| dc.description.abstract | Klimaendringer, delvis utløst av klimagassutslipp, utgjør en kritisk global utfordring. Metan, en svært potent drivhusgass med et globalt oppvarmings potensial på 80 ganger karbondioksid, er en betydelig bidragsyter til denne krisen. Kilder til metanutslipp inkluderer olje- og gassindustrien, landbruket og avfallshåndteringen, med fakling i olje- og gassindustrien som en betydelig utslippskilde.
Fakling, en standard prosess i olje- og gassindustrien, antas ofte å være 98 % effektiv ved omdannelse av metan til mindre skadelig karbondioksid. Nyere forskning fra University of Michigan, Stanford, Environmental Defense Fund og Scientific Aviation indikerer imidlertid at den allment aksepterte effektiviteten på 98 % av fakling ved konvertering av metan til karbondioksid, en mindre skadelig klimagass, kan være unøyaktig. Denne undersøkelsen revurderer fakkelprosessens effektivitet og dens rolle i metankonvertering.
Dette arbeidet fokuserer på å lage en metode for uavhengig å beregne metanutslipp fra olje- og gassvirksomhet for å løse dette problemet. Satellittdata, som er et nyttig verktøy for å beregne klimagassutslipp fra ulike kilder, er inkludert i den foreslåtte metodikken. I tillegg til standard overvåkingsteknikker, tilbyr satellittdata en uavhengig, ikke-påtrengende, rimelig og kontinuerlig overvåkingstilnærming.
På bakgrunn av dette er problemstillingen for dette arbeidet følgende
"Hvordan kan en datadrevet tilnærming utvikles for å forbedre nøyaktigheten og kvaliteten på estimering av metanutslipp fra faklingsaktiviteter i olje- og gassindustrien, ved å bruke satellittdata fra utvalgte plattformer for å oppdage og kvantifisere fremtidige utslipp basert på maskinlæring mer effektivt?"
For å oppnå dette ble følgende mål og aktiviteter utført.
* Teoretisk rammeverk og sentrale begreper
* Teknisk gjennomgang av dagens toppmoderne satellittplattformer og eksisterende litteratur.
* Utvikling av et Proof of Concept
* Foreslå en evaluering av metoden
* Anbefalinger og videre arbeid
Dette arbeidet har tatt i bruk en systematisk tilnærming, som starter med et omfattende teoretisk rammeverk for å forstå bruken av fakling, de miljømessige implikasjonene av metan, den nåværende «state-of-the-art» av forskning, og «state-of-the-art» i felt for fjernmåling via satellitter.
Basert på rammeverket utviklet i de innledende fasene av dette arbeidet, ble det formulert en datadrevet metodikk, som benytter VIIRS-datasettet for å få geografiske områder av interesse. Hyperspektrale data og metandata ble samlet fra Sentinel-2 og Sentinel-5P satellittdatasettet. Denne informasjonen ble behandlet via en foreslått rørledning, med innledende justering og forbedring. I dette arbeidet ble bildene forbedret ved å beregne den normaliserte brennindeksen.
Resultatet var et datasett som inneholdt plasseringen av kjente fakkelsteder, med data fra både Sentinel-2 og Sentinel-5P-satellitten.
Resultatene understreker forskjellene i dekningen mellom Sentinel-2- og Sentinel-5P-data, en faktor som potensielt kan påvirke nøyaktigheten av metanutslippsestimater. De anvendte forbehandlingsteknikkene forbedret dataklarheten og brukervennligheten markant, men deres effektivitet kan avhenge av fakkelstedenes spesifikke egenskaper og rådatakvaliteten. Dessuten, til tross for visse begrensninger, ga kombinasjonen av Sentinel-2 og Sentinel-5P-data effektivt et omfattende datasett egnet for videre analyse.
Avslutningsvis introduserer dette prosjektet en oppmuntrende metodikk for å estimere metanutslipp fra fakling i olje- og gassindustrien. Den legger et grunnleggende springbrett for fremtidig forskning, og forbedrer kontinuerlig presisjonen og kvaliteten på data for å bekjempe klimaendringer. Denne metodikken kan sees i flytskjemaet nedenfor.
Basert på arbeidet som er gjort i dette prosjektet, kan fremtidig arbeid fokusere på å innlemme alternative kilder til metan data, utvide interesseområdene gjennom industrisamarbeid og forsøke å trekke ut ytterligere detaljer gjennom bildesegmenteringsmetoder. Dette prosjektet legger et grunnlag, og baner vei for påfølgende utforskninger å bygge videre på. | |
| dc.description.abstract | Climate change, precipitated in part by greenhouse gas emissions, presents a critical global challenge. Methane, a highly potent greenhouse gas with a global warming potential of 80 times that of carbon dioxide, is a significant contributor to this crisis. Sources of methane emissions include the oil and gas industry, agriculture, and waste management, with flaring in the oil and gas industry constituting a significant emission source.
Flaring, a standard process in the Oil and gas industry is often assumed to be 98% efficient when converting methane to less harmful carbon dioxide. However, recent research from the University of Michigan, Stanford, the Environmental Defense Fund, and Scientific Aviation indicates that the widely accepted 98% efficiency of flaring in converting methane to carbon dioxide, a less harmful greenhouse gas, may be inaccurate. This investigation reevaluates the flaring process's efficiency and its role in methane conversion.
This work focuses on creating a method to independently calculate methane emissions from oil and gas activities to solve this issue. Satellite data, which is a helpful tool for calculating greenhouse gas emissions from various sources, is included in the suggested methodology. In addition to standard monitoring techniques, satellite data offers an independent, non-intrusive, affordable, and continuous monitoring approach.
