Coal fleet’s changing role in the power system

The role of coal-based generation is shifting from steady baseload to flexible balancing resources as variable renewables expand. This will require greater operational flexibility from coal units and, over time, a reduction in their overall utilisation or plant load factors (PLF). Overall utilisation of coal is likely to fall, and the decline is uneven: higher variable-cost, less efficient units are likely to see the sharpest drop in run-hours, raising their effective costs and underscoring the need for operational reforms, flexibility investments, and selective retrofits.

Recent operations already show deep mid-day turndowns and fast ramps, with the grid code defining minimum-load and ramping expectations—but in practice, a handful of centrally dispatched plants shoulder most of the burden and often operate near technical limits. Looking toward the NEP horizon, modelling with the chronological, least-cost operations model of the Indian power sector (PyPSA-India) indicates coal will need to run even closer to minimum load with steeper, more frequent ramps, otherwise the system leans on gas or curtails renewables.

2.1.1 Technical coal flexibility limits can be characterised by the minimum thermal load and ramp rates

Minimum Thermal Load (MTL) — also referred to as the minimum technical limit — is the lowest output level at which a thermal power plant can operate reliably over a sustained period, usually expressed as a percentage of its rated capacity. India’s previous grid codes had recognised the need for flexible operation and the importance of reaching consensus around MTLs. The latest Grid Code (2023) formalises this limit at 55%, based on a detailed Central Electricity Authority (CEA) technical report.

In practical terms, a thermal unit with a rated capacity of 100 megawatts (MW) should be able to operate at a minimum of 55 MW on a sustained basis. If required to go below this level, the plant can legally declare unavailability since it will compromise reliable operations. 

The code also specifies ramping expectations of around 1% of rated capacity per minute, defining the minimum pace at which units should be able to increase or decrease generation. 

Together, these parameters form the operational framework that allows system planners and grid operators to swing coal generation flexibly — a capability that has been clearly demonstrated in recent years as India has managed rising daytime solar generation.

2.1.2 The coal fleet is already flexing substantially during some periods of the year

One of the main requirements of coal-based generation flexibility is that units must stay online at or just above their minimum technical load to be synchronised and ready for the steep evening ramp. They run low through the day to accommodate solar peaks. Turning units off at noon means long start-up times, often hours or even days. It adds thermal stress and wear, raises fuel and operations & maintenance (O&M) costs, and risks missing reserve margin targets. Keeping units warm and online preserves ramp capability and frequency response when solar falls and net load surges.

This flexibility requirement from the coal fleet has increased year-on-year. Coal plants operated with noticeably greater flexibility in 2025 than in 2024, significantly ramping down during solar hours and ramping up to meet the evening peak. This was because of two  factors – rapid expansion of variable renewable energy, especially solar, and slower system demand growth owing to fewer cooling-degree days this year. 

The surge in solar was key to the suppressed midday net load and raised the flexibility required from coal. In 2025 versus 2024, renewables excluding hydro grew most in the solar window (about 09:00 to 15:00), while late evening and night saw only modest gains. 

At 12:00 in January, output rose from 41,918 MW to 59,099 MW, a 41% increase. Across 09:00 to 17:00, generation was up by roughly 30%. At its peak, renewables excluding hydro reached about 76,401 MW (76.4 GW), roughly 37% of the coal fleet’s capacity (about 205 GW), increasingly offsetting coal during solar hours and deepening the net load “duck.”

PLFs have dipped to as low as 55% in the past, driven by a surge in coal capacity additions between 2010 and 2015 and again during the COVID-19 period. But now, the pattern of utilisation across hours and seasons is now changing fundamentally.  The system increasingly needs the thermal coal fleet to ramp more often and operate at lower loads through the day in order to integrate variable renewable energy into the supply mix.

As a result, 2025 has recorded the highest level of thermal-fleet flexing to date. The figure below shows the dispatch pattern for three representative days in April 2025, a period characterised by one of the steepest coal ramping requirements of the year. 

During this period, total generation peaked at ~215 GW (with a near-midnight peak of ~214 GW). Mid-day solar reached ~60 GW, wind contributed up to ~16 GW despite April not being a high-wind month. While hydro supplied as much as ~24 GW during the morning and evening peaks, coal generation needed to ramp down sharply. 

Between 07:00 and 13:00, coal declined from ~165 GW to ~113 GW, a 52 GW reduction over six hours—an average ramp-down of 8.7 GW/hour (≈145 MW/min). This pattern underscores both the depth of the midday net-load trough and the growing need for fast, flexible capacity to manage the evening rebound.

With a 113 GW minimum generation from coal, the fleet operated well inside its theoretical ramping down limit of ~94 GW (see Annexure 1), implying roughly 19 GW of latent downward flexibility. While a portion of this headroom could likely be committed as downward reserves during solar hours, ramping down close to technical limits would absolutely become essential to integrate more renewable energy.

