Non-intrusive Load Monitoring (NILM) is a computational technique that extracts individual appliance consumption and operation state change information from the aggregate power consumption made by a single residential or commercial unit. This technique has emerged as a reliable energy management approach that intends to reduce energy wastage and inform customers about their electricity consumption. NILM is considered as both Supervised and Semi-supervised Learning problems. The main contribution of this thesis is three-fold.

	First, we evaluated some regression algorithms commonly used in NILM research based on 8 different training and testing scenarios which according to our knowledge covered major demographic factors that affect the appliance usage. The dataset used for the evaluation of the regression models, was collected from a research lab at Grenoble INP, in Grenoble, France. Furthermore, a novel Bayesian optimized Ensemble regressor model for predicting individual appliance consumption from aggregated load data is also proposed. Instead of just using the aggregated power information, the proposed model also uses demographic information from the dataset to estimate accurate consumption output of individual appliances.

	NILM research often requires significant labeled data and obtaining such data by installing smart meters at the end of consumers’ appliances is laborious and expensive and exposes users to severe privacy risks. Moreover, most NILM research uses empirical observations instead of proper mathematical approaches to obtain the threshold value for determining appliance operation states (On/Off) from their respective energy consumption value. The second fold of the thesis proposes a novel semi-supervised multilabel deep learning technique based on Temporal Convolutional Networks (TCN) and Long short-term memory (LSTM) for classifying appliance operation states from labeled and unlabeled data. The two thresholding techniques, namely Middle Point Thresholding and Variance Sensitive Thresholding, which are needed to derive the threshold values for appliance operation states, were also compared thoroughly. The proposed models were then evaluated using Redd, Uk-Dale and Refit datasets.

	Third, we propose a novel NILM algorithm that utilizes deep learning Temporal Convolutional Networks (TCN) for the regression and classification NILM tasks. Most NILM models cannot simultaneously classify appliance operational status or estimate individual appliance power consumption. The deep TCN layers in the proposed architecture of the third fold of the thesis allow the simultaneous extraction of complex patterns in the data of the power consumption and the operational state of individual appliances. Refit data is used for the evaluation of this model.