Machine learning in retail petroleum

Russell Dupuy 16 Apr 2018 By Russell Dupuy

Machine Learning in Retail Petroleum

Gone are the days of buying a set and forgetting electronic equipment, especially intelligent embedded electronics that someone [maybe you] very likely paid a handsome price for..!

Connecting once standalone electronic Things is rapidly gaining pace that is expected to far exceed the take-up of the Telephone, TV, Mobile technology and the Internet.

Machine Learning was first coined by Samuel Arthur of IBM in 1959, a computer gaming pioneer. IoT has advanced this concept to reality through Thing connectivity.

The volumes of IoT data is driving pattern recognition coupled with computational learning theory establishing data-driven predictive insights for Petroleum retailers – disrupting traditional re-active asset and Wetstock management.

More pragmatically, this means simple and complex machine fact data [derived from various embedded electronic Things] is providing Petroleum retailers patterns to equipment behaviour to failure and/or equipment issues that impact consumer satisfaction.

Retail petroleum networks feature some very intelligent Things that for years have done their basic job, more often in silence, or worse turned off, or only looked at when they broke resulting in outages and consumer dissatisfaction!

The typical retail service station features; an Automatic Tank Gauge, Electronic Price Sign, intelligent Submersible Turbines Pumps, various measurement sensors, a Forecourt controller, Dispensers, a POS/BOS – all to simply manage fuel movement from underground tanks to the consumers’ vehicle, with the primary focus on managing the financial transaction.

The more modern retail service station features; [in addition to the typical] Pie, Coffee, Slurpee, Car Wash, Outdoor payment, Camera surveillance, security, digital media at the hose to mention just a few ‘Things’ that are becoming increasingly more intelligent through embedded electronics supporting IoT connectivity.

What is most surprising, is the number of individual digital connections into the retail service station; Security, OPT, Secure Bank payment, Vending and other Things as noted above, which all amount to multiple ADSL or GSM connections incurring monthly carriers costs. Clever technically yes, clever financially no. Then there is the support issue of many proprietary systems, which seems to go back in time.

We instigated a research project to connect a whole lot of Things together within a retail petrol environment, to monitor machine behaviour for an extended period of time, without changing and adulterating anything at a Thing level. Meaning if the Things were programmed incorrectly so be it, we then extended our reach to literally hundreds of sites and extended our monitoring period to over twelve months [which continues]!

What we learnt individually by each Thing was kind of expected, what we learnt through blended datasets was very insightful in dimensions we had not even considered. When these Things are monitored in real time, at a raw data level across a network [high population level], is that there is significant Machine Learning that drives significant operational improvements that has measurable and sustainable profit performance improvements for the typical petroleum retailer.

In our example, Fuelscan® is our embedded edge Thing, that connects to and translates the many and different proprietary protocols, that messages data through to Fuelsuite Cloud where Users visualise information to proactively manage their retail network.

If you would like to learn more about Fuelscan® and Machine Learning for your retail petroleum network, connect with us via or visit