{"success":true,"data":{"author":"TU Delft with OpenSynth (LF Energy)","bucketAccessType":"public","bucketName":"OpenSynth/tud-glass","bucketPrefix":"","category":"Energy","cloudProvider":"huggingface","createdAt":"2026-11-21T22:15:33.249Z","description":"GLASS (Generative-AI Large-customer Smart-meter Samples) is a synthetic electricity dataset of daily load profiles for large-customer smart meters, produced by a flow-matching generative model trained on smart meter data from the Dutch distribution grid operator Liander, and published through OpenSynth, an LF Energy project for synthetic energy data. Each of the 43,184,437 rows is one day of one synthetic sample: 96 average-power values in kW at 15-minute resolution, plus scenario, generation type, customer category, date (2022 to 2024), consumption or generation level and sample id. Four scenarios cover pure consumption (power drawn from the grid), solar PV generation, onshore wind generation and hybrid net consumption with solar, across 26 customer categories based on Dutch Chamber of Commerce sector codes mapped to ISIC sections; each condition has at least 100 samples. The data are a single Parquet file (GLASS_v1.1.parquet, with the earlier v1.0 kept), about 27 GB in the repository (measured 2026-11-21); read with polars or pandas. Use it for load forecasting, grid planning and privacy-preserving energy analytics. Released under CDLA-Permissive-2.0; dataset DOI 10.4121/b18de4df-0f67-4a6f-aa84-6634cdd63991 (4TU.ResearchData).","displayTitle":"GLASS: Generative-AI Large-Customer Smart-Meter Samples, 43 Million Synthetic 15-Minute Daily Load and Generation Profiles (TU Delft, OpenSynth)","doi":"10.4121/b18de4df-0f67-4a6f-aa84-6634cdd63991","huggingFaceUrl":"https://huggingface.co/datasets/OpenSynth/tud-glass","id":"603e24c7-bc6a-5665-8ac1-3e68ac79e1d5","license":"CDLA-Permissive-2.0","licenseEvidenceUrl":"https://huggingface.co/datasets/OpenSynth/tud-glass","objectCountEstimate":4,"probeStatus":"ok","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"OpenSynth/tud-glass\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"OpenSynth/tud-glass\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=OpenSynth, license:cdla-permissive-2.0, size_categories:10M<n<100M, format:parquet, downloads=55"],"format":"mixed","generatedAt":"2026-11-21T22:31:34.371Z","layout":{"example":"GLASS_v1.0.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"GLASS_v1.0.parquet","truncatedAt":512}],"sampleObjects":["GLASS_v1.0.parquet","GLASS_v1.1.parquet"],"sidecars":[{"bytes":"---\nlicense: cdla-permissive-2.0\ntags:\n- energy\n- smartmeter\n- synthetic\npretty_name: GLASS\n---\n# GLASS — Generative-AI Large-customer Smart-meter Samples\n\n## Overview\n\nThis dataset contains synthetic daily load profiles for large-customer smart meters, produced by a flow-matching generative model. Each row represents a single day of a single synthetic sample: 96 power values at 15-minute resolution (in kW), conditioned on customer category, generation type, scenario, and consumption/generation level.\n\nThis dataset is published as part of [OpenSynth](https://lfenergy.org/projects/opensynth/), an [LF Energy](https://lfenergy.org/) project that democratizes synthetic energy data to accelerate the decarbonization of global energy systems. [HuggingFace](https://huggingface.co/OpenSynth) and the [SURF 4TU Data Repository](https://data.4tu.nl/) are its distribution channels. The dataset was generated using the [SmartMeterFM](https://github.com/sentient-codebot/SmartMeterFM) model trained on [Liander N.V.](https://www.liander.nl/) smart meter data of large customers. These are \"grootverbruikers\" according to the Dutch Energy law that have a grid connection with capacity between 60 and 160 kW.