Energy storage and firm clean power portfolios

Lesson 5 explained how data centre owners buy renewable energy and why the industry is moving from annual to hourly matching of consumption with clean generation. A data centre runs around the clock, while solar produces only by day and wind when the wind blows. Filling the gaps requires storage to move energy in time and firm clean resources that produce whenever needed. This lesson examines storage technologies and their roles, how storage earns value and is accredited for capacity, how to design a portfolio that matches a constant load hour by hour, why the last few percent of matching are so expensive, and how storage and firm resources are contracted.

Storage technologies

For data centres the most relevant today are lithium-ion batteries, at the facility (DCE-340) or on the grid, and contracts with grid-scale storage of various durations. A battery is described by its power rating (MW) and energy capacity (MWh); a 100 MW, 400 MWh battery can deliver 100 MW for four hours. Usable capacity is less than nameplate to protect cell life, and it degrades with cycling and time, typically to 70% to 80% over 10 to 20 years, so contracts specify augmentation (adding cells) to maintain capacity.

How storage earns value

• Energy arbitrage: charging when prices are low (often midday with abundant solar) and discharging when they are high (evening peaks).
• Capacity: counting toward resource adequacy requirements, paid through capacity markets or contracts (Lesson 2).
• Ancillary services: frequency regulation and reserves, which batteries provide very well because of their fast response.
• Shaping renewable contracts: storing midday solar for evening use, converting an intermittent profile into a firmer one.
• Transmission and distribution deferral: relieving a constraint in a few peak hours, delaying an upgrade.
• Backup and flexibility for the customer (Lesson 4).

Capacity accreditation

How much a storage resource counts toward reliability depends on its duration relative to the length of the system's stress periods. A 4-hour battery covers a short evening peak almost fully, but contributes less when stress periods last longer, as in multi-day winter cold snaps. System operators increasingly accredit resources by effective load carrying capability (ELCC): the additional load a resource allows the system to serve at the same reliability. ELCC values for 4-hour batteries are high when few are installed and fall as more are added, because each additional battery shifts the remaining stress to longer periods. Solar's ELCC falls similarly as its share grows, while firm resources such as nuclear, geothermal and gas keep high accreditation (adjusted for outage rates and, for gas, fuel supply risk in winter).

Designing a 24/7 clean energy portfolio

The hourly carbon-free energy (CFE) score (Lesson 5) measures the share of each hour's consumption matched by clean generation in that hour. For a constant data centre load, achieving a high score requires resources whose output fills different hours:

import numpy as np

rng = np.random.default_rng(5)
h = np.arange(8760); day = h // 24; hr = h % 24
load = np.full(8760, 100.0)                                        # flat 100 MW data centre
season = 0.75 + 0.25 * np.cos(2 * np.pi * (day - 172) / 365)
cloud = np.repeat(rng.uniform(0.4, 1.0, 365), 24)
solar = np.clip(np.sin(np.pi * (hr - 6) / 12), 0, None) * season * cloud     # per MW of capacity
w = np.zeros(8760); x = 0.0
for t in range(8760):                                               # wind: slowly varying random process
    x = 0.97 * x + rng.normal(0, 0.12); w[t] = x
wind = np.clip(0.38 + 0.3 * w + 0.08 * np.cos(2 * np.pi * (day - 15) / 365), 0, 1)

def cfe(sol_mw, wind_mw, batt_mw=0, batt_h=4, firm_mw=0):
    gen = sol_mw * solar + wind_mw * wind + firm_mw
    soc, cap, served = 0.0, batt_mw * batt_h, 0.0
    for t in range(8760):
        surplus = gen[t] - load[t]
        if surplus > 0:
            soc = min(cap, soc + min(surplus, batt_mw) * 0.9); served += load[t]
        else:
            d = min(-surplus, batt_mw, soc); soc -= d; served += gen[t] + d
    return served / load.sum(), gen.sum() / load.sum()

for label, args in (("solar 300 MW", (300, 0)), ("solar 150 + wind 200 MW", (150, 200)),
                    ("solar 150 + wind 200 + 100 MW x 4 h battery", (150, 200, 100, 4)),
                    ("solar 150 + wind 200 + battery + 40 MW firm clean", (150, 200, 100, 4, 40))):
    score, vol = cfe(*args)
    print(f"{label:50s}: hourly CFE {score:5.1%}, annual volume {vol:5.1%} of load")

solar 300 MW                                      : hourly CFE 37.0%, annual volume 49.1% of load
solar 150 + wind 200 MW                           : hourly CFE 85.5%, annual volume 106.4% of load
solar 150 + wind 200 + 100 MW x 4 h battery       : hourly CFE 91.7%, annual volume 106.4% of load
solar 150 + wind 200 + battery + 40 MW firm clean : hourly CFE 99.5%, annual volume 146.4% of load

Solar alone, even at three times the load's capacity, matches only about 37% of hours: it produces nothing at night, and its midday surplus is wasted without storage. Combining solar and wind, with slightly more annual energy than the load consumes, raises the score to about 85%. A 4-hour battery adds several points by moving midday surplus into the evening. Closing the remaining gap, the calm nights and winter weeks when neither solar nor wind produces, takes firm clean power: adding 40 MW of round-the-clock clean supply reaches over 99%, but the portfolio then produces 46% more energy than the load uses over the year, much of it surplus sold back to the grid.

This pattern is general: the cost of each additional percentage point of hourly matching rises steeply above roughly 80% to 90%, because the last hours to be filled are the longest, rarest gaps. Studies of 24/7 procurement find that firm clean resources (nuclear, geothermal, hydro, and in future long-duration storage or gas with carbon capture) greatly reduce the cost of very high matching compared with overbuilding solar, wind and short-duration batteries.

Where the clean energy is: location and deliverability

Hourly matching credibility also depends on location. A clean generator that is not deliverable to the data centre's grid region, because of transmission constraints, does not displace fossil generation serving the data centre in that hour. Emerging standards for hourly accounting therefore require resources to be in the same grid region (or with demonstrated deliverability), and emphasize additionality: new resources that would not otherwise have been built. Proposed revisions to the Greenhouse Gas Protocol's Scope 2 guidance move in this direction, which would change how many current procurement claims are counted.

Contracting for storage and firm power

• Hybrid PPAs: a renewable project paired with storage delivers a shaped or firmed profile, priced as a single contract.
• Tolling agreements: the buyer pays a fixed capacity fee for the right to charge and discharge a storage facility as it chooses, taking the market risk and reward.
• Capacity contracts or resource adequacy agreements: payments for accredited capacity without energy dispatch rights.
• Firm clean energy contracts: long-term agreements with nuclear plants (including restarts and uprates), geothermal developers or hydro utilities, often at premium prices reflecting their round-the-clock value (Lesson 10).
• Utility clean transition tariffs: tariffs, offered by some utilities, under which large customers pay for specific new clean firm resources on the utility system, combining the utility's grid operations with customer-funded procurement.

The data centre owner's choice among these depends on whether it wants financial exposure to market prices, physical firmness, or simply credible hourly matching, and on what its local market and regulators allow.