Probabilistic processing of survey collected data in a residential load area for hourly demand profile estimation

A simulation model developed for prediction of the electric load profile of a residential area is briefly outlined in the paper. The model based on a "bottom-up" approach in order to reproduce customers' daily behaviour concerning elcctricity usage, has been applied for validation to reproduce load diagrams of a sample residential area where field measurements of power demand arid a combined mail survey had been previously conducted. Since the residential electric load was almost exclusively of the discretionary, weather-independent type, the survey ques tionnairc regarded mainly thc lifestyle of customers, thus evidencing the need for probabilistic processing of the data collected by the survey. To this end customers' presence at home and appliance usage have been considered as normal distributions around a incan value as ensued from the survey data. In the paper, a sensitivity analysis is performed on standard deviation values of the above distributions, which allows quite good matching of recorded data with simulation risults. A heuristic approach is proposed for correcting the dispersions of normal distributions to compensate for the averaging inherent in the questionnaire replies compared with actual random behaviour of the persons concerned