Shota Furuya

Pieces日本語版

Solar and wind depend on the weather. How can supply be kept stable?

“Solar and wind depend on the weather, so won’t the lights go out?” is the most common plain-language doubt in any argument about expanding renewables. Yet countries and regions where variable renewable energy (VRE) already supplies 30–60% of electricity run their grids without blackouts. That fact alone disposes of the premise that VRE inevitably destabilises a power system.

A grid is a vast network whose demand and supply have always fluctuated: a dynamic system that keeps them balanced across seconds, minutes, hours, days and seasons. This piece covers (1) why variability is a surmountable problem, (2) how flexibility resources — storage, pumped hydro, demand response, interconnection, forecasting — actually work, and (3) what is hard in Japan and what the answers look like.

“Unstable” means the balancing problem — which means flexibility

Electricity cannot readily be stored, so consumption and generation must be matched in real time. Excess supply raises the frequency and a shortfall lowers it; step outside Japan’s tolerance of ±0.2–0.3 Hz around 50/60 Hz and protective relays trip, with the risk of cascading outages.

What matters here is that demand fluctuates constantly too. Waking up in the morning, the lunchtime peak, the evening return home with heating or cooling, the overnight trough — demand moves minute by minute. Large machines start and stop, temperatures swing, a popular broadcast draws everyone to their screens. Power systems were always designed to absorb this, balancing continuously with thermal output, pumped hydro, frequency regulation and reserves.

Renewable variability, in other words, is an extension of variability management that has always existed. No new mechanism has to be invented from scratch. What counts is the quantity and diversity of flexibility resources.

The IEA describes six phases of VRE integration, and holds that from phase 4 (VRE at roughly 25–50%) the combination of flexibility resources becomes essential. Japan’s solar and wind together were about 11% of generation in 2024 (PV about 10%), which places it in phase 2–3. Germany, meanwhile, met 62.7% of electricity consumption from renewables in 2024, Denmark about 56% from wind alone, and Spain about 56% from renewables across the year — so the engineering is already demonstrated.

Better forecasting, and geographical spread

The first flexibility is making generation predictable. Probabilistic forecasting that combines numerical weather prediction with machine learning has brought day-ahead errors for wind and irradiance down to a normalised mean absolute error of 3–5%, which sharply reduces the reserves that must be held and the cost of balancing power procured after gate closure.

The second is geographical smoothing. Spread generation over a wide area and passing clouds or local lulls cancel out; the relative variation of the combined output falls roughly with the square root of the area.

For Japan, operating across a wide-area grid from Hokkaido to Kyushu lets cloud in Kyushu be offset by sunshine in Tohoku. The capacity of the interregional links sets the limit on how much of that smoothing can be captured. The Hokkaido–Honshu link is being expanded from 900 MW to 2,000 MW towards FY2027, the eastern link between Tokyo and Chubu is being reinforced, and a large expansion between Honshu and Kyushu is planned for the 2030s.

The third is complementarity between solar and wind. Irradiance peaks in the daytime, is weaker in winter and stronger in summer; wind tends to blow harder at night and in winter and spring. Combined, they trace a supply curve closer to a baseload. Designing a portfolio around regional character — wind-led in Hokkaido and Tohoku, solar-led on the Pacific coast — is where stabilisation starts.

Storing energy: pumped hydro, grid batteries, hydrogen

The physical way to fill a gap between supply and demand is storage. Japan already has about 27.5 GW of pumped hydro, the third largest fleet in the world. It was built to shift surplus nuclear output from night to day; now the reverse is becoming the norm, shifting surplus daytime solar to the evening peak.

As ISEP’s analysis showed, 16.5 GW of renewable output was curtailed nationwide on 27 April 2025 while 3,800 MW of pumped hydro and 1,248 MW of downward range on LNG plants went unused. That is not a physical limit; it is a matter of operating rules and the rigidity of day-ahead plans.

The most dramatic change of recent years is in grid-scale batteries (BESS). BloombergNEF’s Lithium-Ion Battery Price Survey 2025 puts the weighted average pack price at US$108/kWh (about ¥16,000/kWh), down 20% year on year, with stationary packs at US$70/kWh (about ¥11,000/kWh), down 45% — for the first time the cheapest segment of all, below EV packs.

That has brought four-hour grid-scale BESS into the range where it earns its keep in Japan’s capacity and balancing markets. Connection applications for grid storage had expanded to about 60 GW by the end of 2024 on METI and OCCTO figures, with projects already cleared in the capacity market due to come online through 2026–27.

Over the longer run, absorbing seasonal swings calls for large-scale, long-duration storage: hydrogen, ammonia and synthetic fuels. For weeks of weak sun or still air, Europe is building power-to-gas designs that store green hydrogen in salt caverns and existing gas infrastructure and convert it back through fuel cells or hydrogen co-firing.