2.1.3 Coal fleet will consistently need to operate flexibly and closer to its technical limit

By FY 2031–32, the structural role of coal will shift decisively toward providing residual and balancing energy, especially during solar hours.

The modelled dispatch for the likely mix in FY 2031–32 illustrates the structural shift expected for coal-based generation in the same period as the 2025 observation. We simulate FY 2031–32 dispatch with a chronological, least-cost PyPSA model (detailed description in the supplementary material)

The model projects maximum demand to peak at around 330 GW in this period, with coal continuing to play a critical role in balancing solar and wind variability.

Two significant shifts are likely in coal generation. First, coal plants will need to operate much closer to their technical minimum to avoid solar curtailment and the use of more expensive gas-based generation. The figure above shows that solar curtailment occurs, which would not take place if the coal fleet could hypothetically ramp lower. 

Secondly, even though coal is not generating at its maximum level, we still see more expensive gas-based dispatch. This happens because the system is optimising while considering the need for coal units to ramp down during the day and the number of units that must remain online to ensure availability for the evening ramp.

According to the model’s results, coal output is likely to peak at about 190 GW during non-solar hours and ramp down to roughly 112 GW at midday solar peaks—a ~78 GW daily swing. 

A peak dispatch of 190 GW (net) implies ~206.5 GW (gross) online, assuming 8% auxiliary consumption. At a 55% MTL, this corresponds to ~104.5 GW (net). The expected midday coal output of 112 GW (net) is, therefore, only ~7.5 GW above the theoretical minimum (see annexure 2). In other words, the fleet has already demonstrated close to this but will need to sustain such performance consistently day after day and across months as solar penetration rises.

Practically, it may not be possible to run the entire fleet at the MTL simultaneously due to unit-specific constraints. This implies some units may need to turn down below 50% and potentially toward ~40%, which is currently being planned, to deliver the required system-wide turndown while keeping enough units synchronised for the evening ramp.

Further, the model also finds that the morning ramp-down is also significantly steeper than today. Coal dispatch falls from 190 GW at 07:00 to 112 GW at 10:00—78 GW in three hours—equivalent to ~26 GW/h (≈ 433 MW/min) or ~13.7% per hour of the peak online net capacity.

Compared to 2025, the FY 2031–32 system demands much higher levels of flexibility from coal. To make this workable operationally, many plants will need to hold around a 40% effective MTL for maintaining sufficient downward (and upward) reserves during solar hours—minimising reliance on limited gas generation and avoiding the perverse outcome of providing reserves via renewable curtailment. 

For the modelled FY 2031–32 system, the 40% gross MTL is ~82.6 GW, which equals ~76.0 GW net. At a mid-day coal output of ~112 GW net, the headroom above this 40% MTL is ~36 GW. This lower effective MTL enables deeper daytime turndown while keeping enough synchronised units for automatic generation control (AGC) response and the evening ramp. This would be more than sufficient to displace the more expensive gas generation and materially reduce renewable curtailment.

If the coal fleet cannot reduce output sufficiently close to its technical minimum, the system must lean more on gas during non-solar hours—not because gas is cheaper, but because coal is deliberately held back to respect unit limits. Once units are kept online, they generally must stay online (min-up constraints) and cannot be turned down below their MTL (typically ~55%). It means the allowable pre-sunrise coal output is effectively capped by the midday target and the MTL of individual generating units. If the desired peak exceeds this cap, the excess appears as gas generation. In our modelling, this shows up as ~10 GW of gas in the evening/morning shoulders owing to a technical-limit outcome to avoid forcing the coal fleet below its feasible minimum at the solar peak and triggering curtailment or infeasible operations.

Considering the scale of the likely problem, there have been multiple piloting efforts by CEA to understand the technical feasibility and cost economics of setting MTLs at 40%. Most of the generators chosen for the piloting study have shown promising results. However, most of them have been newer plants designed for flexible operations. Several intra-state plants have shown concerns about extremely low MTLs (below 55%). 

Therefore, incorporating an overarching 40% MTL would be far from due in the grid code. But there would be many newer plants that could go below 55% and accommodate some of the solar.

2.1.4 Practical limitations to increased coal flexibility

The above calculations are done at a fleet level and are a simplification to build a basic understanding. Such flexibility would exist only as long as the entire coal fleet operates within a centralised dispatch framework. However, in reality, a handful of central ISGS (Inter-State Generating Station) coal plants — whose generation tends to be shared by multiple states and whose operations are overseen by the national grid operator — have been forced to shoulder most of the flexibility burden. In contrast, intra-state generators, operating within individual state jurisdictions, remain unresponsive to flexibility needs. As a result, the burden of adjustment falls disproportionately on a few plants.

In several instances, these ISGS generators have had to ramp down close to — or even below — their MTL to accommodate rising solar generation. This has led to events of forced outages, or conversely, over-injection into the grid to stay online and above MTL — both of which pose significant challenges to grid stability and reliable operations. 

