A Real World Multi Year Hourly District Heating Demand Data for Denmark

This dataset contains retrospective hourly heat consumption data from 4,910 smart meters in a Danish district heating system, aggregated across three District Metered Areas (DMA A, DMA B, and DMA C). The data spans from January 1, 2016 to December 31, 2019, with continuous hourly resolution and no missing timestamps. Since raw smart meter readings are cumulative, a first-order difference has been applied to derive hourly consumption values (in kWh). These values are then aggregated at the DMA level to represent total heat demand at each time step.

For each DMA, the dataset provides three aligned time series: (i) aggregated heat consumption, (ii) normalized aggregated consumption (i.e., consumption per active meter), and (iii) the number of contributing meters at each timestamp. The normalized series is computed by dividing aggregated consumption by the number of active meters, which is always non-zero. This dataset supports research in heat demand forecasting, energy system optimization, and time series analysis, and is associated with the work by Ramachandran et al. (2026) on deep learning-based forecasting using time–frequency representations.