kth.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Comparison Of Convolutional Neural Networks And Long Short-Term Memory For Predicting Hydrogen Sulfide In Sewer Systems
KTH, School of Electrical Engineering and Computer Science (EECS).
KTH, School of Electrical Engineering and Computer Science (EECS).
2024 (English)Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesisAlternative title
Jämförelse av Convolutional Neural Networks och Long Short-Term Memory för predicering av svavelväte i avloppssystem (Swedish)
Abstract [sv]

Bildandet av svavelväte (H2S) i avloppssystem utgör betydande utmaningar för avloppsinfrastruktur och folkhälsa. En djupare förståelse av H2 S-prediktion kan vara fördelaktig för att genomföra förebyggande åtgärder och därmed minska svavelvätets påverkan. Denna studie jämför de prediktiva förmågorna hos Convolutional Neural Network (CNN) och Long Short-Term Memory (LSTM) modeller för att förutsäga H2S-nivåer med hjälp av operativ data från pumpstationer, inklusive temperatur, concentration och vattnets ålder. Med hjälp av statistiska mått som R2 och andra anpassade evalueringsmått fann studien att CNN-modellen presterade bättre än LSTM-modellen. CNN uppnådde en träffsäkerhet på 94,2% jämfört med LSTM 89,8%, inom en definierad felmarginal. Även om LSTM-modellen effektivt fångar toppars utseende, hade den svårigheter med generalisering. Trots CNNs överlägsna prestanda visade båda modellerna starka prediktiva förmågor baserat enbart på operativa data, utan att inkludera kemiska parametrar som syrenivåer, nitrater eller pH. Dessutom undersökte studien strategiska faktorer som är väsentliga för att integrera H2S-prediktioner i företagslösningar. Resultaten lyfter fram överväganden som finansiella, kapacitet, marknadsstrategi och implementeringsrisker.

Abstract [en]

The formation of hydrogen sulfide (H2S) in wastewater and sewage systems poses significant challenges to wastewater infrastructure and public health. A deeper understanding of H2S  prediction could be beneficial for implementing preventive measures, ultimately reducing the impact of hydrogen sulfide. This paper compares the predictive capabilities of Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) models for predicting H2S levels using operational data from pump stations, including temperature, concentration, water age, and flow. Using statistical measures such as R2  and various custom metrics, the study found that the CNN model outperformed the LSTM model. The CNN achieved an accuracy of 94.2%, compared to LSTM’s 89.8%, within a defined margin of error. While LSTM effectively captured peak characteristics, it struggled with generalization. Despite CNN’s superior performance, both models showed strong predictive capabilities based solely on operational data, without incorporating chemical features such as oxygen levels, nitrates, or pH. Additionally, the study examined strategic factors essential for integrating H2S  predictions into enterprise offerings. The findings highlight considerations such as financial, capability, market strategy, and implementation risks.

Place, publisher, year, edition, pages
2024. , p. 14
Series
TRITA-EECS-EX ; 2024:270
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:kth:diva-350769OAI: oai:DiVA.org:kth-350769DiVA, id: diva2:1884899
Supervisors
Examiners
Note

QC 20240819

Available from: 2024-08-13 Created: 2024-07-18 Last updated: 2024-08-19Bibliographically approved

Open Access in DiVA

fulltext(1260 kB)140 downloads
File information
File name FULLTEXT01.pdfFile size 1260 kBChecksum SHA-512
03a4edfb023aef3a9ea1b97229388c08a6b12d66aad6d7ce01d3d772baf6ecac21cfc65acd79c4ed7b699181568fdf1d3eb25182ca35524bd679f8695bf4c501
Type fulltextMimetype application/pdf

By organisation
School of Electrical Engineering and Computer Science (EECS)
Computer and Information Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 140 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 303 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf