Balancing India’s 2030 Electricity Grid Needs Management of Time Granularity and Uncertainty: Insights from a Parametric Model

India has some of the most ambitious targets for renewable energy (RE) in the world, which were formalized by Prime Minister Modi at the UN Climate Change conference COP26, in Glasgow November 2021. India aims for 450 GW of RE (and 500 GW of non-fossil) capacity in its electricity grid by 2030, roughly a quadrupling of RE capacity in eight years.

In a new study, we examine India’s grid and its balancing through 2030, under high Renewable Energy (RE) scenarios, with half-hourly time resolution and high uncertainty across variables including demand, share of wind versus solar, coal power plant flexible operations, etc.

Read the paper (journal version)

Read the author pre print

Read the blog post with follow-up thoughts on Grid 2030: A glass half cloudy

A few of the questions we attempt to answer include:

• How will India manage its extremely aggressive renewable energy (RE) plans?[1] How much RE is feasible, and are there risks of “too much” RE, which would lead to curtailment?
• Does India need more coal-based power? Does it need to build more coal power plants?
• What supply (fuel) options make the most sense for new capacity going forward? This isn’t just a function of their economics but also issues like system security and managing volatility or uncertainty.
• Are batteries the answer in the short term or longer-term? How should they be thought of for planning?
• What are the factors that matter for grid planning and policy?

This study is complementary to a range of other studies on grid 2030, which are summarised in a study by LBL, and uses a different technique based on parametric uncertainty modeling to couple investment decisions for capacity expansion (to meet projected demand over time) with optimal economic despatch of generation fleet as chosen.  It is only an all-India calculation, and hence doesn’t capture any transmission constraints (this is equivalent to “infinite” or “as required” transmission), and also does not vary the load profile shape compared to 2019, the baseline year for data. We use data from CSEP’s carbontracker.in portal, which captures demand and supply mix all-India at 5 minute resolutions for the baseline data, which provides both demand and RE output at high resolution.  We also do not undertake analysis at a per plant level, instead relying on simple buckets of plants based on their fuel costs, e.g., if they have cheap gas supply available or if they are distant plants (which have more expensive coal).  We assume modest growth of hydro and nuclear capacity as the baseline.

Importantly, the focus is on trends, uncertainty, and tradeoffs, and not on specific values of costs of optimal capacity.

The analysis has two main parts. First, we examine what the ability of existing capacity plus planned growth of variable RE (VRE, without storage) would be to meet demand per time period. If it is unable to meet demand, something NEW[1] is required. This analysis factors in issues of coal plant flexible operations, and assumes no daily start-stop (2 shift) operations. It also considers a range of how many coal power plants from the existing fleet retire by 2030. We also assume no shortfalls in fuel supply, be it coal, nuclear, or gas.

For NEW supply, we calculate the lowest cost for a range of options including new coal capacity, new gas (operating in open cycle, combined cycle, and spark ignition modes), biodiesel, and batteries. We also consider a limited range of blended options.

A few of the highlights of the findings and implications include:

• Higher RE as variable RE (VRE) is the optimal strategy for India’s grid. 450 GW RE by 2030 (even without storage) is ambitious but cost-effective, even with high curtailment (varying significantly based on ratio of solar versus wind power). Not only is 450 GW RE not “too much”, even at such high RE capacity India would still have around half its supply coming from coal (Figure 1)

Figure 1: 2030 optimal supply mix based on varying shares of RE in 2030 (before adding something NEW, which might further displace generation from existing gas or coal).

• Adding high Plant Load Factor (PLF, also called capacity utilization factor, or CUF) wind is important at a systems level, even if wind’s levelized cost of power (LCOE) is higher than for. This is offset by producing more output at periods of non-solar-coincident demand, and thus reducing the demand for NEW supply.
• Even with very high RE and minimal coal fleet retirement, modest demand growth for electricity of 5.25% annually means there is insufficient supply at many time periods from the existing capacity plus RE, and so “something new” is required;
• India’s grid has benefited at an operational level from surplus coal power plant capacity, which will exhaust in a few years. 2030 simulations show maxing out of existing capacity, and a rise in gas output, more so in a peaker role. Thus, the Indian grid has not yet faced high costs for peak power, which will be what most of NEW supply will be for. A corollary finding is that India needs to ensure adequate fossil fuel supply, since using existing capacity is typically the cheapest option for supply;
• NEW supply will inherently be relatively expensive because it will operate at low PLFs if its duty cycle is to meet otherwise unmet demand (Figure 2). Stated another way, India will need to embrace the concept of peakers in the future, which have low utilization and higher per unit (kWh) costs. Figure 2 shows how NEW capacity is required in full only for a few hours per year, and for much of the year it isn’t needed at all.

