Every day, Origin manages more than 2,000 gas wells across our gas operations in Queensland’s Surat Basin, supplying gas to large customers such as manufacturers on Australia’s east coast and feeding into the Australia Pacific LNG export facility on Curtis Island in Gladstone.
From time to time, we need to adjust the amount of gas we produce from our fields due to changes in customer demand or to complete scheduled maintenance on infrastructure. Monitoring the wells and deciding which ones to turn down or temporarily suspend used to be performed by one of our operators in charge of analysing up 30 wells a day. But data and AI are now enabling more accurate decisions and helping us to keep more gas in the ground for production at a later time.
Aleta Nicoll has a control engineering background and manages Origin’s Brisbane Central Control Room (BCCR), from where our team remotely operates every gas well, pipeline and processing facility in our upstream operations.
“The BCCR enables us to remotely control our gas field production in Queensland and we’ve been adapting Artificial Intelligence-led machine learning to predict which wells have the best prospects of recoverability after being turned down to meet fluctuations in customer demand,” Aleta said.
“The work we’re doing with digital is starting to marry machine learning with control system design – it’s super exciting!”
A world-first AI production optimisation tool developed by Origin’s digital team takes data such as pump speed and flow rates and decides which wells can be turned down and turned back on again with the lowest probability of failure, which means that we can reduce the number of workovers and extend the life of each well, which is important to help manage costs.
We are also able to reduce our emissions from flaring by analysing data from thousands of wells simultaneously meaning we can shut off production, confident that wells will return to production and we will meet customer requirements. Once the tool is fully operational, we hope to avoid over 500 TJ per annum of flaring through optimising well production, avoiding roughly 25,000 tonnes CO2-e per annum.
Erin Musk delivers key technology projects such as the production optimisation tool as the digital manager of Origin’s Integrated Gas team. Erin said using AI and machine learning helped to make more informed decisions about which wells could be turned down and turned back on without impacting on their performance.
To train the model and develop the algorithm, more than 4,720 historical events were used. The model itself comprises 10 terabytes of operational data – or equivalent to storing nearly 3.8 million songs! The control system receives 881 individual data points for each well every 30 seconds.
“Analysing data and seeing how we can apply that learning is a key part of this – where a human operator could monitor production rates and forecast future maintenance for about 30 wells a day, this tool can do thousands in real time,” Erin said.
“I’m inspired by the creativity of our people who bring new perspectives, insights and approaches to some of our biggest challenges,” she said.
Origin’s generation team has also applied machine learning across the business, with real-time optimisation at Eraring in NSW – an Australian-first digital tool to optimise power plant performance which has the potential to have a huge impact on emissions.
During FY2019, more than 136,000 tonnes of CO2-e were avoided at Eraring and once fully implemented, there’s potential to avoid a total of 1 million tonnes of CO2-e between 2020 and 2025.
Data, AI and machine learning are helping us to deliver energy that is cleaner and smarter.