission constraints, and system reliability. PyPSA (Python for Power System Analysis)[36] is an advanced software toolbox specifically designed for simulating and optimizing modern electrical power systems. PyPSA optimizes the capacities of these components across time periods, considering factors like electricity demand, renewable energy generation, and energy prices. This comprehensive modeling enables informed decision-making and efficient system operation. PyPSA's capacity modeling capabilities address the challenges posed by renewable energy sources’ variability by modeling their capacities and optimizing the overall power system capacity. It also allows analyzing the influence of transmission lines on power systems, enhancing power transmission efficiency and reliability. PyPSA enables users to optimize the capacity mix, promoting the integration of renewables, and improving the reliability and effectiveness of power systems.[33]
While PyPSA is designed for modeling power systems, Neumann et al.[17] utilized an open capacity expansion model called PyPSA-Eur-Sec,[37] which stands out from previous studies due to its comprehensive approach that considers multiple sectors, high spatiotemporal resolution, and multicarrier transmission infrastructure representation. This model allows to address the various transport bottlenecks that hinder the cost-effective integration of variable renewable energy sources. By co-optimizing the investment and operation of generation, storage, conversion, and transmission infrastructures, the model aims to achieve the least-cost outcome through a single linear optimization problem. The model accounts for energy flows between different carriers using various technologies, such as heat pumps, combined heat and power plants, thermal storage, electric vehicles, batteries, power-to-X processes, fuel cells, and underground hydrogen storage potentials. Additionally, the model incorporates data on electricity and gas transmission infrastructure to determine grid expansion needs and retrofitting potentials.[17]
3.3.2 Fluid Dynamic Modeling
In contrast to pure capacity modeling as sketched in the previous section, here, several approaches will be described to cope with the task of physical fluid dynamic modeling in the context of large gas networks. So, not only the amounts in terms of volume or energy are computed but also changes in pressure, density, temperature, velocity, gas mixture, etc. All these approaches have in common that the most general form of modeling consists of systems of partial differential equations (PDEs), which are called Euler equations in literature.[38-40] The complexity of the problem and with this as well the computing time can be reduced by assuming steady-state conditions. In the steady-state problem, all time derivates equal zero and the system of PDEs is reduced to a system of ordinary differential equations.
The system of equations of the Euler equations is underdetermined and must be completed problem specifically. In the case of gas network modeling, the equation of state of real gases is generally used for this purpose.[38-40] Together with algebraic (nonlinear) equations to describe components like compressor stations, a so-called differential–algebraic equation systems are formed.
For the modeling engineer dealing with gas network planning tasks, there are several problems to be tackled: 1) To prove that certain amounts of gas can be transported (at the same time) under defined pressure and temperature conditions from several entry points to many exit points (steady-state problem): state-of-the-art for this kind of problem is to use steady-state models. To prove that a certain amount of gas can be transported without violation of the network constraints, suitable settings for the active elements in the network, which are compressor stations and valves, must be found. This is usually done manually in iterative simulations. This procedure requires a lot of experience and time. Therefore, Koch et al. proposed an automation of this process by means of operation optimization (see c));[31] 2) To solve time-dependent tasks (transient problem): in steady-state problems, the entry and exit mass flows of each pipeline are by definition equal. In the intraday operation of gas networks, however, it may happen that the incoming and outgoing mass flow are explicitly different. This can either happen unintentionally due to delays caused by the limited reaction time of the active elements (compressors and valves) and the limited transport velocity of the gas flow, or it can happen intentionally, e.g., if the internal storage capacity of the network is to be utilized and the pressure in a section is to be temporarily increased or reduced.[41] To capture these effects, transient simulation is necessary. This requires time-resolved contiguous data for the boundary condition as well as knowledge of the initial conditions.[31] To solve the PDE system, a spatial and temporal discretization must be performed. The discretization method must be chosen carefully, since it directly influences the results;[42] and 3) To optimize the possible gas flow in task 1) or 2) under certain constraints (optimization problem): it must be distinguished whether the task is to optimize the topology (network expansion planning) or whether gas flows in a fixed topology shall be optimized (optimization of operation). Network expansion planning is usually done using a steady-state description of the gas flow with the aim of minimizing investment costs. Optimization variables are the investment decisions in individual pipelines and units. The operational optimization of gas flows in a fixed topology can be performed with a steady-state as well as with a transient simulation. Different objectives can be achieved by choosing setpoints for the compressors and valves (so-called active elements). Ríos–Mercado and Borraz–Sánches group those objectives into linepacking tasks, pooling tasks, and minimization of fuel or energy costs for the compressors.[43] The main goal of the optimization proposed in ref. 31 is to find one feasible solution that does not violate the network constraints (validation of nominations problem). In this case, the choice of the objective function is of secondary importance.
