targetfev / AutoGluon ↗Global Energy Forecasting Competition (GEFCom 2014)
GEFCom 2014 (proenfo_gfc14) provides 17,520 hourly observations across a single electrical zone, targeting a 168-hour (7-day ahead) forecast of power grid demand. Conditioned on exogenous air temperature, the benchmark tests medium-term multi-seasonal grid scheduling and dispatch optimization.
Business Objective: Global Energy Forecasting Competition (GEFCom 2014)
01Business Context
Transmission system operators (TSOs) and energy trading desks operate on weekly commitment cycles, scheduling baseload and peaker generation 168 hours ahead to ensure grid stability and clear day-ahead power markets.
02Analytical Objective
Forecast hourly zonal electrical load across a full 168-hour ahead horizon using historical consumption telemetry and known future ambient air temperatures.
03Economic & Decision Impact
Under-forecasting risks reserve margin depletion, localized brownouts, and costly spot-market imbalance penalties. Over-forecasting forces unnecessary thermal plant spinning reserves, curtailment of renewable generation, and elevated operating expenditures.
Classical univariate baselines (AutoARIMA, ETS) struggle over the long 168-hour horizon due to phase error accumulation across multi-day diurnal cycles. Feature-engineered LightGBM models with temperature lag interactions excel, while zero-shot foundation models such as Mistral's Timestral and MOIRAI capture complex non-linear load-temperature dynamics without manual feature crafting.
Target & Feature Column Definitions
Exhaustive business definitions, measurement units, roles, and target variables across all 4 columns.
| Col | Variable / Feature Name | Role | Type | Unit / Scale | Description & Business Meaning |
|---|---|---|---|---|---|
| C | targetZonal Electrical Load | TARGET | NUM | MW | Total hourly electric power consumption observed across the regional utility zone. |
| A | idSeries Identifier | SERIES ID | CAT | — | Unique categorical identifier denoting the electrical zonal load time series. |
| B | timestampTimestamp | TIME | TIME | — | Hourly observation date and time in ISO 8601 format. |
| D | airtemperatureAir Temperature | Feature | NUM | deg F | Observed and forecasted ambient surface air temperature influencing thermal building demand. |
Interactive Data Table
Explore rows, feature values, and target labels for Global Energy Forecasting Competition (GEFCom 2014).
Foundation Models vs Traditional ML & Baselines (41 Methods)
Empirical evaluation directly from the AutoGluon fev benchmark suite evaluating zero-shot foundation models against deep learning, gradient-boosted trees, and statistical baselines across 100 evaluation tasks.
| Rank | Method / Model | Family | Organization | SQL Win Rate?Average win rate (%) against all competing models across the benchmark tasks under Scaled Quantile Loss. | SQL Skill?Normalized skill score relative to the Seasonal Naive baseline (0% = baseline, higher is better). | MASE Win Rate?Average win rate (%) under Mean Absolute Scaled Error. | Latency?Median end-to-end execution time per 100 time series in seconds. | Link |
|---|---|---|---|---|---|---|---|---|
| #1 | Timestral-209M | Foundation Model | Mistral AI | 87.7% | +50.4% | 86.0% | 0.51s | Repo ↗ |
| #2 | Timestral-123M (Carla Engine) | Foundation Model | Mistral AI | 86.8% | +50.2% | 85.0% | 0.32s | Repo ↗ |
| #3 | TimeRSI | Closed-Source API | GaTech & SamayLabs | 84.8% | +51.2% | 83.8% | 2512s | Repo ↗ |
| #4 | t0-gamma | Foundation Model | The Forecasting Company | 84.4% | +49.4% | 81.8% | 4.38s | Repo ↗ |
| #5 | Timestral-28M | Foundation Model | Mistral AI | 79.2% | +48.4% | 77.0% | 0.11s | Repo ↗ |
| #6 | timesfm-3 | Foundation Model | 78.5% | +49.0% | 77.0% | 3.41s | Repo ↗ | |
| #7 | ChakraTS | Baseline | Community | 78.2% | +48.1% | 79.5% | 14.22s | Repo ↗ |
| #8 | Lingjiang2.0 | Baseline | Community | 73.9% | +47.0% | 71.3% | 8.89s | Repo ↗ |
| #9 | Chronos-2 | Foundation Model | AWS | 70.4% | +47.3% | 68.4% | 0.84s | Repo ↗ |
| #10 | timeh2o-v1-tiny | Foundation Model | H2O.ai | 69.9% | +45.5% | 67.0% | 0.19s | Repo ↗ |
| #11 | Timestral-5M | Foundation Model | Mistral AI | 69.0% | +46.2% | 67.7% | 40ms | Repo ↗ |
| #12 | Toto-2.0-2.5B | Baseline | Community | 67.5% | +44.4% | 65.7% | 5.58s | Repo ↗ |
| #13 | t0-beta | Foundation Model | The Forecasting Company | 67.4% | +46.6% | 64.5% | 2.29s | Repo ↗ |
| #14 | Toto-2.0-1B | Baseline | Community | 66.4% | +44.4% | 65.5% | 2.48s | Repo ↗ |
| #15 | TiRex-2 | Foundation Model | NX-AI | 65.8% | +45.5% | 63.4% | 0.27s | Repo ↗ |
