In 2018, India’s annual total emissions were approximately 3 GtCO₂ (gigatonnes of CO₂), of which around 73 per cent (2.1 Gt) were emitted by power plants and the industrial sector (including utility electricity generation, industries, captive power plants, and industrial process and product use emissions) (GHG Platform India 2018). Significant mitigation efforts will be required to limit the global temperature increase to 1.5°C and to meet India’s net-zero target by 2070. While India continues to promote renewable and alternative energy use to meet the growing energy demand in power systems and industries, legacy capacity in industries and the power sector that rely on fossil fuels cannot be easily abandoned before the end of their lifespans. As a result, fossil fuels will need to remain an integral part of India’s energy economy for several decades before the transition to green energy is complete.

To achieve net-zero by 2070, carbon capture and storage (CCS) will be indispensable. It is estimated that CCS can reduce economic losses by 23 per cent between 2030 and 2050 and by 32 per cent between 2050 and 2100 by reducing the pace of deployment of expensive low-carbon technologies and preventing the abandonment of operating assets (Chaturvedi and Malyan 2022). Additionally, the rate of increase in average power generation costs is lower in energy system modelling scenarios with CCS than in other energy modelling scenarios because CCS does not require changes to legacy fossil fuel–based capacity (Chaturvedi and Malyan 2022). To meet the 1.5°C temperature increase cap, India may have to inject 5.3–10 GtCO₂ by 2050 (Chaturvedi and Malyan 2022; Gambhir et al. 2021; Garg et al. 2017; Vishal et al. 2021). In addition to energy efficiency, renewable power, and alternative fuels, CCS could help mitigate up to 57 per cent of emissions from hard-to-abate sectors, such as steel, cement, and aluminium, because it is a significantly lower-cost alternative to carbon capture and utilisation (Elango et al. 2023; Nitturu et al. 2023).

India has a large storage potential of 317 GtCO₂, of which 144 Gt are in saline formations, 170 Gt in basalt, and the remaining in oil and gas and coal fields (Bakshi et al. 2023). CO₂ storage in basalt formations is unique because the injected CO₂ reacts with the basalt to form solid mineral carbonates over time, thereby eliminating the risk of long-term leakage. Since these geological formations can be far from major CO₂-emitting sources, a CO₂ transport network is needed. There are three major ways of transporting CO₂: by trucks, trains, or pipelines. The choice of transport mode depends on the distance, the quantity of CO₂ to be transported, the duration and frequency of transport, and whether the start and end points are stationary or mobile. Since our focus is on the long-term transport of large quantities of CO₂ from stationary sources to stationary CO₂ sinks, pipelines are the most economical choice in our analysis (Lu et al. 2020).

Studies on the design of an optimised CO₂ pipeline network typically recommend the hub-and-spoke model, a logistics strategy in which centralised collection at a specific place, or ‘hub’, serves as a focal point for redistribution to different locations, or ‘spokes’. This model helps manage and optimise network flow efficiently. However, there are two challenges with the hub-and-spoke model in the Indian context. First, Indian industrial and power generation facilities are distributed across the country, unlike in smaller countries, where they are concentrated in certain industrial clusters. This makes identification of hubs to support an industrial cluster difficult and possibly unnecessary. Second, in the context of optimising pipeline networks using a hub-and-spoke model, right-of-way (RoW) issues are not considered. RoW refers to the legal permission to lay and operate a pipeline through third-party-owned land. RoW can pose significant challenges, including land acquisition, compensation disputes, regulatory compliance and interference with existing infrastructure.

Pipeline construction projects in India are often significantly delayed because of RoW-related issues (Business Standard 2024; Department of Economic Affairs 2020; GlobalData 2024; Tewari 2024). Obtaining the necessary permissions to lay pipelines through forest areas, land demarcated for defence, coastal regulation zones, and government-owned lands is a complex and time-consuming process that often takes a year or more. There have been several instances in which pipeline construction had to stop due to RoW issues, sometimes delaying the project by many years, resulting in increased project costs and lost revenues (Economic Times 2020). To address such issues, in this study, we suggest an alternative conceptual approach in which we consider laying trunk CO₂ pipelines parallel to existing natural gas (NG) pipelines with RoW. Logically, this should reduce the time required to lay new pipelines and, hence, the overall cost of such projects. However, RoW will be necessary for spur pipelines that connect the source to the trunk line and the trunk line to the sink. This paper focuses on national-level network optimisation and estimates the costs and infrastructure required to transport CO₂ for CCS in India, using the RoW of an existing natural gas pipeline network to build new trunk CO₂ pipelines.

