Smart home energy management based on deep reinforcement learning technology

• Main Author: Sun, Xinghao
• Other Authors: Xu Yan
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering; Smart home energy management system; Deep reinforcement learning; Deep Q network
• Online Access: https://hdl.handle.net/10356/175417
• Institution: Nanyang Technological University

This study framework designed to enhance home energy management systems (HEMS) in smart homes that incorporate renewable energy and electric vehicles (EVs). Leveraging real-world data within an advanced simulation platform, we delve into the interactions among diverse appliances, home batteries, photovoltaic (PV) systems, and the smart grid. Developed using Python, this framework allows for detailed management of energy resources, including the strategic charging of home batteries and EVs, scheduling appliances, and optimizing air conditioning to boost energy efficiency andreduce costs. This study utilizes Deep-Q-Networks (DQN)-based Reinforcement.Learning (RL) to manage smart homes energy usage more wisely. The goal is tocut down on energy costs while keeping homes comfortable and making sure all the devices work as they should. This system can make decisions on its own, schedules when to turn appliances on or off, based on how much energy is available from solar panels and the cost of electricity from the grid. The simulations in this study are based on real-world PV and price datasets, demonstrating the how the proposed method used and stored energy. The renewable energy changes over time and actions of HEMS agent are compared. The results from these simulations show big benefits. The proposed HEMS method based DQN help reduce the need for electricity from the grid, make more use of clean energy from the solar PV, save money on energy bills, and make homes more self-sufficient. The results show how energy is used in homes and how different energy sources and appliances interact, pushing forward the development of smart homes and green energy. The system designed in this study can be adjusted and improved for future research, including testing more complex situations and using DQN methods to get better at managing home energy.