In the following, we will describe and compare three software tools dealing with several of the aforementioned problems. The well-known gas network simulation tool SIMONE[44] was designed to solve the steady-state problem by numerical methods from the transient area. The advantage of this approach is that only one numerical kernel is needed to cope with both steady-state and transient problems. A drawback in this method lies in the fact that in some cases no solution for the steady-state problem exists and the user will get no hint what to do to make this problem solvable.
The software MYNTS[34] uses two different approaches for steady-state and transient problems. Here, the stationary method also can cope with cases where no physical solution exists. In such cases (e.g., if not enough pressure was given initially), the result will show regions of negative pressure, which is physically impossible but indicates in this context that in such regions the pressure or the volume flow must be increased.
The software framework pandapipes[45] is an open-source software tool specifically designed for grid and network modeling, offering capabilities for simulating and analyzing various aspects of energy networks, including natural gas, hydrogen, and heat networks. It provides functionalities for network topology creation and hydraulic and thermal calculations. Until now, the functionalities have been focused on steady-state solutions. Since the abovementioned optimization problem is often unsolvable due to the complexity of mixed-integer problems,[30-32] pandapipes offers a simplified modeling technique to speed up the computing time for a single network simulation. Thus, the software is able to run a larger stack of batch routines for a Monte-Carlo-like optimization of selected parameters. One important feature is the possibility of the data structure to add costs to each single network asset. So, it is possible to optimize a network based on investment, operational costs, or total costs—in particular with respect to the cost difference between repurposed and newly built hydrogen pipelines.
Table 1 shows an overview of the general advantages and disadvantages of fluid-dynamic models compared to regions/capacities models. A more detailed comparison can be found in the appendix.
| Modeling approach | Regions/capacities models | Fluid-dynamic models |
|---|---|---|
| Long-term applications | Yes, including large time series | For limited network usage cases only |
| Medium-term applications | Yes, including large time series | For limited network usage cases only |
| Short-term applications | Yes, without dynamic effects | Yes |
| Transient modeling | No | Yes |
| Sector integration | Gas/electric/heat/water/biomass | Gas/hydrogen/electricity/heat/water/CO2 |
| Accessibility | Open Source | By license |
| Computational time | Low | Medium-to-high |
| Advantages | Flexible and modular | Captures nonlinear effects of fluid dynamics |
| Disadvantages | No nonlinear effects of fluid dynamics | Requires high computational time |
| Manual adjustments necessary |
3.4 Use and Interplay of Scenario Data Sphere, Network Topology Sphere, and Network Simulation Sphere
The different data spheres of the network modeling systematics—the scenario data sphere for the tasks for the network (scenarios, network usage cases, etc.) on the one hand, and the network topology sphere for the network itself (topology, elements, equipment, settings, and configurations) on the other hand, can combine in different ways and directions. The differentiation into the different spheres is necessary, and the clear assignment to the spheres and documentation in the tools described above is helpful for the efficient design and execution of research projects. This is developed in Table 2 for essential network-related research questions; more details, including predesigns of the procedures are given in the appendix and practical applications in Section 4.
