Reducing stranded assets through early action in the Indian power sector

A policy database, called the Climate Policy Database (CPD) was used to implement national policies into global Integrated Assessment Models (IAMs) and national energy transition models. The database collects information on currently implemented policies⁴ related to climate change mitigation from countries worldwide (Climate Policy Database 2019). Planned policies are excluded from the database, with an exception of energy and GHG (Greenhouse Gas) emission targets announced as Intended Nationally Determined Contributions (INDCs) for the post-2020 period. Policies are considered only until the end of 2016. The duration of most policies in the NDC is 2030, with a few countries (e.g. USA) being up to 2025.

From the CPD, a set of core/high-impact polices were selected for each G20 country, including India (see list in SI section 4). The selection was made with national experts, with the objective of finding policies that would significantly impact GHG emissions.⁵

To be implementable in IAMs, policies were translated into policy outcome indicators, e.g. building standards were translated into final energy reductions in the building sector. Not all high-impact policies could be translated into policy indicators. For a description of how each policy indicator was implemented in each participating IAM see Roelfsema et al 2019 (in review)

The Early action scenario follows the currently implemented policies until 2020. Thereafter, it reaches for the prescribed carbon budget constraint, defined until 2100 for global models, and until 2050 for national models. While some models are cost-optimising, others simulate a carbon price, creating a response in the system that fits the carbon budgets. The global carbon budget is capped at 1000 Gt CO2 (2011–2100). Based on the latest assessment of the remaining carbon budget (Rogelj et al 2019),⁶ this figure represents more than 66% probability for a 2 °C warming but less than 50% probability for a 1.5 °C warming, thus falling within the definition of a well-below 2 °C target. The Delayed action scenario consists of currently implemented policies, as before, and additional pledges mentioned in the NDC until 2030. Thereafter, like early action, delayed action includes the carbon budget constraint, with both scenarios have the same carbon budget. However, unlike the global models, the two national models (AIM/Enduse and India MARKAL) have different targets, both of which are above the 2011–2050 budget observed for India in the global early action 2 °C scenarios. India MARKAL assumes a much higher GDP growth rate (see SI section 8) than AIM/Enduse and does not include CCS (carbon capture and storage), which leads to much higher baselines emissions and constrains how much decarbonization is possible. Furthermore, the scenario setup differs for national and global models and is shown in table 1. For a detailed methodology see SI section 5.

The models used in this study include six global Integrated Assessment Models (IAMs), which help in exploring interactions between the economy, land, and the energy system. They tend to be quite broad and include stylized and simplified representations of these subsystems (Rogelj et al 2018). These are: AIM V2.1, REMIND-MAgPIE, WITCH, IMAGE, GEM-E3 and POLES. Furthermore, two national energy system models are used—India MARKAL and AIM/Enduse (see SI section 6 for a description of each model).

Global models include inter-regional trade, the pace and cost dynamics of new technologies, and the link between the global economy with the global climate system. Most of them are technology rich, giving the energy system a variety of decarbonization options. On the other hand, national models can generally consider national circumstances and constraints in more detail.

The models have different structural representations of the energy system and solution paradigms and differ significantly in their assumptions and implementation of policies.⁷ The diversity of modelling approaches and assumptions reflect the inherent uncertainty about drivers and determinants of social systems, and the comparison of results allows for identification of robust and sensitive effects. For an overview of techno-economic assumptions used by the suite of models, refer to Krey et al (2019) and for a comparison of key socio-economic assumptions like GDP, Population, and Energy demand across the models, see SI section 8.

The modelling of energy storage and batteries is crucial to high shares of VRE in the energy mix. How these are modelled for each of the models is given in table S9 in the SI section 14. In general, storage requirements increase with increasing share of variable renewables and require additional investment which leads to increasing levelized cost of electricity (Pietzcker et al 2017, Ueckerdt et al 2017).

As the energy sector is changing fast and transformative changes are required to drastically reduce emissions, evaluation through bottom-up data allows to put scenario results into the context of current developments. These are especially relevant for the first future years of the modelling which typically takes place in 5-year time steps. We evaluate the near-term feasibility of these pathways by looking in-depth at what plagues or enriches each technological option in the power sector in India. More information about the bottom-up sources, namely the Central Electricity Authority's (CEA) National Electricity Plan (NEP) and Coal Swarm's 'EndCoal' database is available in the SI (section 2).

