Electricity Demand Forecast Accuracy, 30 Day Window to 20 September 2026
How far off was the day-ahead demand forecast? Over each grid's latest 30 day hourly window, measured against reported demand: US grids (EIA): 53 grids, average error from 1.4% (ERCOT) to 33.5% (SPA), data to 20 September 2026; European bidding zones (ENTSO-E): 47 grids, average error from 1.1% (Czechia) to 58.2% (Bosnia and Herzegovina), data to 20 September 2026. The two groups are ranked separately. Each grid is measured over its own latest 30 day window of hourly readings, shown in the last column of the tables.
Key Facts
- How far off was the day-ahead demand forecast? Over each grid's latest 30 day hourly window, measured against reported demand: US grids (EIA): 53 grids, average error from 1.4% (ERCOT) to 33.5% (SPA), data to 20 September 2026; European bidding zones (ENTSO-E): 47 grids, average error from 1.1% (Czechia) to 58.2% (Bosnia and Herzegovina), data to 20 September 2026. The two groups are ranked separately.
- United States (EIA): the most accurate forecast was ERCOT, off by 1.4% on average, with 99% of hours within 5% (21 August 2026 to 20 September 2026).
- United States, among grids averaging 5,000 MW or more, PSCO had the largest average error, 24.2%.
- United States: 20 of 53 grids ran high on average by half a percent or more, 24 ran low by half a percent or more, and the rest were close to balanced.
- Europe (ENTSO-E): the most accurate forecast was Czechia, off by 1.1% on average, with 100% of hours within 5% (23 August 2026 to 20 September 2026).
- Europe, among grids averaging 5,000 MW or more, Netherlands had the largest average error, 26.8%.
- Europe: 21 of 47 grids ran high on average by half a percent or more, 12 ran low by half a percent or more, and the rest were close to balanced.
What Is a Day-Ahead Demand Forecast and Why Do Errors Matter?
Grid operators publish a forecast of how much electricity their area will use, hour by hour, the day before. Power plants are scheduled, fuel is bought, and wholesale electricity is traded against that forecast. Real demand then arrives, driven by weather, holidays and the many small choices of millions of people.
When the forecast runs too low, operators must find extra generation or imports at short notice, which is usually dearer and can strain the grid. When it runs too high, plants are committed that were not needed. Errors are largest when the weather departs from what was expected, and grids with a lot of rooftop solar are harder to forecast because part of the demand is hidden behind the meter.
Average error, in the tables below, is the mean of the absolute percentage gap between forecast and reported demand across all compared hours, so a high miss and a low miss both count. Bias is the signed average: above zero means the forecast was too high on average, below zero too low. The worst miss is the largest single-hour gap in megawatts.
United States: Forecast Accuracy by Grid
Operator data to 2026-09-20, 53 grids| Grid | Average error | Bias | Hours within 5% | Worst miss | Hours compared | Data to |
|---|---|---|---|---|---|---|
| ERCOT | 1.4% | -0.2% | 99% | 4,422 MW(6%, high) | 695 | 2026-09-20 |
| AVA | 1.6% | -0.0% | 98% | 142 MW(8%, low) | 719 | 2026-09-20 |
| BPA | 1.7% | +0.9% | 100% | 3,807 MW(116%, high) | 695 | 2026-09-20 |
| PGE | 1.9% | +1.0% | 92% | 367 MW(14%, high) | 719 | 2026-09-20 |
| CHPD | 2.0% | -0.5% | 96% | 23 MW(11%, low) | 719 | 2026-09-20 |
| TPWR | 2.0% | +1.3% | 95% | 38 MW(7%, high) | 719 | 2026-09-20 |
| NWMT | 2.0% | -0.6% | 93% | 335 MW(20%, low) | 719 | 2026-09-20 |