Reserve obligations/participation tighten the noose further. If units are required to hold reserves (primary/secondary/tertiary), they must keep additional headroom above their minimum to deliver down regulation on command. Practically, that means a unit that could have gone down to its technical minimum (say 55% gross) may have to sit at MTL + reserve headroom (e.g., 55% + 2–5%), simply to remain eligible to provide frequency and balancing services. The same few ISGS plants that already shoulder energy-ramping duties are also the ones carrying reserves — so they cannot descend to the theoretical minimum even when the energy balance suggests they should. If AGC (automatic generation control) participation is concentrated in these units, they end up “hovering” above MTL to maintain down reserve and response quality, which reduces the achievable midday turn-down and pushes more curtailment risk onto solar (or more gas backfill in shoulder hours).

Our model projects the overall utilisation of India’s coal fleet to decline sharply by FY 2031–32. This will be especially true during solar hours, as solar and wind displace coal generation, leading to a predicted fall in average PLFs from about 69% (FY 2024-25) to 55%. This reduction will be uneven, hitting higher-cost and less flexible coal plants the hardest, with some units facing near-complete idleness and significant financial implications due to fixed costs being spread over fewer operating hours.

2.2.1 Reduced overall utilisation of the fleet, especially during solar hours

As coal shifts away from baseload operation—with solar displacing generation during the day and wind during both solar and non-solar hours (particularly in high-wind months)—the PLF of coal units is projected to fall from about 69% in FY 2024–25 to around 55% in FY 2031–32,as per Ember’s modelling).

The maximum instantaneous utilisation of coal (gross generation) was around 85% of the installed coal capacity (≈221 GW) in FY 2024–25. 

In the modelled NEP 2032 scenario, utilisation similarly peaks at ~85% of a larger installed capacity of 259.4 GW. However, the instances of high utilisation become much shorter under the NEP scenario. There also remains sufficient coal capacity to meet peak demand between April and June during non-solar hours. Outside these peak instances, coal utilisation declines even during non-solar hours due to the growing contribution of wind, hydro and storage resources. The reduction is much sharper during solar hours as solar generation directly displaces coal. 

However, adding more coal capacity may not necessarily guarantee availability during these hours, since actual availability depends on how many generating units are online and can be kept synchronised during periods of low generation within the same day.

2.2.2 Reduction in utilisation is non-uniform across the fleet

While the coal fleet will see an overall reduction in utilisation, the impact will be uneven across units. Plants with higher marginal costs are likely to be utilised for less. Conversely, units with lower technical minimums at similar marginal cost are expected to see higher utilisation.

Plants with variable costs higher than roughly INR 2.8/kilowatt-hour (kWh) will see significant reductions in run hours, and those plants with a marginal cost less than INR 3/kWh will operate for only a few hours per year, risking significant underutilisation. The fleet is, therefore, relatively cleanly divided into baseload, flexible and peaker units, with many plants needing to adapt to their new roles. This shift will have significant technical and financial implications, discussed in Chapter 3. 

2.2.3 Planned coal capacity sufficient to maintain supply-side security

The figure below shows the utilisation of the coal fleet, illustrating the percentage of plants and their actual utilisation in FY 2023–24 compared to the modelled utilisation for FY 2031–32. The first clear trend is a significant drop in utilisation for roughly the second half of the fleet—mainly the higher-cost marginal generators.

About one-third of the total capacity could experience a sharp decline in utilisation, ranging from a 20% drop to complete non-utilisation. This means that plants operating with PLFs in the range of 35–40% today may see little to no operation by FY 2031–32, while even units that currently sustain the overall fleet average of ~69% may experience their utilisation reduced by half. 

The results also indicate that ~10% of additional plants are completely idle in FY 2031–32. Taken together with the under-utilisation described above, these suggest more than sufficient coal capacity to meet reliability needs. As a result, India will not need new coal capacity beyond the 35 GW already under construction to maintain supply-side security in FY 2031–32.

This trend becomes even more pronounced under an optimised dispatch framework, where generation is allocated strictly based on cost efficiency. This approach leads to a more concentrated reduction in output among less competitive units. It could lead to an increase in the utilisation of units with a lower marginal cost. While the utilisation of around half of the fleet shows a slight increase, this reflects an optimised dispatch outcome in the model, which assumes perfect system efficiency. 

In practice, such optimisation is difficult to achieve, but ongoing mechanisms such as SCED (Security Constrained Economic Dispatch) and the sale of URS (Un-Requisitioned Surplus) already help move operations closer to that ideal by enabling real-time cost-based dispatch and allowing surplus generation to be sold across states. Moreover, the proposed MBED (Market-Based Economic Dispatch) framework aims to integrate all generators into a unified national merit order, pushing the power system even closer to the optimal dispatch envisioned in the model.