Data overview

The wind data used in the study were derived based on MERRA-217, a NASA reanalysis product publicly available in NASA’s Goddard Earth Sciences Data and Information Services Center. This database defines hourly wind speeds with a spatial resolution of 0.50° longitude by 0.67° latitude from 1980 to present. Wind speeds at 100 m were extrapolated from 10 to 50 m using the vertical profile of the power law described by Archer and Jacobson25. The friction coefficient in the analysis was evaluated using wind speeds represented at 10 and 50 m for each grid cell, as in Lu et al.26. Wind power was computed on an hourly basis using the power curve for the MHI Vestas Offshore V164-8.0 MW wind turbine, a typical system employed currently for offshore applications, and the Goldwind 2.5 GW for onshore. Specifications for each technology are summarized in Supplementary Table 2. Capacity factors (CFs), defined by the ratio of electricity generated by a solar installation relative to the realization of its full capacity over the same period, were evaluated on an hourly basis at the spatial resolution of the NASA database. The solar data used in the study were derived from NASA’s GEOS-5 FP database27, which identifies hourly temperatures and incident solar radiation at a spatial resolution of 0.25° latitude by 0.31° longitude. We employed an integrated solar PV assessment model in evaluating the performance of solar PV systems, following the approach described by Chen et al.18. The spatial variation of factors impacting CF were modeled consistently, accounting for tilt, packing density, sun shading, and temperature. Hourly solar power values were calculated assuming installation of fixed-tilt polysilicon PV modules with a 16.2% conversion efficiency.

Onshore filter

Onshore areas that are forested, urban, or covered with water or ice were filtered according to data from the NASA MODIS (Moderate Resolution Imaging Spectroradiometer) satellite MCD12C1 dataset28. Slope data were derived from the Shuttle Radar Topography Mission (SRTM) Global Enhanced Slope Database29 with a spatial resolution of 1 arc-s (~30 m). Grids characterized by slopes of more than 20% or by heights of more than 3000 m were excluded as inappropriate for deployment of onshore wind power systems.

Offshore filter

To determine locations suitable for offshore wind in India, we filtered data spatially based on a number of criteria. First, only locations within India’s Exclusive Economic Zone (EEZ) were considered. India’s boundaries for the EEZ were taken from Marine Regions, a database which aggregates information from a number of regional and national providers30. Another filter adopted was to consider only fixed-bottom turbines, which require offshore depths of less than or equal to 60 m. The offshore depth data used here were taken from the General Bathymetric Chart of the Oceans (GEBCO) One Minute Grid, a global bathymetric grid providing data at a one-arcminute resolution31. Finally, we removed areas from each grid according to environments designated as either “Special Marine Reserves” (environmentally-protected regions) or shipping routes. Areas for the Special Marine Reserves are defined in ref. 32. The SO2 emissions compilation from MERRA-2 was used as a surrogate in the identification of shipping routes, and 20% of a cell’s area was removed for locations defined as emitting SO2 at a rate higher than 10−11 kg m−2 s−1.

Solar filter

This study used slope, land use type, and solar radiation as criteria to identify areas suitable for solar farm development, following the approach described by Chen et al.18. The maximum permissible slope was set at 5%. As with onshore wind, the SRTM database was used to calculate terrain elevation and slopes for each grid. Suitability factors were selected according to land use types with higher values allocated to land areas with sparse vegetation and low ecological productivity33 (Supplementary Table 6). The MODIS data were used to filter unsuitable land areas from this analysis, excluding forests, water bodies, permanent wetlands, croplands, cropland/natural vegetation mosaic, and snow and ice environments (land classifications are indicated in Supplementary Fig. 4). Areas excluded by these filters were assigned 0% as suitability factors. For exploitable areas, suitability factors ranging from 5 to 20% were assigned to each land use type. The minimum solar radiation required for exploitable land areas was set at 1400 kWh/(m2a), a typical threshold value for acceptable solar resources18. And, it should be noted, the current study does not allow for a potential source of carbon-free electric power from solar panels installed on roof tops, a development that could be facilitated by appropriately targeted policy initiatives.

