Load Profile Assignment for Planning and Operation Support in Distribution Networks Under Partial Smart Meter Penetration

The growing need to enhance observability in distribution networks has driven the development of load pseudomeasurement generation methods, particularly under partial smart meter (SM) penetration. This paper proposes a load pseudomeasurement framework that builds representative daily load profiles (load curves) from hourly SM time series using clustering techniques, with and without weather information. Markov chain models are then used to capture day-to-day dynamics by predicting the most likely next-day profile to be assigned to customers without SM. To enable this transfer, a hierarchical grouping scheme based on monthly energy consumption is introduced to map behaviors from SM-equipped customers to customers without SM measurement. The methodology is validated with real residential data from the Low-Carbon London project under multiple observability scenarios including different SM availability levels, where SM measurements are withheld from the inputs to emulate customers without SM measurement, and the resulting pseudomeasurements are benchmarked against the original measurements. The results show that the Euclidean representative curve method achieved the most robust overall performance, with a minimum MAE of 1.65 in the Reduced × 75% SM configuration. The best-performing configuration depended on the observability level: Reduced was the most robust option under medium-to-high observability, whereas Temp_reduced with a 21-day window performed best under the lowest-observability condition. In addition, the Euclidean method showed low practical deviation in the Reduced × 25% SM case, with a bias of 0.63 and Cohen’s d = 0.27. Overall, the proposed approach accurately reproduces the hourly load shape and captures inter-day variability under partial observability conditions.