Toward global rooftop PV detection with Deep Active Learning

It is crucial to know the location of rooftop PV systems to monitor the regional progress toward sustainable societies and to ensure the integration of decentralized energy resources into the electricity grid. However, locations of PV are often unknown, which is why a large number of studies have pr...

Full description

Saved in:
Bibliographic Details
Main Authors: Matthias Zech, Hendrik-Pieter Tetens, Joseph Ranalli
Format: Article
Language:English
Published: Elsevier 2024-12-01
Series:Advances in Applied Energy
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2666792424000295
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846127142697435136
author Matthias Zech
Hendrik-Pieter Tetens
Joseph Ranalli
author_facet Matthias Zech
Hendrik-Pieter Tetens
Joseph Ranalli
author_sort Matthias Zech
collection DOAJ
description It is crucial to know the location of rooftop PV systems to monitor the regional progress toward sustainable societies and to ensure the integration of decentralized energy resources into the electricity grid. However, locations of PV are often unknown, which is why a large number of studies have proposed variants of Deep Learning to detect PV panels in remote sensing data using supervised Deep Learning. However, these methods are based on annotating datasets and therefore often require relabeling when fine-tuned or extended to a different region. Recent advances in Deep Active Learning offer the opportunity to significantly reduce the number of required annotated images by intelligently selecting the images to label next based on their informative value for the model. In this study, we compare different Deep Active Learning algorithms using a variety of datasets from different regions and compare different model training variants. In the simulations, the entropy-based acquisition function shows the highest performance with only 3% of the data needed in case-imbalanced data, while remaining simple to implement. We believe that Deep Active Learning provides an elegant solution to maintain high model accuracy while reducing annotation effort substantially. This facilitates the development of generalizable models for worldwide rooftop PV detection.
format Article
id doaj-art-1eaec8d822934bf8b37e96ee2467516a
institution Kabale University
issn 2666-7924
language English
publishDate 2024-12-01
publisher Elsevier
record_format Article
series Advances in Applied Energy
spelling doaj-art-1eaec8d822934bf8b37e96ee2467516a2024-12-12T05:23:39ZengElsevierAdvances in Applied Energy2666-79242024-12-0116100191Toward global rooftop PV detection with Deep Active LearningMatthias Zech0Hendrik-Pieter Tetens1Joseph Ranalli2German Aerospace Center (DLR), Institute of Networked Energy Systems, Carl-von-Ossietzky-Straße 15, Oldenburg, 26129, Germany; Corresponding author.German Aerospace Center (DLR), Institute of Networked Energy Systems, Carl-von-Ossietzky-Straße 15, Oldenburg, 26129, GermanyPenn State Hazleton, 76 University Drive, Hazleton, 18202, PA, USAIt is crucial to know the location of rooftop PV systems to monitor the regional progress toward sustainable societies and to ensure the integration of decentralized energy resources into the electricity grid. However, locations of PV are often unknown, which is why a large number of studies have proposed variants of Deep Learning to detect PV panels in remote sensing data using supervised Deep Learning. However, these methods are based on annotating datasets and therefore often require relabeling when fine-tuned or extended to a different region. Recent advances in Deep Active Learning offer the opportunity to significantly reduce the number of required annotated images by intelligently selecting the images to label next based on their informative value for the model. In this study, we compare different Deep Active Learning algorithms using a variety of datasets from different regions and compare different model training variants. In the simulations, the entropy-based acquisition function shows the highest performance with only 3% of the data needed in case-imbalanced data, while remaining simple to implement. We believe that Deep Active Learning provides an elegant solution to maintain high model accuracy while reducing annotation effort substantially. This facilitates the development of generalizable models for worldwide rooftop PV detection.http://www.sciencedirect.com/science/article/pii/S2666792424000295Deep Active LearningPV panel detectionMachine LearningSemantic segmentationRemote sensing
spellingShingle Matthias Zech
Hendrik-Pieter Tetens
Joseph Ranalli
Toward global rooftop PV detection with Deep Active Learning
Advances in Applied Energy
Deep Active Learning
PV panel detection
Machine Learning
Semantic segmentation
Remote sensing
title Toward global rooftop PV detection with Deep Active Learning
title_full Toward global rooftop PV detection with Deep Active Learning
title_fullStr Toward global rooftop PV detection with Deep Active Learning
title_full_unstemmed Toward global rooftop PV detection with Deep Active Learning
title_short Toward global rooftop PV detection with Deep Active Learning
title_sort toward global rooftop pv detection with deep active learning
topic Deep Active Learning
PV panel detection
Machine Learning
Semantic segmentation
Remote sensing
url http://www.sciencedirect.com/science/article/pii/S2666792424000295
work_keys_str_mv AT matthiaszech towardglobalrooftoppvdetectionwithdeepactivelearning
AT hendrikpietertetens towardglobalrooftoppvdetectionwithdeepactivelearning
AT josephranalli towardglobalrooftoppvdetectionwithdeepactivelearning