We suggest an approach of laying CO₂ pipelines along existing natural gas pipelines to limit land acquisition challenges and delays.

To evaluate the proposed CO₂ transport pipeline network, we leveraged the RoW for the existing and under-construction NG pipeline network. The analysis excludes crude oil (~11,000 km), petroleum products, and liquified petroleum gas (LPG) pipelines (~23,000 km) (PPAC 2024). The optimal solutions for a CO₂ transport pipeline network may be shorter in length and thus cost less to build if crude oil, petroleum product, and LPG pipelines were included in the analysis. However, due to a lack of geographic information system (GIS) data and limited computing power, we included only the RoW of approximately 18,000 km of the NG network’s trunk pipelines, as shown in Figure 1. The 18,000 km network length indicates the route length (the RoW along the trunk pipelines), and not the length of the pipelines (which total to ~32,500 km, because, in many routes, there are multiple NG pipelines that run in parallel, such as the Hazira–Vijaipur–Jagdishpur pipeline section, where three pipes are laid in parallel).

We designed an optimised pipeline network for transporting CO₂ from 765 power and industrial plants to potential onshore underground storage sites.

We refer to all new CO₂ pipelines laid down using the RoW for the existing NG pipeline network as trunk pipelines (to mean the same as associated routes). We also analysed how much of the new CO₂ spur pipelines overlap with existing NG spur pipelines Spur pipelines carry CO₂ from the source to the trunk pipelines and from the trunk pipelines to the sinks.

Our model used the following four-step process to build the CO₂ pipeline network.

i. Find the shortest/cheapest path between each source–sink pair

First, we mapped the existing, under-construction, and approved NG pipeline routes on GIS software using data provided by Genesis Ray (2022). Before using this as an input in our model, the spur pipeline routes of the existing NG pipeline network were removed to reduce complexity and runtime. The resultant network (or graph data structure) consisted of linear splines with a total route length of approximately 18,000 km. To find the shortest possible path between each source and sink, we first defined a weighted graph data structure consisting of vertices and edges. The vertices were defined such that the route between any two adjacent vertices is a linear edge of length less than or equal to 100 km.

The values or weights of the edges were defined as follows: If two vertices are directly connected by an NG pipeline route, the weight of that edge is simply taken as the length of that route. Otherwise, the weight of the edge is equal to the Euclidean distance multiplied by the RoW factor. The RoW factor was introduced to discourage the model from building pipelines when an existing NG RoW is nearby. This factor needs to be large enough to maximise the selection of the existing RoW routes.

The existing NG trunk pipeline routes are not necessarily connected to the existing sources and future sinks. Therefore, the graph trunk pipeline routes were extended to include the sources and sinks, with new edges. The weights assigned to these new edges were also defined as the product of the Euclidean distance and the RoW factor, since they will all become new spur pipelines. As sinks are regions and not points, the distance from a vertex to a sink was defined as the shortest distance between the vertex and the sink boundary.

The weighted graph was used as the input for the Floyd–Warshall algorithm to compute the shortest distances (based on edge weights) between each source and sink. The Floyd–Warshall algorithm is one of the most efficient algorithms for finding the shortest paths between points in a weighted graph and is used to solve network and connectivity problems in which all the nodes in a network are connected. The algorithm output is a table that comprises the weighted lengths of the shortest paths between each source–sink pair.

ii. Optimally allocate the CO₂ flow rate

In this step, we used a linear programming-based model to connect each source to a sink (based on the shortest paths identified in Step 1, such that the average cost of CO₂ transport across the entire network was minimised). The model used the following inputs:

• A matrix containing the shortest/cheapest distances from each source to each sink for all source–sink combinations

• Yearly emissions from each source

• Storage potential of each sink

• Duration for which we need to optimise, i.e. the number of years over which CO₂ will be injected in the sink

Additional decision variables were included in the model to represent the CO₂ flow rate from each source to each sink. The model allocates a positive value to only one of the variables for each source– sink combination when an optimal solution was found, indicating that only one of the many possible connections was made between a source and a sink. An objective function subject to constraints allocated CO₂ from each source to a given sink based on the minimised cost of transport cost, as shown here:

