الفهرس 
INTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 97 |  |  |
| | | from | 97 | | |
Abstract
Residential, commercial, and industrial customers are continuously increasing their demands on the nation’s already-stressed utilities. This challenge may be solved by either expanding the number of energy-generating units or optimizing their utilization. The first approach is far more expensive and time demanding than the second. As a result, effective Demand Side Management (DSM) of current electrical resources became crucial. By integrating information technology into the grid structure and enabling real-time monitoring and control of energy production and consumption, traditional power grids have been recently turned into ”Smart Grids” (SG). DSM is a critical component of a smart grid since it enables users to take control of their energy usage patterns. DSM might be thought of as a shift in consumer behavior that influences the load curves of the electrical system. Forecasts of the electricity consumption and individual home loads are required to carry out the Demand Response (DR) initiatives. Recent studies reveal that Deep Learning seems to be an effective tool capable of deciphering the massive volumes of data generated by an IoT-based grid system. It comprises the data collection, processing, and decision-making processes associated with the smart grid system. Through distributed energy resources, improved metering infrastructure, dynamic load behavior, and DSM, the smart grid empowers customers. Data gathering and analysis are often used to aid in the design and operation of power systems. Data analysis in real time is carried out for making the best possible decision. The need for developing effective load forecasting tools to cope with such new trends has become crucial. Furthermore, the deployment of thesetools within the unprecedentedly developed infra-structure of modern power systems turned to be urgent. This thesis aims to develop a load forecas working under IoT-Enabled Residential Households. The proposed load forecasting model is constructed using a deep learning technique based on Long Short-Term Memory (LSTM) units. The suggested model is used to forecast both power station-level and customer-level loads. The suggested model’s performance is tested using a real-world energy consumption dataset from individual smart residential buildings. The output of the proposed load forecasting model is compared against the output of the same dataset generated using a range of approaches reported in the literature. The proposed LSTM model proves to provide the most precise predictions among all other competing techniques. This thesis also aims to establish a paradigm for the scheduling of a two-step energy consumption. The first step is achieved by the utility, where it enables the utility to determine the maximum permissible energy consumption per hour for an IoT-enabled residential unit. The second step is to be completed at the customer side, where it is used to distribute the load throughout the day and to provide a schedule for the operation of household appliances. This strategy may be used to indirectly decrease the utility’s peak-toaverage and, by extension, to improve the load curve. The DSM program is implemented using CVX program in MATLAB® (R2018b). Finally, this thesis introduces an integrated cloudbased framework that makes use of the previously described deep learning-based load forecasting model and the DSM model to perform real-time monitoring. The model is implemented using ThingSpeak™ platform to demonstrate the efficacy of smart grid energy management and real-time monitoring.