3.1.1. Coal expansion till 2030

Pledges under the NDC take into effect during the period 2020–2030. Under India's NDC, models project coal-based⁹ generation to increase relative to current generation, although there is a wide-spread (1158–2025 TWh in 2030 for global models in grey ribbon with upper limit until 2764 TWh when including all models, figure 1). One outlier to this result is the national model AIM/Enduse which projects a decrease in coal-based generation from 2020 onwards. It projects this transition by rapid expansion of natural gas-fired generation and CCS (Carbon capture and storage) and overall lower demand growth (See SI sections 7 and 8 for more information). The bottom-up projections from CEA, in black triangles, also show an increase and fall within the range of model results. Thus, both the modelling results and bottom-up projections show that under currently implemented policies and NDC pledges coal-based power generation likely continues to increase in India. This is consistent with India's NDC, which states that 'coal will continue to dominate power generation in future' and Coal India Limited, provider of over 80% of domestic coal, significantly increasing investment in exploiting the country's coal reserves (Press Trust of India 2019a). To provide further context, results from two other national modelling studies have been included—the NITI Aayog's India Energy Security Scenarios (IESS)¹⁰ and the NITI Aayog—India Energy Model (Thambi et al 2017), both falling well within the range of model results.

Zoom In Zoom Out Reset image size

Recent developments in the coal sector indicate that the periods of high capacity additions are over. The number of cancelled plants is increasing and those of planned plants decreasing (figure S1 in SI). In 2018–19, the capacity addition fell to a record low of 3.6 GW (Asian News International 2019) (compared to an average of 10 GW per year additions during 2015–2018) and some of the major power developers have vowed to move away from coal (Press Trust of India 2019b). One reason for this slowdown is the over-capacity in power generation which has led to around 40 GW of 'stressed' coal capacity¹¹ and many plants running at well-below their operating capacities (see SI section 12 for discussion on its drivers).

Thus, although there might be a reduction in the pace of addition of coal-based power generation in the coming years, the overall generation would likely continue to increase.

3.1.2. Solar and wind expansion

In 2015, the target for solar under India's solar mission was increased five-fold from 20 GW to 100 GW (India's NDC 2015) in 2022. This also became part of a target of 175 GW of renewable energy (excluding large hydro (> 25 MW)) by 2022. Additionally, in its National Electricity Plan, the CEA projects an addition of 50 GW solar and 40 GW wind during the period 2022–2027 to achieve 275 GW of renewable capacity in 2027 (black triangles in figure 2). Solar capacity has exponentially grown over the last five years (the current capacity¹² is 28 GW for solar and 35 GW for wind—see SI section 2). Even without a carbon price,¹³ new solar has become competitive to new coal and two-thirds of the existing coal power plants (Greenpeace 2017, Oliver 2018)

Zoom In Zoom Out Reset image size

Figure 2(a), for solar, shows that, in 2030, the national (270–380 TWh) and global (192–455 TWh) models are broadly in line with the projections (243 TWh in 2027) from the CEA. The results are similar for wind (figure 2(b)) with national (117–334 TWh) and global (151–312 TWh) projections broadly in line with projections from CEA (188 TWh in 2027).

3.1.3. Near-term projection of coal alternatives—gas

In the energy transformation pathways for many regions of the world, gas is the most important alternative to coal in the near-term. Secondly, gas provides peak capacity, increasing the flexibility of the power system as more renewables are integrated.

Compared to projections from CEA, many models project significant near-term increase in gas-based generation under current policies (figure 3). However, several factors make this scenario unlikely for India.

Zoom In Zoom Out Reset image size

Low supply of domestic gas and high prices of imported LNG (Liquefied Natural Gas) have left around 50% or 14 GW of the plants stranded; the current PLF (plant load factor) of gas plants is 25% (Central Electricity Authority 2018b). Secondly, gas for power plants competes with other uses, which are given a higher priority (Standing Committee on Energy 2019). These include cooking (as PNG or Piped Natural Gas and LPG or Liquefied Petroleum Gas), as CNG (Compressed Natural Gas) in the transportation sector, and the fertiliser sector (as raw material).¹⁴

The construction of the long-distance TAPI (Turkmenistan-Afghanistan-Pakistan-India) pipeline woud eventually¹⁵ supply India with more gas (almost half of the current domestic production of 32 billion cubic metres). However, as mentioned before, power generation competes with other uses and as the pipeline connectivity within the country improves (Ministry of Petroleum & Natural Gas 2015), the demand in these high-priority sectors will also increase. Under these circumstances gas power plants are unlikely to receive a dominant share of the incoming gas.

Thus, unless there is a further decrease in international gas prices and subsequent ramp-up of LNG terminals or exploration and ramp-up of shale-gas production, gas is unlikely to play a significant role in near-term power production in India (Sen 2015, Chaturvedi et al 2018).