| Category | Interchange between spheres | Application |
|---|---|---|
| Network adequacy | ||
| → Can the network fulfill changing requirements? | Scenario data change → network topology remains unchanged | Regions/Capacities or Fluid-Dynamic Case Security of Supply (SoS) France, Section 4.2 |
| Bottleneck analysis | ||
| → Which network elements limit the network flows? | Scenario data change → network topology remains unchanged | Fluid-Dynamic Context Case Distribution, Section 4.1 System Development, Section 4.3 |
| Capacity assessment | ||
| → To what extent can flow be increased? | Network topology unchanged → scenario data are maximized | Regions/Capacities or Fluid-Dynamic Context Case SoS France, Section 4.2 System Development, Section 4.3 |
| Expansion planning | ||
| → Which measures are adequate? | Scenario data are given ↔ network topology changes are analyzed | Regions/Capacities or Fluid-DynamicCase SoS France, Section 4.2 System Development, Section 4.3 |
| Network repurposing | ||
| → Which parts of natural gas networks to repurpose for hydrogen? | Scenario data are given ↔ network topology changes are analyzed | Fluid-Dynamic Partly Case Distribution, Section 4.1 System Development, Section 4.3 |
4 Applications
4.1 The Future of Gas Distribution Networks
The grid simulation framework pandapipes in combination with network and scenario data offers a valuable application in the context of transitioning from a natural gas-based distribution system to a pipeline-based hydrogen infrastructure in the distribution network sector. The overall sense of this idea is still under debate. A scenario analysis published by the German Ministry of Economics concludes that hydrogen, power-to-gas (P2G), and power-to-liquid (PtL) will play no or only a very minor role in meeting heat demand in the building sector. However, due to the current connection of about 50% of the building stock in Germany to the gas grid, the possibility of using hydrogen as a gaseous energy carrier should not be disregarded.[15] So, although hydrogen should not play a relevant role in the building sector, its partial use, for example, as a transition technology, must be clarified regionally. Here, various factors have a decisive influence on whether hydrogen is a realistic alternative for meeting heat demand in the building sector. Besides the technoeconomic aspects, there must be sufficient availability of hydrogen, both nationally and internationally, and skilled personnel for necessary conversion measures within the gas infrastructure must be available in a timely manner. The corresponding monetary investments must be made for the provision of hydrogen and retrofitting measures.
A technoeconomic analysis is presented in ref. 46. To conduct an analysis of the transition from gas distribution networks to hydrogen, several important steps including data fusion and grid simulation need to be taken.
4.1.1 Data Compilation for Scenario Data Sphere and Network Topology Sphere
A gas network dataset is generated, e.g., by georeferencing a gas network map provided by the network services operator. Georeferencing involves assigning spatial coordinates to the gas network map, allowing it to be accurately aligned with real-world geographic locations. This process enables the integration of the gas network map with other spatial data and facilitates subsequent analysis and modeling tasks: 1) Obtain data on the distribution of buildings with gas grid connection for the years 2020 to 2045: this information is crucial for understanding the existing infrastructure and identifying areas where the transition to hydrogen may be required; 2) Model the distribution of annual energy consumption for space heating and hot water among buildings: this step involves analyzing the energy consumption patterns of different buildings and determining the demand for heating and hot water services in each area; and 3) Define the distribution of building types: this includes categorizing buildings based on their characteristics, such as residential, commercial, or industrial, as well as their size and energy needs. This information helps in assessing the diversity and scale of energy demand across the distribution network.
4.1.2 Applications in the Network Simulation Sphere
The gas mass flow within the gas pipelines is determined using pandapipes. By simulating the behavior of the gas distribution network, pandapipes enables the calculation of gas flow rates, pressures, and other relevant parameters. This step is crucial for evaluating the feasibility and efficiency of hydrogen distribution, as well as identifying potential bottlenecks or areas that require infrastructure upgrades.
One of the scenarios which can be analyzed is the complete conversion of the natural gas grid in 2035 to 100 vol% in a part in a German City.
4.1.3 Results Documentation and Interpretation—Conversion Costs of the Distribution Network
The cost of retrofitting for a hydrogen operation is determined by summing the individual cost involved. In-service gas pipelines and pressure reduction and metering stations were determined by gas flow simulation in pandapipes.
A distinction was made between costs incurred by the distribution system operator and those borne by the end users. It was assumed that the distribution system operator bears the costs to retrofit gas network elements that are located outside buildings.