Figure 2: Load Duration Curve for NEW Supply required (half hourly stacked time periods, i.e., not chronological) in 2030

• In case a battery is used (with RE), it is the cheapest of NEW supply options not because it is cheaper than coal on a standalone (LCOE) basis but because it can displace expensive fossil fuels at the margin, and so offers value beyond simply meeting NEW supply (otherwise unmet demand). Batteries also benefit from the ability to use surplus (curtailed) RE, modelled at marginal costs, i.e., free RE; This is one reason we find high energy (TWh) coming from batteries by 2030 – much of it is because the capacity displaces fossil fuels in some time periods when otherwise not required (as shown in Figure 3b). If we do add batteries (which assumes a price-performance curve improvement over time), it also reduces curtailment of RE. Batteries remain a cheap option for NEW supply but per kWh will cost measurably more than existing coal capacity. This is because in many time periods, especially seasonally, one doesn’t need a battery to meet otherwise unmet demand, and hence its value is only the marginal cost of fuel displaced, predominantly coal. This can be perhaps under Rs. 3/kWh if at the pithead in 2030, which doesn’t cover the projected full costs of the battery in 2030.

Figure 3: Generation mix over time with 450 GW RE (a) prior to adding NEW; (b) post adding a battery for NEW

(a)                                                                                         (b)

• The kink in coal output seen in Fig 3b isn’t just due to use of batteries for NEW but also rising RE. The trajectory of RE capacity addition matters – linear vs. exponential/fixed CAGR, impacting for how soon and by how much we need something NEW before 2030.  If there is high short-term energy growth outside the VRE (variable RE) window, there could be challenges for meeting peak demand (esp. measured as “net demand” = after removing demand met by VRE, typically in the evening). Any such challenges are heightened if there are fuel supply challenges. In the paper, we assume a steady CAGR for growth of RE. Slight acceleration of RE growth can help but it’s also not easy to achieve in practice.
• A disproportional share of incremental costs is for selected (peak) time periods, especially when measured by “net demand” (demand after removing RE supply). Time of day (ToD) pricing will be important for both shifting consumption to off-peak periods and signalling for the correct investments in capacity.This includes the need for a smarter and resilient grid under uncertainty and volatility, which will only worsen as RE’s share increases. The use of ToD pricing for wholesale procurement is likely to impact VRE more than traditional (despatchable) supply, and based on the coincidence of solar across most of India (especially compared to wind), ToD pricing will ultimately reduce the value of solar mid-day;
• India will need to work backwards based on construction schedules to see by when it wants to procure for NEW supply. By that time, there would be greater clarity on fossil fuel prices and technology prices, especially for RE and batteries. India should be cautious in premature retirement of existing capacity, even if this is rarely used, as it provides disproportional value at the margin.[3] The emissions from rarely used coal plants will be low, but the value of such capacity is very high.

As this study shows, LCOE is a poor measure of costs of supply, and instead a portfolio approach is required, which inherently factors in implications upon other fuels; this is before considering transmission costs.  Even at Rs 7,500/kWh for a battery inclusive of the inverter and with low operations & maintenance (O&M) costs, its LCOE is about Rs. 4/kWh, excluding the RE that charges the battery. An LCOE calculation assumes that the battery is used about 90% every day, but as we show, it isn’t needed every day at this same level.  Ultimately, “are batteries cheaper than coal?” is the wrong question, and we need to ask “what’s the cheapest way to meet demand as a portfolio?” which means we compare new batteries with new coal plants, not existing coal plants. Batteries with RE would take longer to displace existing coal plants at the margin (assuming they operate on domestic fuel).