Smart Home Energy Management System

A tétel rövid nézete
dc.contributor.advisorSarvajcz, Kornél
dc.contributor.authorMehanna, Raghid
dc.contributor.departmentDE--Műszaki Kar
dc.date.accessioned2025-09-04T15:26:12Z
dc.date.available2025-09-04T15:26:12Z
dc.date.created2025-05-12
dc.description.abstractWith raising energy cost, environmental concerns, and the widespread use of the smart home technologies, efficient energy management in residential buildings is becoming increasingly important. In this thesis, I present an end to end design for a Smart Home Energy Management System (SHEMS) that utilises the ZWave sensors and the smart devices with the Home Assistant on a Raspberry Pi. Reinforcement Learning (RL) is used in the system to autonomously optimize energy consumption and comfort level by using the algorithm of Proximal Policy Optimization (PPO). The agent was trained on a simulated environment, with realistic home scenarios that bypass practical limitations due to hardware. The model was tested thoroughly including a large number of smoke tests and more than 80,000 timesteps of training. Visualizations in TensorBoard provided good insight into how the agent was learning and how effective the actions it made were. This paper demonstrates that the RL based SHEMS not only achieves excellent tradeoff between energy efficiency and user comfort but also achieve super energy conservation compared to traditional management. The findings complement the possibilities of utilizing advanced machine learning methods on top of residential home automation frameworks to make valuable contributions and operational concepts for an environmentally friendly and intelligent energy operation in residences.
dc.description.courseMechatronikai mérnöki
dc.description.degreeMSc/MA
dc.format.extentThis thesis builds a fully virtual smart-home testbed using Home Assistant on a VirtualBox hosted Raspberry Pi image with simulated Z-Wave sensors and smart plugs , frames energy optimization as a reinforcement learning task in a custom OpenAI Gym environment using occupancy, temperature, humidity, dynamic tariffs and energy use as state inputs and simple on/off actions for plugs and lights implements and trains a Proximal Policy Optimization agent with a combined energy cost/comfort reward over 100 000+ timesteps (iteratively tuned via TensorBoard), and benchmarks it against static rule based and manual schedules on simulated energy cost, comfort compliance and control efficiency. All within a single home, simulation only setup that deliberately excludes physical hardware deployment, thermostat control by RL, multi home scenarios and real world user data.
dc.identifier.urihttps://hdl.handle.net/2437/397246
dc.language.isoen
dc.rights.infoHozzáférhető a 2022 decemberi felsőoktatási törvénymódosítás értelmében.
dc.subjectSmart Home Energy Management System (SHEMS), Reinforcement Learning (RL), Proximal Policy Optimization (PPO), Internet of Things (IoT)
dc.subject.dspaceMűszaki tudományok
dc.titleSmart Home Energy Management System
Fájlok
Eredeti köteg (ORIGINAL bundle)
Megjelenítve 1 - 1 (Összesen 1)
Nincs kép
Név:
Thesis - Raghid Mehanna_DEA.pdf
Méret:
2.45 MB
Formátum:
Adobe Portable Document Format
Leírás:
Megjelenítés
Engedélyek köteg
Megjelenítve 1 - 1 (Összesen 1)
Nincs kép
Név:
license.txt
Méret:
2.19 KB
Formátum:
Item-specific license agreed upon to submission
Leírás:
Letöltés
Gyűjtemények
Hallgatói dolgozatok (Villamosmérnöki és Mechatronikai Tanszék)