Data-driven power demand disaggregation to the substation level

Detailed representations of power systems are critical for understanding the operational characteristics and flexibility offered by smart grid technologies. Many details of power grids, however, remain proprietary assets, limiting the transferability of research results based on reference test networks to the real world. This work develops and validates a new methodology to create representative demand profiles at the transmission substation level. The proposed methodology integrates system demand, power grid, and context layers—each utilizing publicly available data. In the demand layer, K-means clustering captures the core statistical time series characteristics of demand data, increasing computational tractability of the problem. The power grid layer establishes a substation contribution index to system level demand, considering the grid’s network structure. In the context layer, demographic data serve as a proxy of demand. Taken together, the layer information creates a substation demand index, which disaggregates the system level demand profile to each transmission level substation. The method is validated on data from the Alberta transmission system operator (AESO), with results showing a mean percent error around zero, and a maximum percent error at 10%. The hourly, disaggregated substation demand profiles are useful within larger modeling efforts, such as transmission expansion planning and hosting studies.