Regional power capacity

Musial et al.34 estimate that the spacing appropriate to minimize turbine-turbine interference for offshore wind is equivalent to ~7 rotor diameters, corresponding to a deployment density for turbines of one per 1.04 km2. The area for each latitude/longitude grid cell was divided by this value to compute the number of turbines that could fit maximally into a given cell. It should be noted that this spacing does not account for the downstream wake effect, which is of too small scale to be modeled accurately using the MERRA-2 data. Given that the average downstream power loss is on the order of 5%35, the wake effect should not have a significant bearing on the present results. The potential installed capacity (in GW) is computed by multiplying the number of turbines in a cell by the turbine power (8 MW in this case). Onshore power is calculated similarly, using a spacing of 9 rotor diameters (one turbine per 0.64 km2) and turbine power of 2.5 MW. The solar power PV capacity potential (in GW) is defined by the packing factor obtained by multiplying the power per unit area of the PV panels (161.9 Wm−2) by the area available for their placement (factoring in solar filter constraints as described before). The spatialized packing factor here refers to the effective panel area per square meter of land area, which is determined by the solar PV tilt, azimuth angle (east-west orientation), and the spacing between neighboring PV panel footprints. The tilt setting assumed in the study follows the method proposed by Jacobson36, and the orientation of the panels was set to face the equator. The principle to determine the spacing between footprints is to ensure that minimal shading will occur for most of sunlight hours throughout the year. The spacing was calculated using the solar altitude angle for 3 PM at the winter solstice, the day for which shading is likely to be most significant.

Power generation

The next step is to quantify the power that could be supplied to individual regions. For offshore wind, we assumed that the wind resource available over a given location in India’s EEZ was under the jurisdiction of the country’s nearest region. These regional divisions, along with the mean of on- and offshore wind and solar PV CFs over the year 2016 are indicated in Fig. 2. From the installed capacity and CF data, estimates of available energy for each technology E(lat,lon,t) (in kWh) were computed using the equation:

$$Eleft({lat,lon,t} right) = CF(lat,lon,t) times C(lat,lon) times 8760$$


where C(lat,lon) represents the installed capacity at a given location, CF(lat,lon,t) is the CF at the location and 8760 defines the number of hours in a year.

Projected costs for each technology

The globally averaged price for PV decreased from $4.60/W in 2010 to $1.20/W in 201837. Pachouri et al.3 argued that these prices should continue to decline, projecting a decrease of 3% per year to 2024, 2% per year from 2024 to 2027, and 1% per year thereafter. For present purposes, we assume a range of prices for PV panels in 2040 varying from a low of $0.55/W to a high of $1.65/W. All of the costs quoted here are defined in terms of 2018 US dollars. Prices for onshore wind have also declined, dropping from $1.90/W in 2010 to $1.20/W in 201837. Pachouri et al.3 projected a more modest decrease for future prices in this case, 1% per year. For present purposes, we consider a range of costs for 2040 onshore wind installations varying from $0.98/W to $1.95/W, with higher prices, $1.30/W to $2.30/W, assigned for offshore facilities. Current trends would appear to favor the lower of the costs quoted here for all three applications. Accordingly, we elected to emphasize for purposes of the standard model in what follows the lower of the ranges of values indicated here ($0.55/W for solar PV, $0.98/W for onshore and $1.30/W for offshore wind)37. The low-cost projections for these renewables are consistent with cost estimates from NREL38. The sensitivity of results to the choice of costs will be discussed later and more extensively in the SI.

India has abundant reserves of coal, fourth largest in the world trailing only the US, Russia and China. The contemporary price for coal in India averages about $3.5/MMBTU39. We assume that this price is unlikely to change much by 2040 and adopt accordingly a reference future cost for coal of $3.6/MMBTU. It is more difficult to predict future prices for gas, given the sensitivity of prices for this commodity to vagaries of the international market. NITI Aayog and IEEI39 suggest a range for future prices from $6/MMBTU to $15/MMBTU. For present purposes, we adopt a value of $7.65/MMBTU, near the midpoint of this range. The sensitivity of conclusions to this choice will be discussed in what follows.

As indicated earlier, in seeking the least cost strategy to minimize future costs for electricity while organizing a significant shift from coal to wind and solar, we propose to allow for cost-effective expansions of the interregional transmission grid in addition to investments in storage. Estimated costs for expansion of the interregional transmission grid are summarized in Supplementary Table 2. These costs were defined by considering expenditures per unit of power for investments involved in development of the current grid40. The higher costs associated with specific interconnections reflect primarily the greater distances involved with these links.

Options for storage

A variety of options are available for storage of power. Mechanical systems include pumped hydro, compressed air, and flywheels. Chemical options refer mainly to batteries. Two considerations are involved in assigning relevant costs: the peak power capacity of the system (measured for example in kW), and the capacity of the system to store energy (measured specifically in kWh). A range of prices for different systems, adopted from Safaei and Keith41, is presented in Supplementary Table 3. For purposes of the standard model, we select the option identified as Medium Cost. The optimization model described below is charged with exploring the least cost option for any particular application, recognizing the distinctions between the power and energy capabilities of individual systems. Pumped hydro is responsible for the bulk of the 140 GW of power storage currently deployed globally and is likely to play an important role in the future also for India. Capital expenditures for construction of pumped hydro facilities are high, however, relative to costs for batteries42. Responding to the disparity in prices for capital investments in pumped hydro versus batteries, the current analysis concludes that batteries are likely to provide the option of choice for storage of power for India at least over the time interval considered here.