| MISO | 2.1% | +1.8% | 96% | 7,452 MW(7%, high) | 719 | 2026-09-20 |
| TVA | 2.2% | +1.0% | 94% | 4,310 MW(17%, high) | 719 | 2026-09-20 |
| BANC | 2.3% | +0.0% | 93% | 2,403 MW(52%, low) | 719 | 2026-09-20 |
| NYISO | 2.6% | -2.0% | 89% | 1,698 MW(8%, low) | 695 | 2026-09-20 |
| SOCO | 2.6% | +0.9% | 86% | 13,299 MW(47%, high) | 688 | 2026-09-20 |
| PACW | 2.6% | +0.8% | 86% | 263 MW(11%, high) | 719 | 2026-09-20 |
| SCL | 2.7% | -1.0% | 92% | 89 MW(10%, high) | 695 | 2026-09-20 |
| DUK | 2.8% | +0.6% | 84% | 3,613 MW(22%, high) | 719 | 2026-09-20 |
| FPL | 2.9% | -0.0% | 82% | 2,932 MW(14%, high) | 718 | 2026-09-20 |
| CPLE | 3.0% | -0.5% | 84% | 1,685 MW(18%, high) | 719 | 2026-09-20 |
| ISO-NE | 3.0% | -2.5% | 76% | 1,425 MW(11%, low) | 719 | 2026-09-20 |
| EPE | 3.2% | -1.9% | 82% | 272 MW(16%, low) | 695 | 2026-09-20 |
| TEC | 3.3% | -0.3% | 78% | 544 MW(16%, high) | 622 | 2026-09-20 |
| TAL | 3.4% | -1.2% | 80% | 79 MW(14%, low) | 719 | 2026-09-20 |
| DOPD | 3.4% | +2.2% | 76% | 34 MW(16%, high) | 695 | 2026-09-20 |
| TIDC | 3.4% | -0.4% | 80% | 98 MW(19%, high) | 717 | 2026-09-20 |
| CPLW | 3.5% | -2.6% | 72% | 107 MW(17%, high) | 719 | 2026-09-20 |
| GCPD | 3.6% | -3.4% | 76% | 61 MW(7%, low) | 681 | 2026-09-20 |
| PJM | 4.0% | -2.0% | 64% | 12,035 MW(10%, high) | 719 | 2026-09-20 |
| TEPC | 4.0% | +0.4% | 74% | 821 MW(55%, high) | 719 | 2026-09-20 |
| AECI | 4.2% | +0.4% | 65% | 446 MW(12%, low) | 719 | 2026-09-20 |
| JEA | 4.4% | +2.3% | 66% | 468 MW(25%, high) | 719 | 2026-09-20 |
| FMPP | 4.4% | +2.5% | 67% | 688 MW(28%, high) | 719 | 2026-09-20 |
| SC | 4.8% | +3.6% | 66% | 3,274 MW(299%, high) | 671 | 2026-09-19 |
| PNM | 4.9% | -4.4% | 48% | 222 MW(8%, low) | 719 | 2026-09-20 |
| IID | 5.1% | +1.5% | 70% | 394 MW(116%, high) | 718 | 2026-09-20 |
| SCEG | 5.4% | -4.7% | 53% | 586 MW(16%, low) | 719 | 2026-09-20 |
| IPCO | 5.7% | -0.5% | 52% | 638 MW(29%, high) | 719 | 2026-09-20 |
| BHBA | 7.1% | -6.9% | 29% | 171 MW(16%, low) | 719 | 2026-09-20 |
| SPP | 7.2% | +7.0% | 46% | 15,947 MW(40%, high) | 688 | 2026-09-20 |
| LDWP | 7.4% | -6.2% | 36% | 1,441 MW(28%, low) | 718 | 2026-09-20 |
| NEVP | 7.8% | -7.5% | 38% | 1,609 MW(20%, low) | 717 | 2026-09-20 |
| PACE | 7.8% | +5.3% | 43% | 2,006 MW(28%, high) | 719 | 2026-09-20 |
| LGEE | 7.8% | -7.8% | 14% | 852 MW(15%, low) | 717 | 2026-09-20 |
| CAISO | 8.6% | -8.0% | 54% | 11,666 MW(30%, low) | 719 | 2026-09-20 |
| HST | 9.4% | +5.5% | 44% | 46 MW(61%, high) | 695 | 2026-09-20 |
| AZPS | 9.8% | -4.8% | 29% | 4,278 MW(125%, high) | 714 | 2026-09-20 |
| GVL | 10.7% | -9.5% | 16% | 95 MW(25%, low) | 626 | 2026-09-20 |
| SRP | 11.5% | +10.1% | 41% | 2,977 MW(82%, high) | 719 | 2026-09-20 |
| SEC | 11.8% | +4.9% | 38% | 345 MW(53%, low) | 717 | 2026-09-20 |
| WALC | 14.0% | -0.7% | 22% | 627 MW(78%, high) | 688 | 2026-09-20 |
| SWPW | 14.5% | +14.1% | 33% | 1,418 MW(46%, high) | 719 | 2026-09-20 |
| FPC | 23.3% | -23.3% | 0% | 3,369 MW(29%, low) | 719 | 2026-09-20 |
| PSCO | 24.2% | -24.2% | 0% | 3,266 MW(40%, low) | 719 | 2026-09-20 |
| PSEI | 31.8% | -31.7% | 3% | 1,671 MW(64%, low) | 719 | 2026-09-20 |
| SPA | 33.5% | +23.8% | 12% | 28 MW(280%, high) | 600 | 2026-09-20 |
Ranked from the smallest average error. Bias is signed: plus means the forecast ran above reported demand. Worst miss shows its size, its percentage of the reading and its direction. Each grid uses only the hours in which both a forecast and a positive demand reading exist; the latest hours, for which the forecast exists but demand is not yet reported, are left out.