The system operator is responsible for gas pipeline lines, valves for interrupting the gas flow (gate valves), and gas network components in gas technical installations for pressure regulation and gas measurement.
The end users bear the costs of gas network elements, which are located inside the buildings. These include gas network components in the building installation area as well as gas combustion equipment.
Table 3 shows the costs incurred for the complete conversion to 100 vol% hydrogen in 2035. Costs are incurred in the area of gas-technical systems for pressure control and gas measurement, which are allocated to the distribution network operator. In addition, costs are incurred in the area of building installation and for gas combustion systems, which are allocated to the end users. The list of total costs over the years to 2035 is shown in Table 4.
| Costs type | In operation | Cost in € |
|---|---|---|
| Pressure control system | 3 | 89 820 |
| House connection | 905 | 235 300 |
| Gas combustion system | 1,813.01 | 13 960 177 |
| Cost for distribution grid operator | Cost for end user | Cost per dwelling unit | |
|---|---|---|---|
| 2035 | 89 820 | 14 195 477 | 7,829.78 |
| Total cost in € | 14 285 297 | ||
The study shows that utilizing pandapipes, researchers and planners can analyze the feasibility, performance, and potential challenges associated with the conversion of gas networks to a hydrogen distribution system, supporting informed decision-making and facilitating a smooth transition toward a sustainable and low-carbon energy future. More use cases and assumptions can be read in ref. 46.
In the context of future energy network planning, it becomes increasingly important to couple gas distribution network simulations with other energy network simulations, such as electricity and heat networks, particularly for investment planning purposes. This integration allows for an analysis of different energy technologies and their impacts on multiple networks. One such application is the evaluation of heating technologies for future energy systems. In a previous study conducted in 2019,[47] we compared a gas-based heating system with an electricity-based heating system, considering not only the investment and operational costs for households but also the costs associated with the electricity and gas networks. This analysis provided valuable insights into the economic feasibility and network implications of different heating options. The integration of these simulations can be facilitated by utilizing the “multinet” module of pandapipes, which enables the simultaneous modeling and analysis of multiple energy networks.[48, 49] The investment and operational costs for households and network infrastructure can be derived.
In conclusion, pandapipes, with its versatile module multinet, offers the capability to simulate integrated energy infrastructures encompassing gas, electricity, and heating networks. This enables researchers and practitioners to conduct comprehensive technoeconomic analyses for future coupled networks, considering the interdependencies and synergies between different energy systems. By providing a holistic modeling approach, pandapipes empowers decision-makers to explore optimal strategies and make informed choices toward a sustainable and efficient energy transition.
4.2 Natural Gas Security of Supply France
Network modeling and simulation are not restricted to planning tasks. It can be utilized to answer the question of supply security, as demonstrated by the authors in ref. 50. In this study, the urgent question was answered, how to reverse the natural gas flow in Europe after the stop of Russian gas deliveries in 2022. Here, again the hierarchical modeling approach showed its strength: starting with a simplified European flow model it could be shown that it is not sufficient just to add up the liquified natural gas (LNG) capacities in European ports.
The flow model gives valuable hints at which places the main flow direction should be reversed, especially at interconnectors between certain countries. Reversing the flow at compressor stations often needs changes in the infrastructure, but with significantly lower investments compared to a new pipeline construction. The next deeper level in the hierarchical method resulted in refined models for selected countries in Europe to make sure that the transport task that had been estimated on the higher level could also be fulfilled by each country's infrastructure. Figure 4 (hierarchical modeling; left: France detailed, right: European coarse grid.[34]) displays the hierarchical structure of the grid modeling techniques used for the analysis.
4.2.1 Data Compilation for Scenario Data Sphere and Network Topology Sphere
The most important topology data for the simplified European flow model are the interconnector capacities between the countries and the LNG capacities of the individual countries. For the detailed analysis of the individual countries, e.g., France as presented here, data on parallel pipelines, pipeline diameters, and on the locations of compressor stations are required to be able to estimate the available transport capacities as accurate as possible. For this, the data basis of SciGRID_gas[51] was supplemented and improved by detailed research for the individual countries. The locations of natural gas storage facilities were also included in this step.