Optimization model for India’s energy system and its capacity expansion

The generation and transmission capacity expansion results for different levels of renewables were obtained based on a capacity expansion model optimizing jointly investment decisions and hourly system operations accounting for a full set of flexibility constraints. The model allows for potential deployment of defined renewable resources, for thermal generation, for energy storage and for upgrades in interregional transmission.

The decision variables for the energy system capacity expansion model (ESCEM) involve two components. For capacity investments, the decision variables account for invested capacities for each type of generation technology in each region, the capacity of storage deployed, and the capacity for transmission between different regions. For system operation, the decision variables allow for the available capacity and for the hourly dispatched output for each category of generation and storage for each region. The capacity available during the dispatch phase is interlinked with the investment decisions.

The objective of the ESCEM is to minimize the overall system cost, which includes two parts: (1) system annual operational costs, the sum of hourly fuel costs, start-up costs and operational costs for storage, thermal power, hydropower and nuclear power systems; and (2) amortized capacity investment costs, fixed O&M expenses and costs for the interregional network expansion.

The model considers a full set of constraints for the system operation. Hourly power balance as well as reserve constraints are incorporated for each region. Flexibility constraints for thermal units are also included with maximum and minimum generation limits defined, and with specification of ramping and minimum on/off time constraints. Operational constraints relating to energy storage are also considered based on different characteristics of storage technologies. Limitations on interregional power flow are incorporated in optimizing regional power exchange. Finally, renewable portfolio requirements are incorporated as an additional constraint.

To accelerate the calculation at such large scale, a novel flexibility method described in ref. 43 is employed to reduce the modeling complexity and improve the computational efficiency. The Units are grouped in the model with similar operational characteristics (same fuel type, similar nameplate capacity) to be dispatched based on aggregated power generation. There are six groups (categories) for each of the five regions in India and the total online capacity at each time interval is calculated by a combination of on-off status for all individual units in the group.

The mathematical formulation of the proposed optimization model is detailed further in the Supplementary Method and validation of the model is discussed in Supplementary Note 2.

Projecting India’s energy system in 2040

The proposed ESCEM accounts both for the expansion projected in power demand and the annual hourly operation for India’s energy system in 2040 covering five regions (East, North, South, West and Northeast), accounting for regionally distributed load, power plants (thermal, hydro, and nuclear), renewables (solar and onshore/offshore wind), possible energy storage systems, and exchanges for the interregional power grid. A detailed description of the energy system configuration in India is presented in the SI. Following results from Spencer et al.44, we scale the hourly power demand to 2040 for each region assuming a compound annual growth rate of 6.5%, which accounts for population and economic growth. Estimates of the hourly demand in 2016 were obtained from POSOCO16, and are indicated in Supplementary Fig. 5.

To project the role of coal in India’s future energy system, we consider technological improvements for coal-fired units in 2040 motivated by India’s Ministry of Power proposal to renovate and replace inefficient coal-fired units. Details of the updated configuration can be found in Supplementary Table 3. Other unit information (gas, hydro, and nuclear) is similarly derived from Ministry of Power proposals. Nuclear units are fixed; i.e., power output is time independent because the plant must be operational all of the time. Because of the relatively low cost and high flexibility of hydropower plants, we assume a fixed capacity of 55 GW based on the Indian government’s future investment plan for hydropower45. Locations for coal, gas and nuclear plants are indicated in Supplementary Fig. 3.

Costs for renewables are projected to decline significantly in the future, in response to technological improvements and benefits from learning experience. We consider fixed cost reduction rates of 15% for onshore, 35% for offshore, and 45% for solar PV from present-day values to 2040. To cover the most conservative and optimistic estimates for renewable investment costs, we consider two scenarios: a low-cost scenario based on the lowest renewable investment costs ($975 kW−1 for onshore wind, $1300 kW−1 for offshore wind, and $550 kW−1 for solar PV) and a high-cost scenario based on the highest renewable investment costs ($1955 kW−1 for onshore wind, $2300 kW−1 for offshore wind, and $1650 kW−1 for solar PV), following analyses of current renewable systems from IRENA37. Results derived from other future pathways are indicated in Supplementary Figs. 610.

As an important operational component of India’s energy system, capacity expansion for seven interregional transmission corridors (indicated in Fig. 2) is considered in the model. We assume that the number and location of corridors are fixed and the transmission capacity for a given corridor can be expanded according to an expansion factor, defined as the ratio of total current investment cost for a corridor to its capacity. The calculation process and expansion factors are presented in the Supplementary Method and Supplementary Table 2, respectively.

Total sector emissions of CO2 are aggregated from hourly emissions of all thermal generators. Emissions factors of CO2 for coal and gas-fired units are derived from the Ministry of Power46. Detailed description of emission factors and calculations of emissions for different scenarios are indicated in the Supplementary Note.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *