K-DEMOS: Long-term Projections of Municipal-level Hourly Electricity Demand by End Use in Korea

K-DEMOS (Korea DEMand at municipal-hOurly Scale) provides hourly electricity demand for all 229 Korean municipalities, disaggregated into four end uses: residential, industrial, commercial-and-other, and transport.

Historical demand for 2004–2025 is reconstructed by fusing four official observations of the demand field — KEPCO municipal billing statistics, KOSIS load-research profiles, KPX wholesale market metering, and KMA station temperatures. The reconstruction reproduces official accounting to within 0.13% and is published at monthly-by-hour resolution for 2004–2025 and at full hourly resolution from 2024. Future demand for 2030–2050 is projected in five-year steps at hourly resolution, anchored to national end-use totals from GCAM v9.1 under a reference and an economy-wide net-zero scenario, with explicit layers for electric-vehicle charging and heating electrification and a climate adjustment under SSP2-4.5 and SSP3-7.0. Under the net-zero scenario, electrified heating raises the winter peak to nearly match the summer peak by 2050.

The record ships as five self-contained archives, so that users can download only the branch they need; every archive includes the README and license:

- K-DEMOS_v1.0_core.zip (66 MB) — observed monthly-by-hour fields 2004–2025 (CSV), the incremental EV charging layer for both scenarios (Parquet), and all national accounting ledgers, climate adders, and five-year GCAM anchor originals
- K-DEMOS_v1.0_hourly_reference_ssp245.zip (1.2 GB) — hourly demand fields: observed 2024–2025 and the Reference × SSP2-4.5 branch
- K-DEMOS_v1.0_hourly_reference_ssp370.zip (1.1 GB) — Reference × SSP3-7.0 branch
- K-DEMOS_v1.0_hourly_netzero_ssp245.zip (1.1 GB) — Net-zero × SSP2-4.5 branch
- K-DEMOS_v1.0_hourly_netzero_ssp370.zip (1.1 GB) — Net-zero × SSP3-7.0 branch

The README documents the schema, English column aliases, day-type and leap-year conventions, and the sales/metered basis accounting. The dataset is described in a manuscript submitted to GEO DATA; the full processing pipeline is openly available at https://github.com/hyuntaechoi634/k-demos.