Source: U.S. Energy Information Administration (EIA). (US Government work, public domain) Data as of 2026-09-20. Accuracy is computed by World Power Monitor from the operator's hourly forecast and reported demand.
Europe: Forecast Accuracy by Grid
Operator data to 2026-09-20, 47 grids| Grid | Average error | Bias | Hours within 5% | Worst miss | Hours compared | Data to |
|---|---|---|---|---|---|---|
| Czechia | 1.1% | -0.3% | 100% | 313 MW(4%, low) | 686 | 2026-09-20 |
| Spain | 1.2% | +0.2% | 100% | 1,446 MW(5%, high) | 685 | 2026-09-20 |
| France | 1.4% | -0.1% | 100% | 2,372 MW(5%, high) | 687 | 2026-09-20 |
| Portugal | 1.5% | -0.2% | 98% | 521 MW(7%, low) | 686 | 2026-09-20 |
| Norway South-West (NO2) | 1.8% | -1.3% | 99% | 249 MW(6%, low) | 683 | 2026-09-20 |
| Norway Central (NO3) | 2.0% | -0.6% | 96% | 209 MW(9%, high) | 684 | 2026-09-20 |
| Denmark West (DK1) | 2.1% | +0.2% | 91% | 517 MW(16%, low) | 685 | 2026-09-20 |
| Poland | 2.1% | +0.8% | 93% | 1,846 MW(10%, high) | 686 | 2026-09-20 |
| Slovakia | 2.1% | +0.5% | 92% | 298 MW(10%, low) | 687 | 2026-09-20 |
| Denmark East (DK2) | 2.2% | +0.4% | 90% | 235 MW(18%, high) | 685 | 2026-09-20 |
| Austria | 2.2% | +0.3% | 91% | 1,330 MW(16%, low) | 685 | 2026-09-20 |
| Norway South-East (NO1) | 2.5% | -1.6% | 91% | 303 MW(9%, low) | 684 | 2026-09-20 |
| Switzerland | 2.5% | +0.1% | 85% | 1,555 MW(21%, low) | 686 | 2026-09-20 |
| Finland | 2.6% | +0.4% | 86% | 1,249 MW(16%, high) | 688 | 2026-09-20 |
| Germany and Luxembourg | 2.7% | -0.1% | 86% | 5,247 MW(11%, high) | 672 | 2026-09-19 |
| Norway North (NO4) | 2.7% | -1.6% | 90% | 291 MW(13%, low) | 684 | 2026-09-20 |
| Sweden South-Central (SE3) | 2.8% | -2.3% | 86% | 995 MW(11%, low) | 686 | 2026-09-20 |
| Belgium | 3.1% | -0.2% | 82% | 1,262 MW(12%, high) | 687 | 2026-09-20 |
| Romania | 3.4% | +0.8% | 81% | 1,130 MW(31%, high) | 688 | 2026-09-20 |
| Croatia | 3.4% | +1.9% | 80% | 391 MW(17%, low) | 686 | 2026-09-20 |
| Georgia | 3.5% | +0.5% | 78% | 267 MW(17%, high) | 664 | 2026-09-20 |
| Norway West (NO5) | 3.5% | -0.1% | 74% | 215 MW(14%, high) | 684 | 2026-09-20 |
| Serbia | 3.6% | +0.9% | 75% | 898 MW(24%, low) | 686 | 2026-09-20 |
| Italy North | 3.6% | -0.7% | 72% | 3,106 MW(13%, low) | 686 | 2026-09-20 |
| Hungary | 3.9% | -2.6% | 71% | 533 MW(12%, low) | 686 | 2026-09-20 |
| Italy Centre-North | 4.0% | +0.8% | 72% | 650 MW(17%, low) | 686 | 2026-09-20 |
| Greece | 4.0% | +2.6% | 67% | 1,143 MW(17%, high) | 687 | 2026-09-20 |
| Albania | 4.2% | +1.9% | 71% | 453 MW(57%, high) | 711 | 2026-09-20 |
| Italy Sicily | 4.4% | -0.1% | 65% | 545 MW(15%, low) | 686 | 2026-09-20 |
| Bulgaria | 4.5% | +3.2% | 77% | 1,547 MW(57%, high) | 683 | 2026-09-20 |
| Kosovo | 4.7% | +4.1% | 67% | 247 MW(68%, high) | 686 | 2026-09-20 |
| Italy Centre-South | 4.7% | -1.9% | 61% | 1,508 MW(21%, low) | 686 | 2026-09-20 |
| Slovenia | 5.7% | +0.1% | 67% | 812 MW(162%, high) | 686 | 2026-09-20 |
| Cyprus | 5.8% | -2.3% | 52% | 349 MW(45%, high) | 687 | 2026-09-20 |
| Italy Sardinia | 5.9% | +3.7% | 55% | 244 MW(35%, high) | 686 | 2026-09-20 |
| Latvia | 6.2% | +1.8% | 51% | 242 MW(48%, high) | 687 | 2026-09-20 |