Based on this, the problem statement for this work is the following
“How can a data-driven approach be developed to enhance the accuracy and quality of methane emission estimation from flaring activities in the Oil and Gas industry, using satellite data from selected platforms to detect and quantify future emissions based on Machine learning more effectively?"
To achieve this, the following objectives and activities were performed.
* Theoretical Framework and key concepts
* Technical review of the current state-of-the-art satellite platforms and existing literature.
* Development of a Proof of Concept
* Proposing an evaluation of the method
* Recommendations and further work
This work has adopted a systematic approach, starting with a comprehensive theoretical framework to understand the utilization of flaring, the environmental implications of methane, the current state-of-the-art of research, and the state-of-the-art in the field of remote sensing via satellites.
Based upon the framework developed during the initial phases of this work, a data-driven methodology was formulated, utilizing the VIIRS dataset to get geographical areas of interest. Hyperspectral and methane data were aggregated from the Sentinel-2 and Sentinel-5P satellite dataset. This information was processed via a proposed pipeline, with initial alignment and enhancement. In this work, the images were enhanced by calculating the Normalized Burn Index.
The result was a dataset containing the location of known flare sites, with data from both the Sentinel-2, and the Sentinel-5P satellite.
The results underscore the disparities in coverage between Sentinel-2 and Sentinel-5P data, a factor that could potentially influence the precision of methane emission estimates. The applied preprocessing techniques markedly enhanced data clarity and usability, but their efficacy may hinge on the flaring sites' specific characteristics and the raw data quality. Moreover, despite certain limitations, the combination of Sentinel-2 and Sentinel-5P data effectively yielded a comprehensive dataset suitable for further analysis.
In conclusion, this project introduces an encouraging methodology for estimating methane emissions from flaring activities within the oil and gas industry. It lays a foundational steppingstone for future research, continually enhancing the precision and quality of data in combating climate change. This methodology can be seen in the flow chart below.
Based on the work done in this project, future work could focus on incorporating alternative sources of methane data, broadening the areas of interest through industry collaboration, and attempting to extract further features through image segmentation methods. This project signifies a start, paving the way for subsequent explorations to build upon.
Climate change, precipitated in part by greenhouse gas emissions, presents a critical global challenge. Methane, a highly potent greenhouse gas with a global warming potential of 80 times that of carbon dioxide, is a significant contributor to this crisis. Sources of methane emissions include the oil and gas industry, agriculture, and waste management, with flaring in the oil and gas industry constituting a significant emission source.
Flaring, a standard process in the Oil and gas industry is often assumed to be 98% efficient when converting methane to less harmful carbon dioxide. However, recent research from the University of Michigan, Stanford, the Environmental Defense Fund, and Scientific Aviation indicates that the widely accepted 98% efficiency of flaring in converting methane to carbon dioxide, a less harmful greenhouse gas, may be inaccurate. This investigation reevaluates the flaring process's efficiency and its role in methane conversion.
This work focuses on creating a method to independently calculate methane emissions from oil and gas activities to solve this issue. Satellite data, which is a helpful tool for calculating greenhouse gas emissions from various sources, is included in the suggested methodology. In addition to standard monitoring techniques, satellite data offers an independent, non-intrusive, affordable, and continuous monitoring approach.
Based on this, the problem statement for this work is the following
“How can a data-driven approach be developed to enhance the accuracy and quality of methane emission estimation from flaring activities in the Oil and Gas industry, using satellite data from selected platforms to detect and quantify future emissions based on Machine learning more effectively?"
To achieve this, the following objectives and activities were performed.
* Theoretical Framework and key concepts
* Technical review of the current state-of-the-art satellite platforms and existing literature.
* Development of a Proof of Concept
* Proposing an evaluation of the method
* Recommendations and further work
This work has adopted a systematic approach, starting with a comprehensive theoretical framework to understand the utilization of flaring, the environmental implications of methane, the current state-of-the-art of research, and the state-of-the-art in the field of remote sensing via satellites.
Based upon the framework developed during the initial phases of this work, a data-driven methodology was formulated, utilizing the VIIRS dataset to get geographical areas of interest. Hyperspectral and methane data were aggregated from the Sentinel-2 and Sentinel-5P satellite dataset. This information was processed via a proposed pipeline, with initial alignment and enhancement. In this work, the images were enhanced by calculating the Normalized Burn Index.
The result was a dataset containing the location of known flare sites, with data from both the Sentinel-2, and the Sentinel-5P satellite.
The results underscore the disparities in coverage between Sentinel-2 and Sentinel-5P data, a factor that could potentially influence the precision of methane emission estimates. The applied preprocessing techniques markedly enhanced data clarity and usability, but their efficacy may hinge on the flaring sites' specific characteristics and the raw data quality. Moreover, despite certain limitations, the combination of Sentinel-2 and Sentinel-5P data effectively yielded a comprehensive dataset suitable for further analysis.
In conclusion, this project introduces an encouraging methodology for estimating methane emissions from flaring activities within the oil and gas industry. It lays a foundational steppingstone for future research, continually enhancing the precision and quality of data in combating climate change. This methodology can be seen in the flow chart below.
Based on the work done in this project, future work could focus on incorporating alternative sources of methane data, broadening the areas of interest through industry collaboration, and attempting to extract further features through image segmentation methods. This project signifies a start, paving the way for subsequent explorations to build upon. | |