Objective function

The objective was to minimise the total transport cost. Ideally, the pipelines would be represented as a number of discrete standard pipeline sizes. However, this type of representation would require a large number of continuous and binary variables and constraints in a mixed-integer linear programming formulation, significantly increasing the computational resources required. Instead, our function represented the pipelines as a single linear function (passing through the origin) of cost and pipeline capacity (flow rate) for a given length, ensuring that the cost estimate yielded by the model is realistic (Middleton et al. 2020). The cost of each pipeline was directly proportional to the product of flow rate and pipeline length. Thus, our objective function minimised the sum of the products of the flow rate and the weighted length for each source–sink connection, as described in Step 1. Mathematically, the objective function can be represented as:

Note that although we used a simplified pipeline cost function for optimisation, the actual pipeline cost function (without simplification) was used to calculate the final costs, as described in Step 4.

Constraints

• The CO₂ flowing out of a source should be equal to its emissions

• The total CO₂ flowing into a sink in a decided time period should be less than the capacity of that sink

where

life = assumed lifespan of the pipeline (25 years)

The output of this linear programming model provided an allocation of the CO₂ flow rate, which minimised the total cost of the pipeline network using the simplified cost function with a zero y-intercept. The output obtained in this step established an independent connection from each source to the chosen trunk pipeline and further to the sink through that trunk pipeline. To leverage economies of scale and reduce costs, it was practical to aggregate the flows going through overlapping or similar routes of spur pipelines. Hence, the solution obtained in this step was suboptimal.

iii. Minimise the spur pipeline route length

To address suboptimality and consolidate multiple spur pipelines into a single pipeline, we further reduced the total length of the spur pipelines using Kruskal’s algorithm, which is highly efficient when the number of splines is minimal/sparse (GeeksforGeeks 2023). In this step, we combined multiple spur pipelines in the same vicinity and connecting to the same trunk pipeline into a single spur pipeline. Therefore, the output was a reduced length of spur pipelines in the overall network.

iv. Identify the overlap of CO₂ spur pipelines with existing NG spur pipelines

Step 1 excluded the existing NG spur pipeline routes to keep the computation process manageable. However, the RoW of the existing NG spur pipelines would reduce the land access burden for the CO₂ 2. Average cost break-up provided by a large gas pipeline operator in India. spur pipelines, similar to trunk pipelines. Therefore, in this step, the new CO₂ spur pipelines were overlaid on the existing NG spur pipelines to identify overlaps. For those spur pipelines that overlap, we assumed zero RoW cost. The pipeline network obtained at this stage is the final network, for which the total cost was again calculated using the actual cost vs. flow rate function based on the FECM/NETL CO₂ Transport Cost Model developed by the National Energy Technology Laboratory (Morgan et al. 2023) instead of the simplified function used in Step 2. Since material costs are only one component of the total costs, total costs were assumed to be a product of the material costs based on the assumptions mentioned in Section 3.1. For the trunk pipelines, we assumed zero RoW costs.

The actual cost of the final pipeline network may not be the absolute minimum possible given the simplifications assumed, but it is a good enough approximation to guide the future efforts of all relevant stakeholders and policymakers.

We considered multiple scenarios to help relevant stakeholders make informed decisions regarding goals, planning, and research. Each scenario includes three variables: the sources (industries and power sector) considered, the type of sinks considered, and the constraint parameters that limit the feasible storage potential.

Industries considered

We considered plants from the following industries as sources:

• Aluminium

• Cement

• Fertiliser

• Iron and steel

• Power

• Refineries

Furthermore, industries were grouped based on their ability to reduce emissions using methods other than CCS and on the remaining life of existing plants. The following three source groupings were considered:

• Power plants (PP)

• Iron and steel and cement plants (ISC)

• All plants from all industries (All)

Power plants were considered separately because many may be shut down over time as the overall power system moves towards renewable energy sources. In such a scenario, dedicated trunk pipelines to potentially end-of-life power plants may not make sense. However, at this time, there is no clarity on how power plants will transition in the future; hence, we assumed the continuation of operations for the existing power plants. Iron and steel and cement plants were also considered separately because a significant portion of their process emissions cannot be reduced through other interventions, and legacy plants have significant residual life remaining. Thus, there are four possible choices when considering the sources:

• Power plants (PP)

• Iron and steel and cement plants (ISC)

• Power and ISC plants (ISC + PP)

• All plants (All)

Figure 2 presents the total emissions and geographical locations of all source categories considered in this analysis.