Two different scenarios with network usage cases for the years 2023–2026 were considered in the study: 1) a base case with an average hourly demand and no natural gas injection or withdrawal from the storages on the basis of the data from the TYNDP scenario “Distributed Energy”;[6] and 2) a winter case, which assumes a day in a 2 week cold period with low availability of wind and solar generation on the basis of the data from the TYNDP scenario “Distributed Energy”.[6] In this case, the underground gas storages are emptied. Furthermore, analyses have shown that filling the storages in summer does not place higher stress on the network than the case in winter. Thus, the winter scenario was investigated as the peak load scenario.
The primary objective of the France case study was to show that the interconnector capacities can also be utilized in the reverse flow direction (This was examined from a network capacity perspective. Independent of this is the technical conversion of the interconnectors, which in most cases requires a longer construction project.) (from West to East), and thus that the French LNG terminals can contribute to security of supply in Germany and beyond in Eastern Europe. First, however, it must be shown that security of supply can be guaranteed in France itself (and in the other countries considered in the study). To this end, natural gas demand data on country level from ref. 6 was disaggregated to NUTS 3 level. The data were split into industry and household demand and distributed to the regions proportionally using data from Eurostat.[52] The NUTS3 data were then mapped to the individual network nodes using a Voronoi polygon-based algorithm.[53, 54]
4.2.2 Applications in the Network Simulation Sphere
For the analysis, the authors followed a three-step approach[55]: 1) A regions/capacities model, which uses a pure capacity calculation to optimize the gas flows via the European interconnectors. The model uses a comprehensive approach, encompassing multiple sectors and multicarrier transmission infrastructure representation, addressing transport bottlenecks for cost-effective integration of variable renewable energy sources by co-optimizing generation, storage, conversion, and transmission infrastructures through a single linear optimization problem that incorporates various energy technologies and considers electricity and gas transmission infrastructure data for grid expansion and retrofitting, see ref. 17, Section 3.2.1; 2) In parallel to this and with the same granularity, a European flow model, which is based on a fluid mechanics model, is utilized for validation of the results of the linear model under 1) above; the model is based on nonlinear flow equations, but due to the strong spatial aggregation, here, it takes over the task of a capacity model. The European flow model in the software MYNTS is not used for optimization here. Instead, under given boundary conditions, gas flow corridors and possible bottlenecks can be visualized in the simulation. Thus, in addition to the regions/capacities model under 1), by visualizing the bottlenecks and flow directions, it can be determined which are the most important interconnectors where reverse flow technical capability should be retrofitted and which countries are of particular interest for detailed analysis; and 3) A detailed steady-state fluid dynamic simulation of individual countries. This should validate whether the interconnector flows from the upstream models can be served from the country networks. Otherwise, bottlenecks within countries can be identified. The detailed analysis is done with the software MYNTS and SIMONE. In both tools, a so-called manual adjustment has to be made to find a feasible solution.
4.2.3 Results Documentation and Interpretation
All calculations in the three model layers were carried out for a data set for 1 h of the year, representing one specific network usage case. The identified deficits, indicating threats for security of supply for years before 2025, were therefore identified in MWh/h or MW. The results comparison between the three model approaches showed a high comparability.
4.3 System Development Strategy
The impact of political guidelines and decisions on network topologies and their immanent elements should not be underestimated. In particular, import options and routes, the locations of electrolyzes, re-electrification power plants, and hydrogen storage facilities are defining for network topologies.
Considering the enormous transformations that need to be undertaken to achieve carbon neutrality, it is essential that these are designed efficiently. Furthermore, the transition of the energy system requires significant investments, involving extensive planning, implementation, and long life cycles, particularly in terms of infrastructure. For that purpose, the German Federal Ministry of Economics and Climate Protection has launched the so-called system development strategy. The ongoing development of this policy strategy represents a crucial initiative aimed at effectively coordinating the planning processes for gas infrastructures and aligning them with strategies related to diverse sectors and energy carriers impacted by these infrastructures. The objective of conducting scenario-based analyses is to assess multiple factors that impact the entire system and analyze their effects on the demand sectors, with the aim of identifying synergies and key elements.
