Methodology

Why a grade, not a date

Grid operators publish reinforcement schedules, but those dates slip. A specific promise (“connection in 2027”) is misleading in a congested area — the relevant question isn’t when a project can connect but how long relative to other areas. Speed to Power therefore reports a national percentile on an A–F scale — relative, and therefore stable as the whole market shifts.

Consumption vs feed-in

Every feeding area gets two independent grades:

• CConsumption — how quickly a new data centre, factory, heat-pump cluster, or EV hub can be connected.
• FFeed-in — how quickly a new solar farm, wind project, or battery in feed-in mode can be connected.

An area can be A on feed-in and F on consumption (or vice versa). The map shows one at a time; toggle via the Direction control at the top. For the Dutch grid terms behind these labels, see the glossary.

The two-layer trap

The Dutch grid has two layers: the DSO (Liander, Stedin, Enexis, Coteq, RENDO, Westland Infra) runs the medium-voltage network; TenneT sits above with high voltage. An area can have headroom at the DSO while TenneT is already saturated. When this happens the grade is capped at C until TenneT’s published relief year passes — at which point the cap drops and the grade recovers to the uncapped percentile.

The composite score

Per feeding area, four signals are blended into a 0–100 composite, then ranked into national percentiles:

The weights are fixed and applied identically to every area. The grade is the national percentile of the composite within each direction.

How the grade moves along the timeline

Dragging the timeline forward asks a single question of every region: which of the reinforcements the operators have published will have landed by this year?Two things can hold a region down, and each lifts on its own published year. The region’s own reinforcement clears the congestion signals that were dragging its composite down — the operator’s colour code and the waiting queue — and it is re-ranked against today’snational distribution, so a B in 2031 means the same thing as a B today. Separately, the TenneT cap above it drops on TenneT’s own relief year. A region only ever improves as you scrub forward.

Two deliberate conservatisms. We do not credit the megawatts a reinforcement adds — the operators do not publish them, so a region’s headroom signal is carried forward unchanged rather than invented; the improvement comes only from the congestion clearing. And a region with no published reinforcement year never improves on its own: we will not put a date on the map that the operator has not put in writing. These are operator estimates, not commitments — Dutch grid reinforcements slip by years routinely.

Stress inference (power-flow)

Alongside the published grades, Speed to Power runs a steady-state DC power-flow over a unified model of the network to estimate per-substation loading at 15-minute cadence. This drives the curtailment estimator and the Solar Workbench. It’s classification-grade — right roughly 75-85% of the time on the binary “is this substation stressed?” question, ±20-30% on absolute loading.

Accuracy is measured against published grid-operator congestion events on a rolling 30-day window. A high score means the model is tracking where real congestion occurs; a falling score means it needs recalibration.

The same load-flow powers the Congestion Calculator (per-area grid congestion) and the Stress Lens (inferred substation congestion); the Congestion Calculator methodology documents exactly how those per-area figures are derived.

Grid Drag and the Grid Drag Index

Grid Drag is a yes/no read of one thing: at the current queue state, can a battery at this node secure a firm connection, or only a non-firm one? A Dutch non-firm right — an ATR85 / TDTR — guarantees transport for only 85% of the hours in a year and lets the operator restrict the other 15% (1,314 hours) in the constrained direction, a day ahead. We answer the firm-vs-non-firm question by reading the operators’ own published congestion; that is the axis the model is good at (it rarely misses a node the operators already flag as full). How much a non-firm cap actually costs in earning hours is a downstream, node-specific question we only estimate indicatively — never a bankable number.

How much that costs depends on the node. Congestion is directional — the grid sets separate limits for drawing power off it (charging) and feeding power back (discharging) — and a battery uses both. Its normal cycle, charging in the midday solar surplus and discharging into the evening peak, often runs counter to the local constraint, which relieves congestion and means the cap barely bites. Where a node is constrained in both directions at once, like the Eemsdelta or the Sloegebied, the cap can land on the hours it most wanted to trade, and the bite is real. Grid Drag is the number that says which case a node is in.

It runs on the same inference model described above and inherits the same accuracy bands. What the model does well is detection — flagging whether a node sits on a congested or non-firm corridor, calibrated against the operators’ own published congestion. What it does not do is certify a magnitude: we rank corridors as more- or less-stressed as a directional signal, never as a per-line verdict or a bankable number. The IRR figure is a business-case model you parameterize — your MW/MWh and a market-spread assumption you set — run over the curtailment estimate, giving an indicative delta between free and grid-constrained dispatch, not a return you should bank.

The Grid Drag Index is a periodic aggregate across the announced and candidate Dutch battery pipeline, measured on that same detection axis: the share of sites that can only secure a non-firm / ATR85 connection at the current queue state. (A typical share of peak earning hours taken back is reported alongside it as an indicative, node-dependent read — not a severity score we stand behind.) It is published indicative and relative, with a snapshot date and a link back to this page every time. Treat it as a second opinion, not a single source of truth.

From data to grades

Data moves through three stages. First, the data from each source is kept exactly as published, for audit and change tracking. Next, it is unified into a single, consistent model of the network. Finally, that model is turned into the per-region figures shown on the site — every chart you see reads from this last stage, never from raw source data.

Accuracy bands

The A–F grade is a planning-grade classification of speed-to-power headroom — not a metering-grade reading or a live loading signal. It is inferred from public TSO/DSO data: no real-time SCADA, no N-1 contingency, no sub-substation granularity, and no bankable guarantees. The Netherlands is a single bidding zone, so price does not locate congestion. Use it to compare and shortlist regions, then validate locally before committing.

Every figure derived from the inference model carries this accuracy band. The bands are a hard requirement — no figure is published without them.

Not bankable. For commitments above €5M CAPEX, validate against the DSO-specific ATR85 contract template. More on the curtailment estimator.

Sources & update cadence

Every source behind the grade — the national capaciteitskaart, TenneT’s investment plan and relief years, the DSO waiting-list disclosures, and the feeder-area geometry — is listed on the Data sources page, with provenance, license, and the last successful fetch per pipeline. We keep that one catalogue rather than a second, drift-prone copy here.

Every new national snapshot recomputes all grades and re-prerenders the region pages. The last-updated stamp on every page comes from that snapshot.