| Italy South | 7.5% | -1.8% | 35% | 983 MW(22%, low) | 686 | 2026-09-20 |
| Sweden South (SE4) | 8.1% | +7.3% | 31% | 570 MW(30%, high) | 686 | 2026-09-20 |
| Italy Calabria | 8.6% | +0.9% | 33% | 271 MW(21%, low) | 686 | 2026-09-20 |
| North Macedonia | 8.7% | -2.3% | 43% | 158 MW(26%, low) | 517 | 2026-09-19 |
| Montenegro | 9.1% | +3.9% | 45% | 211 MW(67%, high) | 672 | 2026-09-19 |
| Sweden North-Central (SE2) | 9.6% | +3.4% | 36% | 682 MW(84%, high) | 686 | 2026-09-20 |
| Sweden North (SE1) | 11.8% | +7.1% | 31% | 587 MW(34%, low) | 686 | 2026-09-20 |
| Estonia | 13.4% | +10.5% | 31% | 472 MW(104%, high) | 685 | 2026-09-20 |
| Lithuania | 14.1% | -13.9% | 9% | 661 MW(38%, low) | 684 | 2026-09-20 |
| Netherlands | 26.8% | +25.2% | 13% | 8,217 MW(281%, high) | 673 | 2026-09-20 |
| Bosnia and Herzegovina | 58.2% | +57.8% | 6% | 1,148 MW(356%, high) | 670 | 2026-09-20 |
Ranked from the smallest average error. Bias is signed: plus means the forecast ran above reported demand. Worst miss shows its size, its percentage of the reading and its direction. Each grid uses only the hours in which both a forecast and a positive demand reading exist; the latest hours, for which the forecast exists but demand is not yet reported, are left out.
Source: ENTSO-E Transparency Platform (transparency.entsoe.eu), CC BY 4.0. Aggregated and converted by World Power Monitor; ENTSO-E does not endorse this site. (ENTSO-E Terms of Use plus CC BY 4.0 for listed data items) Data as of 2026-09-20. Accuracy is computed by World Power Monitor from the operator's hourly forecast and reported demand.
How These Figures Are Built
- Source: the last 30 days of hourly demand and the operator's own day-ahead demand forecast. US grids come from EIA and European bidding zones from ENTSO-E; they are ranked separately and are not combined with each other or with Ember statistics.
- Error for an hour is (forecast minus reported demand) divided by reported demand. Hours with a missing or zero demand or forecast value, or a value under 30% or over 3 times the grid's own median for the window (a partial or garbled report), are ignored. A grid needs at least 168 compared hours (seven days) to appear.
- Some grids show a steady gap in one direction (for example a forecast that runs well above reported demand every day); that can come from a difference in what the two series cover rather than poor forecasting, and this page does not tell the two apart. The forecast is whatever the operator last published for the hour, so it may be refreshed during the day for some operators and stay fixed at the day-ahead value for others. This page does not check which.
- Reported demand is preliminary and can be revised. A short window is a small sample: one heat wave or holiday can move a grid's average error a lot, and the figures are not a long run measure of forecast skill.
- Grids with a forecast series but fewer than 168 usable hours, and grids that publish no forecast, are left out. A page notes when a grid's window has fewer than 600 of 720 hours compared.
- See methodology and data quality for known gaps in the operator series.