Through the examination of extreme cases, the scenarios explored in energy system analysis thus far demonstrate the potential possibilities and concurrently highlight the intricacy involved in the gradual conversion of the German natural gas grid to hydrogen as part of the energy transition. These scenarios provide insights into what can be achieved while acknowledging the complex nature of this undertaking. Furthermore, this enables the coverage of a robust bandwidth to achieve the highest possible degree of resilience. The conducted analyses demonstrate that the precision and relevance of the results are highly reliant on the level of detail within the temporal and spatial dimensions.
However, given that the temporal and spatial framework conditions are already undergoing various transformations in the context of the energy system transition toward climate neutrality, it becomes imperative to consider numerous additional changes or alternative developments across different scenarios for future energy system analysis modeling tasks. This recognition emphasizes the need to account for a broader range of factors and potential future developments to ensure comprehensive and accurate energy system analyses going forward. In a similar vein, future analyses should incorporate additional dimensions of data collection, as outlined in ref. 56 for the German gas transmission network.
5 Conclusion and Recommendations
The network modeling systematics analyses the specific complexities of network modeling challenges alongside the dimensions time, space, and network resolution/accuracy and presents a systematic approach to structure the broad range of research questions and political and economic decisions problems into a coherent use and interplay of separate spheres for scenario data, network topology, and network simulation. The applicability and reasonable combination of two basic approaches for the network simulations, regions/capacities models and fluid-dynamic models, are shown.
Embedded in general energy systems analyses, network modeling requires generally a far more disaggregated data resolution in terms of time and space. Their bidirectional aggregations and disaggregations are well-established and can be managed consistently in the scenario data sphere and the network topology sphere. Interlinked with this, but to be managed separately, the network simulations sphere covers the task and data management challenges that are specific for network modeling in contrast to general energy systems analyses.
This network simulation sphere comprises the interfaces to the scenario and topology spheres; in addition, it contains one or several approaches of network simulation, regions/capacities models, and fluid-dynamic models. Both approaches show specific advantages and disadvantages which can be summarized between the poles of technophysical accuracy versus applicability for large time series and network systems. The combination of both, high accuracy and large systems and time frames, exceeds computational and expert capacities and resources. Thus, the larger the systems, the more recommendable are regions/capacities models—at the expense of accuracy; the higher the accuracy requirements, the more fluid-dynamic modeling becomes unavoidable.
From fully detailed fluid-dynamic models to high levels of aggregation, there exist interim levels that are described as layers of network resolution/accuracy. To compensate for and overcome the accuracy limitations for the modeling of large network systems over large time frames, the hierarchical iteration has been developed and applied; it combines fluid-dynamic and simplified modeling approaches. While this does not yield exact optimization results, the heuristic compilation of results can achieve an approximation. It provides feasible and viable solutions which might be suboptimal but guarantee—in contrast to only regions/capacities modeling—the technophysical applicability of the developed solutions in terms of demand satisfaction and security of supply.
Among a variety of software tools for the model implementation, PyPSA for the regions/capacity's models and MYNTS and pandapipes for the fluid dynamic models have been applied with particular relevance for the cases presented, in addition, to SIMONE. Strengths of PyPSA and pandapipes are their open-source availability and adaptability. Weaknesses are the missing fluid-dynamic accuracy (PyPSA) and limited user interface availability. Strengths of MYNTS are its high accuracy for fluid dynamics and multisectoral coverage, weaknesses are the high computations time and necessary manual adjustments.
Independent from the utilization of a specific fluid dynamic network modeling software, the applicability and possible combination of regions/capacities models and fluid dynamic models can be attributed to typical research questions. The recommended combinations have been successfully applied for investigating and assessing the network adequacy of an existing network or a given network development plan (i.e., can the network fulfill the requirements of the network task under changing conditions?) for the assessment of natural gas security of supply in France; in that same application also for expansion planning (i.e., which elements of the modeled network limit the maximum entry and/or exit flows?); and for the assessment of natural gas network repurposing for hydrogen for a municipal distribution network.
