Resource constraint Federated Learning for Energy Smart Meters

Fuente: Zenodo
Saved in:
Bibliographic Details
Main Author: Duttagupta, Abhishek
Format: Recurso digital
Published: Zenodo 2023
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866901709421281280
author Duttagupta, Abhishek
author_facet Duttagupta, Abhishek
contents <p>The thesis focuses on simulating a Federated Learning framework using the London Household smart energy meter dataset. The primary objective is to achieve comparable accuracy in load forecasting to state-of-the-art (SOTA) technologies while demonstrating the preservation of data privacy through FL. In real-world scenarios,  the challenge of noni.i.d. connections persists. To replicate real-world conditions, limited random connections are considered for each round of FL. Additionally, since smart meters often have limited computational power, small sequential dense neural networks (DNN) are utilized to address this constraint. These DNN models are well-suited for low computational power devices and contribute to solving the computational challenges posed by smart meters. Through this project, the aim is to showcase the potential of FL in maintaining data privacy while achieving accurate load forecasting results, even in resource-constrained environments.</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_15375891
institution Zenodo
language
publishDate 2023
publisher Zenodo
record_format zenodo
spellingShingle Resource constraint Federated Learning for Energy Smart Meters
Duttagupta, Abhishek
<p>The thesis focuses on simulating a Federated Learning framework using the London Household smart energy meter dataset. The primary objective is to achieve comparable accuracy in load forecasting to state-of-the-art (SOTA) technologies while demonstrating the preservation of data privacy through FL. In real-world scenarios,  the challenge of noni.i.d. connections persists. To replicate real-world conditions, limited random connections are considered for each round of FL. Additionally, since smart meters often have limited computational power, small sequential dense neural networks (DNN) are utilized to address this constraint. These DNN models are well-suited for low computational power devices and contribute to solving the computational challenges posed by smart meters. Through this project, the aim is to showcase the potential of FL in maintaining data privacy while achieving accurate load forecasting results, even in resource-constrained environments.</p>
title Resource constraint Federated Learning for Energy Smart Meters
url https://doi.org/10.5281/zenodo.15375891