Power consumption forecasts are among the vital data used by Distribution Utilities (DUs) in daily power supply procurement. The most inconsistent load profile among the consumer types are from residential consumers due to fluctuations in their power consumptions. This is due to a wide variation of daily routine and non-routine activities performed by every household member at varied times of the day. The routine activities contribute largely to the consumer's daily load profile and power consumption. Consequently, determining consumers routine activities enables the DUs to have a more precise day-ahead forecast. This paper proposes a method that identifies consumers' routine activities and reconstructs their corresponding daily load profiles using the historical power consumption data. This is accomplished by finding the power consumption curves appearing regularly in many of consumers' data to identify candidates of routine activities and further evaluating probabilities that a specific consumer performs those activities to determine whether each of them is actually a routine activity. Probabilistically, the daily load profile is reconstructed using the determined routine activity. Attributing to the probabilistic nature of the analysis, a 95% confidence limit is calculated for every 15-minute power consumption so that DUs have a reliable data on residential consumers' daily power demand.