The natural gas security of supply analysis for France has also successfully proven the applicability of hierarchical iterations with the combination of a regions/capacities model with the fluid-dynamic network simulation based on MYNTS.
These achievements pave the path for future more widespread applications of the network modeling systematics. This is particularly promising in the framework of system development strategies for electricity, natural gas, and hydrogen transmission systems. These strategies will, by definition, require long-term assessments for strongly varying scenario conditions due to interactions and repercussions between the infrastructures for different energy carriers (to be managed in the scenario data sphere). They also require a high temporal and spatial resolution for investment decisions (to be managed in the network topology sphere) and, finally, a high technophysical accuracy (to be achieved by adequate layering and hierarchical iterations in the network simulation sphere). Such system development strategies are under discussion on the European level and currently entering into binding legislation in the German energy Act. This provides a strong legal basis for the outlook of this perspectives article.
Acknowledgements
The authors gratefully acknowledge the financial support by the Federal Ministry of Education and Research (Bundesministerium für Bildung und Forschung, BMBF), grant number 03HY201S, as part of the public research project “TransHyDE.”
Open Access funding enabled and organized by Projekt DEAL.
Conflict of Interest
The authors declare no conflict of interest.
Author Contributions
J.M.-K.: Conceptualization, methodology, investigation, writing (original draft), writing (review and editing), visualization, supervision, and project administration; M.R.: Investigation, writing (original draft), supervision, and funding acquisition; T.K.: Investigation, writing (original draft), and visualization; B.K.: Investigation, writing (original draft), and visualization; T.M.: Investigation, writing (original draft), writing (review and editing), visualization, and project administration; U.H.: Investigation, writing (original draft), and visualization.
Appendix
| Features of the modeling software | Regions/capacities model, e.g., PyPSA | Fluid-dynamic model, e.g., MYNTS |
|---|---|---|
| Suitability for long-term applications | Yes, including large time series | Yes, for limited network usage cases |
| Suitability for medium-term applications | Yes, including large time series | Yes, for limited network usage cases |
| Suitability for short-term applications | Yes, without transient effects | Yes |
| Suitability for transient modeling | No | Yes |
| Properties | Open-source Python-based optimization model | Advanced physical modeling of gas/hydrogen |
| Sector coupling possible (gas/electric/heat/water/biomass) | Sector coupling possible (gas/electric/heat/water) | |
| Liquid CO2 transport with phase changes | ||
| Accessibility | Open source | By license |
| Computational platform | PC (Win/Linux) | PC (Win/Linux) |
| Computational time | Low | Medium to high, additional manual adjustments required |
| Advantages | Designed to be flexible and modular | Hierarchical modeling in form of subnets |
| Interoperability, interface with other tools possible | Captures nonlinear effects of fluid dynamics | |
| Optimization functionalities included | Advanced graphical editor | |
| Several numerical solvers | ||
| Disadvantages | Does not capture nonlinear effects of fluid dynamics | Requires high computational time |
| Limited user interface | Manual adjustments necessary for many Network Usage Cases |
1 Use and Interplay of Scenario Data Sphere, Network Topology Sphere, and Network Simulation Sphere
1.1 Adequacy of the Existing Network or a Given Network Development Plan → Can the Network Fulfill the Requirements of the Network Task under Changing Conditions?
The changing circumstances can be manifold, we give some examples: 1) Short term—can the existing network, including the storages with given filling levels, comply with the network usage tasks that arise from unchanged demand? 2) Mid term—can the existing network comply with increasing demand of gas-fired power plants? 3) Long term—is the planned network able to cover the network usage cases for different scenarios, e.g., strong increases in hydrogen demand?
The data changes for these research questions primarily take place in the scenario data sphere only, the topology remains unchanged. The different tasks and the results for their utilization/testing in network model runs are documented and analyzed with the results for the respective runs; possible outcomes are the positive/negative answer and/or changing values for pressure, flow rate, and volume flow conditions.
