Machine learning techniques for fuel loss detection at service stations
By Sourced Externally
June 5, 2020
Fuel losses such as leakage from underground fuel storage systems is a common problem globally. It causes significant contamination of soil and groundwater, with the resulting health implications, and often requires costly remediation. Traditional compliance based tools and software used to monitor fuel losses rely heavily on conventional statistical methods and have limitations in practical use. But what if there was a viable and powerful alternative which can potentially revolutionize the business practice in this industrial sector.
A breakthrough project led by EMS & RMIT’s School of Science Researchers, has been awarded a grant to develop effective machine learning techniques for identifying different sources of fuel losses, including leaks. The outcomes of this project will enable an increase in productivity and provide powerful techniques for accurate identification and quantification of fuel losses, whether these are temperature related losses, or losses through incorrectly calibrated equipment.
Australian Research Council (ARC) has announced that RMIT researchers in collaboration with industry partner Environmental Monitoring Solutions Pty Ltd (EMS) (specialists in Wetstock Management, SIR Leak Detection, Real-time Fuel Data analytics and Forecourt Automation for the past 25 years), has been one of the 5 organisations to receive the grant as part of the recent ARC Linkage Projects Grants Scheme funds for collaborative research and development projects.
The Linkage Projects Grants Scheme funds collaborative research and development projects between higher education researchers and partner organisations. Deputy Vice-Chancellor Research and Innovation Professor Calum Drummond said RMIT was renowned for the strength of our national and international industry relationships. “These grants are a recognition of RMIT’s continued success in collaborative research partnerships,” he said. “They’ll see our researchers continue to do what they do best – making a real and positive difference that will benefit our whole community.”
EMS and partner RMIT will use the funds to develop data analytic practices and automation, as well as to enable better and faster decision making through harnessing effective machine learning methods.
A COMPLEX CHALLENGE
Current software packages used in the industry by EMS and others are mostly based on rudimentary statistical methods, which have serious limitations in practical use. Some parts of the process are still carried out manually, which are labour intensive and can have a slow turnaround time. Thus, clear guidelines for data analytic practices and automation, as well as better and faster decision making through harnessing effective machine learning methods are urgently needed.
Machine learning provides a viable and powerful alternative which can potentially revolutionize the business practice in this industrial sector.
This project will help develop innovative and effective Artificial Intelligence (AI) and Machine Learning methodologies to overcome the current limitations faced and will empower fuel loss monitoring software to quickly and accurately identify, quantify and categorize unacceptable losses. This system will benefit the global petroleum industry by reducing costs incurred by wasteful leaks and losses, and assist it in meeting global health, safety and environmental regulations.
DESIGNING THE NEXT GENERATION OF WETSTOCK MANAGEMENT
This project will provide us with a significantly improved fuel leak detection model, that will result in improved data analytic capabilities, reduced misclassification rate, and faster response to issues and turnaround
time for fuel loss investigations. For our clients, these translate into cost saving, reduced environmental impacts, less manual inputs, and improved business practices (e.g., improved ATG programming practise to reduce calibration error in new tanks).
Our techniques will be fast and reliable so that the service operators can be informed in a timely manner, which is presently not possible with the existing technologies deployed in industry.
We will gain invaluable insights into the existing practice, concerning how to best use the data available to carry out predictive analytics and anomaly detection. Gaining such insights will allow us to develop a general framework providing guidelines for better data analytic practices in this particular industry sector.
Current practice is lagging behind in terms of pre-processing of data, detection accuracy, labour intensiveness, and long turnaround time. This much needed general framework represents a methodological innovation in this industry sector.
THE BEST OF RMIT & EMS LEADING THE PROJECT
The Chief Investigators Prof. Xiaodong Li and Dr. Jeffery Chan (RMIT) and Partner Investigator Erica Scott (EMS) & Dr. Li Chik (EMS) will be mainly responsible for carrying out the project.
EMS Managing Director, Russell Dupuy said: “EMS is very excited to work in partnership with RMIT. Our research team has been modelling machine learning for a few years now, this collaboration will further enhance our technology solutions for the benefit of all clients and the market through proven science and research. We look forward to celebrating this outcome with all involved and contributing to improving current Fuel Loss detection techniques significantly.”
The announcement underpins the work and research undertaken at RMIT to support initiatives in AI and machine learning. The successful outcomes of this research will bridge the gap between the research community and practitioners from the energy industry. It is expected to bring greater understanding underpinning UPSS problems and provide tangible benefits to the industry.
Contact us for more info P: +61 3 9785 5000 E: email@example.com W: drivingfueliq.com
THE TRUE COST OF FUEL LOSS
In the US, there have been over 548,000 confirmed cases of underground storage tank releases of hazardous substances, which are predominantly petroleum, and about 484,000 contaminated sites have been cleaned up as of March 2019 (Source: United States Environmental Protection Agency (USEPA)). The clean-up costs are approximately $1 billion USD per year. In Australia, a performance audit published by the Audit Office of New South Wales in 2014 reported that out of the 1,600 contaminated sites notified to the EPA, two-thirds of those are current or former petroleum industry sites, which includes more than 800 service stations.
By Russell Dupuy, Environmental Monitoring Solutions