Define the flexibility need before selecting an asset
Power systems must balance supply and demand continuously while maintaining frequency, voltage, thermal limits, reserves, and recovery capability. Flexibility is the ability to change injection or withdrawal when and where the system needs it. The product is not a battery, gas turbine, interconnector, or responsive load; it is a controlled change with a response time, duration, ramp rate, location, availability, rebound, and state-restoration requirement. Procurement that begins with a technology capacity target risks buying a resource that cannot solve the binding event.
Needs should be mapped by timescale. Inertial and fast-frequency behavior acts in fractions of a second to seconds. Regulation and reserves operate over seconds to minutes. Ramping, peak shifting, and congestion management span minutes to hours. Multi-day wind or solar deficits, hydrological variation, fuel security, and seasonal balance require longer-duration resources or portfolios. IRENA estimates that global daily flexibility needs in 2030 will be around three times their 2019 level under its assessed transition pathways, rising to roughly ten times by 2050. These are modelled system needs, not universal storage-duration prescriptions.
Batteries are fast and modular, but duration still matters
Utility-scale batteries can provide rapid balancing, ancillary services, capacity, local congestion relief, renewable shifting, and black-start support when designed and contracted accordingly. The IEA estimated 124 GW of installed utility-scale battery power at the end of 2024 after a record 63 GW was added that year. Project costs fell about 40% in 2024 to around $150 per kilowatt-hour, although that global figure masks chemistry, duration, interconnection, civil works, financing, and safety differences. A power rating without energy capacity and usable state-of-charge policy cannot describe the service.
Evaluation should report megawatts, megawatt-hours, discharge duration at rated power, round-trip efficiency at representative duty cycles, degradation, auxiliary load, ambient-temperature effects, response time, availability, warranted throughput, and capacity retained through the contract. Simulate the actual event distribution rather than one perfect daily cycle. A two-hour battery may reduce an evening ramp yet contribute little during a multi-day deficit. Conversely, procuring long duration for a local second-by-second constraint can be unnecessarily expensive. Hybrid portfolios can assign fast response to batteries and sustained energy to demand, interconnection, reservoirs, thermal storage, or firm low-emissions resources.
Demand is an underused grid resource
Demand response changes or shifts consumption in response to operational or market signals. The IEA estimates that only around 100 GW was utilized globally in 2024, including roughly 75 GW in industry and 30 GW in buildings, with rounding across categories. For context, it identifies about 160 GW of peak electricity demand in aluminium production and around 600 GW from residential air conditioning, while noting that only a small or marginal share is used flexibly. Those gross loads are not all technically or economically shiftable; they define an opportunity space that requires equipment, controls, incentives, and service constraints.
A credible demand resource needs a baseline method, telemetry, dispatch test, customer override, rebound model, persistence estimate, and measurement of service quality. A cooling load can pre-cool a building, but thermal comfort and humidity set limits. Cold storage can shift compressors, but product temperature is the constraint. Industrial batches may move, yet production schedules and restart costs matter. Aggregators should be paid for measured availability and response, with penalties symmetric enough to make the resource dependable. Avoided grid cost should be calculated at the constrained location and hour, not assigned as a generic benefit to every flexible kilowatt.
Connection queues reveal a network and governance problem
More than 2,500 GW of generation, storage, and large-load projects were stalled in grid connection queues worldwide in the IEA's 2026 assessment. Queue capacity is not equivalent to shovel-ready supply; projects can be duplicated, speculative, or awaiting permits and finance. Still, the scale indicates that building generation faster than networks and interconnection processes can absorb it creates stranded options. The IEA estimates annual grid investment would need to rise roughly 50% from about $400 billion to meet forecast electricity demand through 2030. That is a modeled investment requirement, not a project budget.
Faster connection does not always mean new lines first. Dynamic line ratings, advanced conductors, topology optimization, grid-enhancing technologies, non-firm connections, co-located storage, improved outage scheduling, and transparent queue reform can unlock capacity sooner. The IEA estimates such measures could connect roughly 1,200 to 1,600 GW of advanced-stage queued projects, subject to its assumptions. Each intervention needs a counterfactual and operational limit. A non-firm connection, for example, transfers some curtailment risk to the project; it is valuable only if the rule, forecast, compensation, and maximum interruption are clear.
Procure a portfolio against observed events
A flexibility plan should start with chronological production-cost and network studies calibrated to measured load, weather, outages, and generator behavior. Model sub-hourly constraints and multiple weather years, then stress fuel interruption, transmission outage, heat waves, drought, and correlated low-renewable periods. Candidate portfolios should include grid expansion, interconnection, batteries of several durations, pumped storage, flexible generation, demand response, thermal storage, electric-vehicle charging, and curtailment. Curtailment is not automatically failure: small, deliberate volumes can be cheaper than infrastructure sized for rare output peaks.
Compare portfolios using expected unserved energy, loss-of-load risk, reserve shortfall, congestion, renewable curtailment, emissions, system cost, dependence on critical materials or fuels, construction lead time, and performance under adverse scenarios. Location-specific distribution constraints must survive aggregation. Value stacking should avoid counting the same megawatt as simultaneously available for incompatible services. Market design is part of the physical solution: products need settlement intervals, locational signals, access for aggregated demand and storage, and rules that reward availability during scarcity rather than installed nameplate alone.
Publish operational evidence, not transition theatre
The share of global generation from solar PV and wind is forecast by the IEA to rise from about 17% to 27% by 2030. IRENA reports that renewable power additions reached 692 GW in 2025, but capacity growth and secure integration are separate achievements. A public flexibility scorecard should report hourly curtailment, connection delays, congestion costs, battery availability and duration, demand-response tests and delivery, reserve scarcity, negative-price hours, outage performance, grid investment, and restoration results. Values should be normalized where useful but retain system totals and definitions.
Claims should distinguish historical observations from forecasts and scenario requirements. The IEA's 1,500 GW storage requirement by 2030 belongs to its net-zero pathway for supporting the renewable tripling goal; it is not a forecast and not a recommendation that every system procure the same mix. The research objective is to identify the lowest-risk portfolio that meets a stated reliability and emissions constraint under local conditions. A world-class flexible grid is not the one with the largest battery announcement. It is the one that exposes its needs, procures the right response across timescales, and verifies delivery during real stress.
Scope and limitations
IEA and IRENA global values combine systems with different definitions, market designs, and reporting quality. Queue capacity includes projects with varying maturity. Battery cost and capacity figures do not specify duration or local balance-of-system costs. Forecasts and transition pathways are conditional on policy, technology, demand, fuel prices, weather, and model assumptions; they are not observed outcomes. System planning requires local chronological and network data.
References
Source review: 20 August 2026. Quantitative values retain their original definitions, periods, and boundaries.
- 01Electricity 2026: Executive Summary
International Energy Agency · 2026
www.iea.org ↗ - 02Electricity 2026: Flexibility
International Energy Agency · 2026
www.iea.org ↗ - 03Flexibility for a Secure and Affordable Power Sector Transformation
International Renewable Energy Agency · 2026
www.irena.org ↗ - 04Batteries and Secure Energy Transitions: Executive Summary
International Energy Agency · 2024
www.iea.org ↗ - 05Renewable Capacity Statistics 2026
International Renewable Energy Agency · 2026
www.irena.org ↗