1.2 Bottleneck Analysis for an Existing or Planned Network → Which Elements of the Modeled Network Limit the Maximum Entry and/or Exit Flows?
Real-world applications for this generic research question can relate to areas, sources, borders, or specific points, as the following examples show: 1) What are the limiting elements for the changing of flow direction (reverse flow) from East → West to West → East to accommodate for new imports? 2) What are the limitations for maximum utilization of gas withdrawal from underground storages for hydrogen—and how many are there? 3) What are the limiting elements for the import of hydrogen from different directions under the assumption of full “repurposability” of the existing natural gas network?
These analyses do not require changes in the used network topology, just like in the cases under Section A1.1; but they require one additional step in the analysis of the results of the network simulation results, and that is the examination in which parts of the network topology and its elements’ technical limitations were reached or overstepped.
1.3 Capacity Assessment for an Existing or Planned Network → to what Extent Is It Possible to Increase or Decrease the Entry and Exit Flows?
For these assessments, the logical direction of the analysis is reverted, even though the workflow through the procedural steps remains the same. Here, we still change the numerical values for the network usage cases, then run the network simulation and check for the results to reach the limitations—but now not with an attempt to most accurately represent current or future tasks for the network in the network usage cases. Instead, incremental changes to the entry and exit values are applied systematically with the intent to reach the limitations and track them back to the applied extreme values for the entry and exit values—which then can be understood as the capacities of the analyzed topology.
There is no biunique definition of how to increase/decrease the flow values as there are several degrees of freedom for their distribution. Increases, e.g., can be modeled for individual, clustered, or widely distributed entry point as well as exit points (entry and exit changes overall must be balanced). However, with systematic patterns, interesting insight can be generated, which we illustrate with two examples: 1) Uniform increases in percentage or absolute terms to all entry points and all exit points lead to an assessment of the overall network capacity; 2) Increases for those entry points related to the same direction of origination, e.g., the same export country, give an assessment for the import capacity in terms of resilience for different sources’ distributions; and 3) Increasing, in addition to the last point, the exit values in a regional concentration toward the opposite direction of the network construes a stress test for finding out the maximum throughput capacity of the network.
1.4 Expansion Planning to Cope with Higher or Changed Flows → Which Are Adequate Measures for the Expansion of a Network?
This application adds changes in the network topology sphere, too. The bottlenecks first determined in analyses according to Section A1.2, or the network limitations that turn out to define the capacity limit according to Section A1.3, are addressed by changeovers in the network's topology, like, e.g., additional pipelines, replacement of pipeline segments with larger diameters and/or installation of additional or more powerful compressor stations. These alternatives and choices can be integrated into the network topology data and then be tested for the fulfillment of the network tasks. Among those options that cope with the network usage cases and thus overcome the bottlenecks or limitations, based on technoeconomic comparisons the most efficient network expansion measures can be determined.
1.5 Repurposing of Natural Gas for Hydrogen Transport → Which Segments and Subnetworks Can and Should Be Repurposed?
This research question is more complex as it requires a more intrusive and repeated alteration of the topology database, notwithstanding complex analyses being necessary in the scenario sphere as well. The stepwise approach for the network repurposing, including a split of the existing infrastructure for the transition period, has to consider: 1) Spatial and temporal development of hydrogen demand and natural gas demand; 2) Availability of network segments that can be taken out of the natural gas network and are suitable for hydrogen transport; 3) Combination of the segments into a hydrogen network; 4) Maintaining a fully functional; and 5) Expansion of the hydrogen network over time.
With a view to the first point, for the early phase of hydrogen network development, the location of large industrial hydrogen consumption is a prime driver. In addition, and with increasing relevance over time, the sources and their internal as well as external distribution define the shape and capacity of the network. Approaching higher levels of decarbonization, storage locations and capacities as well as the power plants for re-electrification become dominant. On the side of the natural gas transmission system, in the starting phase, the restriction for the release of element for hydrogen transport is the highest and sees some relaxation over time.
Thus, a close interchange of data between the scenario data sphere and the network topology sphere is of utmost importance for the analysis of the split of natural gas and hydrogen networks.