3.1.4. Other technologies

In the delayed action scenario, the power system follows the NDC trajectory until 2030 (as presented in 3.1). Thereafter, the global IAMs achieve cost-effective mitigation (through a carbon price) under a carbon budget constraint up to 2100. On the other hand, an 'immediate or early action' scenario introduces a carbon price already after 2020. The difference between the two scenarios can shed light on the path-dependency of near-term actions. The concept of carbon lock-ins suggests that near-term addition of carbon infrastructure makes stringent mitigation targets more difficult and costlier to achieve—by prohibiting alternatives to emerge and wasting investments through premature retirement and stranded assets. The rest of the section will show key differences between these two scenarios.

3.2.1. Stranded capacity

Under a climate policy based on carbon pricing, the carbon price increases with time. Such a policy forces power plant operators to run their plants at a load factor well-below its optimal design to reduce operating costs. Furthermore, if the load factor falls below a certain point, the coal plant cannot recover fixed and variable operational costs leading to premature retirement, i.e. before its expected lifetime, and the plant is said to be stranded. However, such a chain of real-life decisions are not represented in models because of their inability to include single power plants and track their age over time. We thus use an illustrative way to calculate stranded capacity, which is uniform across models (See SI, section 10 for details).

Figure 4 illustrates four aspects—(1) figures 4(a) and (b) show coal capacity as it retires naturally¹⁶ (dark blue line), starting in 2020 (early action) and 2030 (delayed action). The bars represent coal capacity and are color-coded according to the age-group of the plants in each year. How the plants are tracked over time is explained in detail in the SI section 10 but an illustration is provided in figure 4(c) on how vintages are calculated); (2) The black lines are the early and delayed mitigation pathways compatible with the Paris Agreement (with the global model REMIND as example), (3) The stranded capacity is represented by the region above the black line (as an example in figure 4(a) the purple line depicts stranded capacity, for the years 2030 and 2040 in early action). For both—the early and the delayed scenario in REMIND, roughly all plants older than 20 years are retired, but the magnitude of stranded capacity is higher in delayed action (∼ 300 GW vs ∼ 150 GW in early action in 2050); (4) The Total stranded capacity (polygonal area) in the delay scenario (figure 4(b))—shows that the magnitude of stranded capacity from plants yet to be built and currently installed plants would be similar.

Zoom In Zoom Out Reset image size

Results for the other models are presented in SI section 10 but main results are given below.

The range of stranded capacity in the period from 2030 to 2050 across models for Delayed action is 133–227 GW and for Early action, over the 2020 to 2050 period, 14–159 GW (SI section 10, table S7). In general, although delayed action leads to higher total stranded capacity, early action leads to slightly higher stranding of younger plants (in the age group of 11–20 years), (SI, section 10, table S7 for details). This is because today's plants, most of which are quite young, are stopped almost immediately in the early action scenario.

Thus, in scenarios where a carbon price is enacted early, the amount of stranded capacity is reduced but not eliminated. Importantly, however, in the early scenario, only already existing plants become stranded, many of which have been planned and constructed without anticipation of the Paris Agreement or of the cost reductions seen for crucial decarbonisation technologies like solar, wind, and battery storage. In the delay scenario, a sizeable share of stranding is from plants yet to be built (see bottom right in figure 4, panel b).

3.2.2. Solar and wind potential

Early action scenarios introduce stringent climate policy in the form of a carbon price. A higher carbon price makes fossil fuels more expensive and incentivises alternatives to emerge (see SI section 9 for a carbon price comparison across models). In 2030, the delayed action scenario has no carbon price.

In early action scenarios, more renewable energy and nuclear is added to the power system (see figure 5, except IMAGE- see discussion in SI section 7), while reducing coal power need by 557–1320 TWh (see 'difference' in 2030, excluding AIM/Enduse). Thus, given the option to start mitigation earlier, many models decide not to build coal (SI figure S10). Although the long-term (2050) final power demand across models does not differ significantly in the two scenarios (SI figure S4), for the global IAMs, the demand dips following the introduction of a carbon price (visible by the sum of the bars in figure 5). For this reason, not all coal is replaced by renewables in the delayed action. Lower electricity demand in the early action leads to lower generation requirements in the near-term. This is in contrast with the much lower elasticity of national models, where electricity demand in the near-term is almost unaffected across the two scenarios.

Zoom In Zoom Out Reset image size

Importantly, IAMs and national models project that, under a constrained carbon budget, an even more rapid scale-up of (primarily) solar and wind compared to early action (figure 2) is cost-efficient.

