Nowadays,the home dwellers want to check their energy usage due to the rising inflation and higher electricity costs in the world.Therefore,it is a potential market for residential builders and contractors provide inhabitants with tools to accurately monitor their energy consumption.The requirement is to monitor each device with their power consumption information displaying on the smart appliances which is not the case for smart systems and equipment nowadays.Non-Intrusive Load Monitoring（NILM）is a method which provides its users the estimated power consumption at the device level depending on the cumulative readings taken at the electric meter or main circuit display panel.Typical NILM technique monitors the changes in the power signal of the various devices when they turn on or off.Based on these results,it estimates the power consumption of the device.The essential element for this approach to work is that the models are need to be trained the various type of devices present at home in order to accurately estimate their power consumption.Due to this limitation,the literature does not discuss the disaggregation problem much but rather emphasizes on the signal/change detection and determining the power.This thesis discusses a better version of a typical NILM problem and utilizes a un-supervised and data based algorithm for solution.A technique named power consumption clustered nonintrusive load monitoring（PCCNILM）is proposed in which the power consumption of various devices is clustered together by their energy consumption levels.The algorithm that is developed to solve the PCC-NILM problem is called approximate power trace decomposition algorithm（APTDA）.This algorithm does not require learning as compared to others and provides projected power intake for various groups of equipment.