| #16 | Toto-2.0-313m | Baseline | Community | 65.1% | +44.2% | 63.5% | 1.06s | Repo ↗ |
| #17 | TiRex | Foundation Model | NX-AI | 58.0% | +42.6% | 54.1% | 0.24s | Repo ↗ |
| #18 | Toto-2.0-22m | Baseline | Community | 57.6% | +42.9% | 56.0% | 0.25s | Repo ↗ |
| #19 | TimesFM-2.5 | Baseline | Community | 54.6% | +42.2% | 53.5% | 1.87s | Repo ↗ |
| #20 | TabPFN-TS-3 | Foundation Model | Prior Labs | 52.4% | +43.1% | 51.3% | 235s | Repo ↗ |
| #21 | TS-ICL | Foundation Model | EDF-Lab | 51.2% | +43.1% | 49.8% | 2.49s | Repo ↗ |
| #22 | FlowState | Foundation Model | IBM | 49.3% | +39.9% | 49.9% | 2.29s | Repo ↗ |
| #23 | citras-fm | Foundation Model | Hitachi | 48.8% | +41.2% | 46.9% | 0.60s | Repo ↗ |
| #24 | Toto-1.0 | Baseline | Community | 47.5% | +40.8% | 45.9% | 22.35s | Repo ↗ |
| #25 | TabPFN-TS | Foundation Model | Prior Labs | 47.3% | +41.5% | 45.4% | 109s | Repo ↗ |
| #26 | Toto-2.0-4m | Baseline | Community | 46.2% | +40.7% | 43.6% | 0.20s | Repo ↗ |
| #27 | Moirai-2.0 | Baseline | Community | 41.9% | +39.3% | 40.5% | 0.35s | Repo ↗ |
| #28 | Chronos-Bolt | Foundation Model | AWS | 41.4% | +38.9% | 39.3% | 0.26s | Repo ↗ |
| #29 | TFT | Deep Learning | — | 31.7% | +32.2% | 29.2% | 829s | Repo ↗ |
| #30 | Stat. Ensemble | Baseline | Community | 27.5% | +20.7% | 30.6% | 147s | Repo ↗ |
| #31 | Sundial-Base | Foundation Model | Tsinghua University | 27.4% | +33.4% | 33.6% | 8.01s | Repo ↗ |
| #32 | PatchTST | Deep Learning | — | 26.8% | +30.1% | 25.1% | 664s | Repo ↗ |
| #33 | DeepAR | Deep Learning | — | 26.1% | +29.1% | 23.6% | 1113s | Repo ↗ |
| #34 | AutoARIMA | Statistical Baseline | — | 23.3% | +21.3% | 21.5% | 20.14s | Repo ↗ |
| #35 | AutoETS | Statistical Baseline | — | 20.3% | -26.8% | 22.5% | 3.48s | Repo ↗ |
| #36 | CatBoost | Tree Ensemble | — | 19.1% | +23.0% | 35.1% | 29.48s | Repo ↗ |
| #37 | LightGBM | Tree Ensemble | — | 17.8% | +21.0% | 33.1% | 3.03s | Repo ↗ |
| #38 | AutoTheta | Statistical Baseline | — | 15.6% | +5.5% | 19.8% | 3.28s | Repo ↗ |
| #39 | Seasonal Naive | Baseline | Community | 10.3% | 0.0% | 11.8% | 0.48s | Repo ↗ |
| #40 | Naive | Statistical Baseline | — | 6.7% | -45.4% | 10.9% | 0.47s | Repo ↗ |
| #41 | Drift | Statistical Baseline | — | 6.1% | -45.8% | 10.0% | 0.45s | Repo ↗ |
Variable Schema & Summary Distributions
Observed numerical ranges, category cardinalities, missing rates, and sample values across 17,520 rows.
| Col | Variable Name | Type | Missing | Distinct | Summary Stats / Distribution | Sample Values |
|---|---|---|---|---|---|---|
| C | targetTARGET | NUM | 0 | 2,502 | [2107, 5224] μ=3328.2 σ=561.3 | 286627102605 |
| A | id | CAT | 0 | 1 | — | loadloadload |
| B | timestamp | CAT | 0 | 17,520 | — | 2013-01-01T01:00:002013-01-01T02:00:002013-01-01T03:00:00 |
| D | airtemperature | NUM | 0 | 311 | [-13.670000076293945, 95] μ=46.8 σ=19.8 | 29.32999992370605529.32999992370605530 |
Frequently Asked Questions: Global Energy Forecasting Competition (GEFCom 2014)
Common questions regarding the Global Energy Forecasting Competition (GEFCom 2014) dataset, machine learning task formulations, and in-browser tabular inference.
What is the forecasting horizon for GEFCom 2014 (proenfo_gfc14)?
The horizon is 168 hours (7 full calendar days), reflecting standard transmission grid scheduling and weekly generation dispatch cycles.
Can I run zero-server forecasts on this dataset in Google Sheets?
Yes, Carla enables zero-server in-context time series forecasting directly in spreadsheets using Mistral's Timestral foundation model without moving data outside your sheet.
Which models perform best on GEFCom 2014?
Modern zero-shot time series foundation models like Mistral's Timestral and tuned gradient-boosted trees (LightGBM) with temperature interaction terms outperform classical statistical baselines like AutoARIMA and Exponential Smoothing.
Test TabICLv2 on Global Energy Forecasting Competition (GEFCom 2014) Yourself
Open the pre-loaded Google Sheet and let Carla configure the target and task for local, zero-cloud tabular machine learning.
Launch Carla
Click once to open the spreadsheet and Carla side panel together.
Review the Setup
Carla selects the dataset target and task from this page automatically.
Evaluate & Predict
Run predictions and compute metrics with zero server uploads.
Provenance & Attribution
Hong, T., Pinson, P., Fan, S., Zareipour, H., Troccoli, A., & Hyndman, R. J. (2016). Probabilistic energy forecasting: Global Energy Forecasting Competition 2014 and beyond. International Journal of Forecasting, 32(3), 896-913.