Type of sinks considered

We only considered onshore sinks in our analysis. Since onshore basalt and saline aquifers constitute approximately 99 per cent of India’s onshore storage capacity, only these two types of sinks were considered.

There are three possible choices in terms of sinks:

• Basalt only (170 Gt)

• Saline aquifers only (54 Gt)

• Basalt and saline aquifers (224 Gt)

Figure 3 presents the spatial distribution of these sinks (Bakshi et al. 2023).

Constraint parameters

Various above-ground challenges considerably influence the realisable underground storage potential. Population density is one such factor. The realisable storage potential increases when we allow more densely populated areas to be used as storage sites. According to a previous study by CEEW, there are two inflexion points on the curve of storage potential vs population density at 200 and 400 people/km2, where the area available and storage potential change substantially (Bakshi et al. 2023). Thus, for each of the three choices of sinks mentioned above, the storage potential at both population density levels was evaluated. Population density is a good indicator of local resistance and other above-ground challenges.

Population density is also a significant factor in the safe design and construction of pipelines. The ASME B31.8 Code for Pressure Piping (2003) is a commonly used standard for design, construction and maintenance requirements. The code defines ‘Location Class’ based on the number of buildings intended for human occupancy along any 1-mile section of a pipeline. Location Class 1 applies to a section that has 10 or fewer buildings and reflects areas, such as “wasteland, deserts, mountains, grazing land, farmland, and sparsely populated areas”. Location Class 2 applies to a section that has more than 10 but fewer than 46 buildings and reflects “fringe areas around cities and towns, industrial areas, etc.” Location Class 3 applies to a section that has 46 or more buildings and reflects areas, such as “suburban housing developments, shopping centres, residential areas, industrial areas, and other populated areas”. Location Class 4 includes “areas where multistorey buildings are prevalent, where traffic is heavy or dense, and where there may be numerous other utilities underground”.

The allowable pressure decreases with increasing location class; alternatively, pipe thickness must be increased for the same pressure with increasing location class. In the context of our analysis, storage sites may be possible only in Location Classes 1 and 2, which have a population density of approximately less than 200 people per km2 and Location Classes 2 and 3, which have a population density of approximately less than 400 people per km2 . In some instances, the pipeline network may pass through Location Class 4.

Considering all the possible combinations of these sources, sinks, and population density levels, 24 combinations can be created. We did not include location classes because of a lack of granular data at the sectional level of pipelines.

• Availability of RoW: We assumed that the existing pipeline network has sufficient space to lay a parallel CO₂ pipeline across the entire network. However, this may not hold in practice, and several sections may require additional RoW. Expanding the existing RoW may be easier than acquiring land for a new RoW. Nonetheless, our analysis indicated a shorter pipeline length and lower cost than is likely in the real world.

• Exclusion of crude oil and petroleum product pipelines: This analysis only relied on the RoW of NG pipelines. Adding crude oil and petroleum product pipelines would increase the total length of existing RoW, thereby reducing the network length for new CO₂ pipelines. Accordingly, the overall pipeline length was overestimated in our analysis.

• Euclidean distance of spur pipelines: We used the actual network pathways of existing NG pipelines. However, spur pipelines were estimated using straight-line Euclidean distance, which does not reflect real-world conditions. Hence, the spur pipeline lengths were underestimated in our analysis.

• Cost breakdown for the pipeline network: We relied on a rough breakdown of cost components (materials, laying, RoW, and miscellaneous). In practice, this breakdown will vary across different pipeline sections. We assumed that, on average, the costs would hold at the national level. However, we could not determine uncertainty due to the lack of granular, section-level data.

• Storage capacity of sinks: The CO₂ storage capacity of the sinks is an approximation, and there is inadequate detailed geological data to ascertain the exact storage potential. Hence, if there were a change in the storage capacity of the sinks, the pipeline network would also change; however, this uncertainty cannot be determined at this stage.

• Emissions trajectory: We used a static emissions profile for power plants and industrial facilities considered in this analysis. However, overall emissions are expected to grow over time. Depending on the change in location of emissions abatement (due to plant closures) or the emergence of new emission sources (due to new plant establishment or capacity expansion), the pipeline network would evolve over time. However, this aspect was not captured in our analysis.