Coal-fired power plants planned for the next decade would constitute a significant share of stranded capacity under a climate policy compatible with the Paris Agreement (figure 4). Avoiding this requires further investment in solar and wind. There are some indications that India might actually raise its solar and wind targets to 300 GW and 140 GW respectively (Central Electricity Authority 2019, Reuters 2019), from the 150 GW and 100 GW (solar and wind respectively) in 2027 included in the NEP. Thus, such an ambition would be a step towards decarbonizing the power system. However, increasing coal generation (aided by the absence of an explicit policy limiting coal generation) and with increasing penetration of variable renewable energy (VRE) in the power system, new coal generators could face low or falling load factors (Hirth et al 2015, Palchak et al 2017, Scholz et al 2017, Chaturvedi et al 2018), exacerbating the current stressed assets in the sector. At the same time, although it might be 'economically' rational to lower coal power PLF under such a scenario (especially for newer coal plants with higher tariffs), political economy factors surrounding coal power and electricity pricing in India could mean that VRE is curtailed in spite of its must-run status. Curtailment is already a serious issue for VRE investments in Andhra Pradesh and Tamil Nadu (Jawar 2020) .

The sections before highlighted the policies and targets in place for renewable technologies and outlined the current situation of the coal power sector—one reeling under low plant load factors and many stressed assets. Taking these, and other literature studies into account, this section explores the various trajectories of coal capacity development in India and their implications.

In 2015, a new legislation was introduced requiring all coal power plants to control the concentration of certain pollutants (Central Electricity Authority 2018b). In general, the implementation of this legislation will significantly reduce the share of power plants in total SO2, NOx and PM2.5 emissions (Purohit et al 2019), thereby reducing the possibility of them being prematurely shutdown due to air pollution concerns. However, uncertainty arises due to the heavily regularised and politicised nature of the electricity sector and the location of certain plants close to large cities (see SI section 14 for details).

Both the early and delay action scenarios (orange and green dashed lines- figure 6) consider the implementation of a high carbon price which gradually increases over time (SI section 9). Such high carbon prices could result in disruptive changes and financial instability (Campiglio et al 2018, Kriegler et al 2018)—as also shown by the large amounts of stranded capacity (SI section 10). Therefore, such a policy would be especially avoided by risk-averse policy makers in a growing but still developing country like India. Secondly, although carbon pricing is the principle policy instrument used in IAMs to mitigate emissions, previous studies have shown that policy makers favour to implement a mix of multiple, overlapping instruments over carbon prices to achieve climate mitigation (Jenkins 2014); often starting with the power sector (Murdock et al 2019). Such policies are not only more politically feasible to implement but give rise to coalitions and constituencies supporting low-carbon transformation, essential in the political economy of decarbonization (Meckling et al 2017).

Zoom In Zoom Out Reset image size

The purple line represents a continuation of coal capacity growth as projected till 2030 and a natural decline thereafter, and thus shows the risks of what the continuation of NDC policy ambition until 2030 entails. Under such a scenario, coal power in India alone would take up ∼ 11% of the global carbon budget for a 1.5 C target (see SI section 3).

Thus, a more politically feasible pathway in the short-term and an intermediate between the two policy scenarios is represented by the spectrum spanned by the green and purple lines. Here, no coal additions take place (beyond plants under-construction in turquoise) and the coal plants run till the end of their lifetime. Such a policy ('coal moratorium' in (Bertram et al 2015a) will bring additional benefits—keeping the plant load factor of existing coal plants at the current level, preventing the power system to get further locked into coal and thus necessitating large stranded assets in the future, opening the possibility of integrating emerging and cheaper power technologies in the future, and as mentioned before—laying down important groundwork for ambitious future climate policy. However, such a policy would necessitate a moderate increase in power from other sources, but at a reduced rate compared to the 'Early action' scenario. As presented in preceding sections, Solar PV and Wind could take the bulk of the additional electricity demand.

The study quantifies the stranded coal power capacity in India in the context of declining costs of renewables, especially solar, long life of coal power plants, and policies in line with the Paris Agreement. However, a number of other factors could influence the stranded capacity of coal generators, which have been either partially considered or absent in this study.

The study does not explore specific environmental constraints like water-use in coal power plants, which will become increasingly relevant for India (Caldecott 2015, Manthan India 2017, Vishwanathan et al 2018, Tang et al 2019), nor does it include local environmental damages from mining (Worrall et al 2019). Other factors unique to India which affect the operation of coal generators, like the severe debt of distribution companies and implicit and explicit subsidies to coal (Worrall et al 2019) have also not been considered.

Other limitations include inherent methodological challenges—like the inability to run at hourly timescales (which is important to explore grid and plant flexibility at high VRE penetration rates), although models use various approaches to represent integration challenges of VRE in the grid (Pietzcker et al 2017), further information in SI section 14). Furthermore, how higher shares of VRE in the grid could exacerbate stressed assets in the power section and lead to stranding has only be mentioned qualitatively.