Using the demand side: DR, VPPs and sector coupling

Absorbing all variability on the supply side is inefficient. Building in demand response — shifting demand in time — delivers the same stability without adding capacity.

The characteristic form is the virtual power plant (VPP), which gathers consumer energy resources — home batteries, EVs, heat-pump water heaters, air conditioners, industrial BESS — through communications and algorithms and bids them into the market as if they were a single plant. In Japan, aggregators entered the balancing markets in earnest from FY2025, with ENERES, Tokyo Gas, Kansai Electric Power and NTT Anode Energy among those building businesses.

In homes, shifting heat-pump water heaters into the daytime to soak up surplus solar is becoming common. They were designed to heat water on cheap overnight power — a habit from the nuclear-centred era — whereas heating with surplus solar at midday and using the water in the evening now makes more sense. METI has issued guidance encouraging the daytime shift since 2024. Even an ordinary household is now an actor in grid stability.

EVs work the same way: through V2H and V2G, batteries of tens of kilowatt-hours join the grid as moving storage. Japan’s EV fleet is still small, but as light EVs and commercial vehicles electrify, it becomes a flexibility resource of gigawatt scale in the 2030s. In the context of distributed renewables, these are the core technologies of local microgrids and community energy.

That said, as Boyle et al. (2026) point out, VPPs and demand response involve remote control by aggregators, which easily creates relations of power that participants cannot see. The integration of renewables is not only a technical argument; it runs directly into energy justice — who controls what, and who carries the benefits and the burdens.

Downward range on thermal plants, and other existing resources

It may be surprising, but how flexibly existing thermal and nuclear plants are run is also a powerful answer to variability. An LNG combined-cycle plant can run at a minimum output of around 50% and change output within minutes, which is why North America and Europe treat them as quasi-balancing resources.

In Japan, though, capacity market rules, fuel contracts and operating conventions often mean LNG plants do not offer enough downward range. As for nuclear, Japanese plants essentially do not perform the load-following that France and Germany have long treated as routine; ISEP estimates that adjusting nuclear output during the low-demand months of spring and autumn would cut renewable curtailment by about 19%. Turn down thermal and nuclear before you turn off renewables — the priority dispatch rules say as much, and practice has not caught up. That gap is a structural problem in Japan.

Alongside this, non-firm connections — connection on condition that output is limited when the grid is congested — have been rolled out nationwide since 2023, squeezing the last of the spare capacity out of the network. In Hokkaido and Tohoku they have made additional connections of several gigawatts possible.

International comparison: stable operation, already demonstrated

Finally, the evidence on the fundamental question of whether a grid holds up at a high VRE share.

Denmark met its electricity demand in 2024 with 59% wind and more than 9% solar while keeping supply reliability (SAIDI) among the best in Europe; South Australia reached a 12-month average of 74% renewables in December 2024; Germany recorded 62.7% across 2024. Spain had a day in May 2024 when wind and solar alone briefly covered almost all demand, and Portugal ran on 100% renewables for six consecutive days on several occasions in 2024.

California (CAISO) is working its way through the duck curve created by large volumes of solar by building grid batteries fast: capacity grew 26-fold from 500 MW in 2020 to over 13 GW in 2024, and discharging into the evening peak has sharply reduced gas plant starts. These cases are living evidence that a VRE-led system and supply reliability are not a trade-off.

Summary

  • A power system has always absorbed variation from seconds to seasons; VRE variability can be handled with enough flexibility resources, of enough kinds.
  • Stabilisation rests on five pillars: (1) better forecasting and geographical spread, (2) storage through pumped hydro, grid batteries and hydrogen, (3) the demand side through DR, VPPs and sector coupling, (4) downward range on thermal plants and load-following nuclear, and (5) stronger interregional links.
  • Battery prices fell to US$108/kWh (−20%) in 2025, and US$70/kWh (−45%) for stationary packs on BNEF figures, while Japanese grid-storage connection applications expanded to about 60 GW.
  • On 27 April 2025 Japan curtailed 16.5 GW while 3,800 MW of pumped hydro and 1,248 MW of LNG downward range sat unused — a problem of rules and operation, not physics.
  • Germany at 62.7%, Denmark above 60%, South Australia at 74%, CAISO with 13 GW of batteries: a VRE-led system and high reliability go together, as the world has demonstrated.

“Renewables are unstable because they depend on the weather” is a premise more than a decade behind the facts. The real work is not technical but institutional: market design, operating rules, and a social design that brings regions and the demand side in.

References and data sources

System integration and flexibility

VRE shares achieved internationally

Japan’s grid and curtailment

Battery costs and markets

Pumped hydro, hydrogen and sector coupling

← Back to list