\n\nThe references of this dataset can be found at the end of this file.  \n\n- **File:** `GLASS_v1.1.parquet`\n- **Rows:** 43,184,437\n- **Columns:** 104 (8 metadata + 96 power values)\n- **SHA-256:** `815ac9fc98cf2a2823492a894ac81b132c88d6637a9e1642beb280b38d112343`\n\n### Quick start\n\n```python\nimport polars as pl\nimport matplotlib.pyplot as plt\n\ndf = pl.read_parquet(\"GLASS_v1.1.parquet\")\n\n# Pick one sample: consumption_only, category 001, 2023-01-15, level 0, sample 0\nsample = df.filter(\n    (pl.col(\"scenario\") == \"consumption_only\")\n    & (pl.col(\"category\") == \"001\")\n    & (pl.col(\"date\") == \"2023-01-15\")\n    & (pl.col(\"level_id\") == 0)\n    & (pl.col(\"sample_id\") == 0)\n)\n\n# Plot the 96 power values for that day\npower = sample.select(\n    pl.exclude(\n        \"scenario\",\n        \"generation_type\",\n        \"category\",\n        \"date\",\n        \"level_name\",\n        \"level_id\",\n        \"level_value\",\n        \"sample_id\",\n    )\n).row(0)\nplt.plot(power)\nplt.xlabel(\"15-min interval\")\nplt.ylabel(\"Power (kW)\")\nplt.title(\"Daily load profile\")\nplt.show()\n```\n\n\n## Schema\n\n### Metadata columns\n\n| Column | Type | Description |\n|---|---|---|\n| `scenario` | String | One of 4 scenarios (see below) |\n| `generation_type` | String | `none`, `solar`, or `wind_on_land` |\n| `category` | String | Customer category. 26 distinct values. |\n| `date` | String | Calendar date (YYYY-MM-DD), ranging 2022-01-01 to 2024-12-31 |\n| `level_name` | String | Human-readable level: `low`, `mid_low`, `mid_high`, `high` |\n| `level_id` | Int64 | Level index: 0 = low, 1 = mid_low, 2 = mid_high, 3 = high |\n| `level_value` | Float64 | Target total energy for the month (kWh/month) |\n| `sample_id` | Int64 | Index of the sample within its condition group, starting from 0, with at least 100 per condition |\n\n### Power columns (96)\n\nColumns `00:00:00` through `23:45:00` (Float32) — average power in kW for each 15-minute interval of the day.\n\n- **Consumption** scenarios: values are non-negative (power drawn from grid)\n- **Generation** scenarios: values are non-positive (power exported to grid)\n- **Hybrid** scenario: values can be positive or negative\n\n\n## Samples and conditions\n\n### Condition structure\n\nEach sample is generated from a specific **condition**: a combination of scenario, category, year, month, and level. The dataset covers a controlled product of these dimensions.\n\n### Scenarios\n\n| Scenario | Generation type | Sign convention | Description |\n|---|---|---|---|\n| `consumption_only` | `none` | non-negative | Pure consumption, no local generation |\n| `generation_zon` | `solar` | non-positive | Solar PV generation only |\n| `generation_wopl` | `wind_on_land` | non-positive | Onshore wind generation only |\n| `hybrid_net_consumption` | `solar` | mixed | Net consumption (consumption minus generation) |\n\n### Customer Categories (26 categories)\n\nConsumption Categories: E3D, E3A, E3B, E3C, 00","encoding":"utf8","key":"README.md","name":"README.md","size":10147,"truncated":true}],"sourceObjectKey":"GLASS_v1.0.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"OpenSynth/tud-glass\")\nprint(ds)","starterSamplePath":"GLASS_v1.0.parquet","totalObjects":4},"summary":"GLASS (Generative-AI Large-customer Smart-meter Samples) is a synthetic electricity dataset of daily load profiles for large-customer smart meters, produced by a flow-matching…","title":"GLASS: Generative-AI Large-Customer Smart-Meter Samples, 43 Million Synthetic 15-Minute Daily Load and Generation Profiles (TU Delft, OpenSynth)","updatedAt":"2026-11-22T01:02:21.671Z","visibility":"public","websiteUrl":"https://lfenergy.org/projects/opensynth/"}}