How to Minimize Greenhouse Gases with Predictive Maintenance

Imagine you’ve just been appointed the CEO of an energy supermajor and consider these daunting facts:

• The world has passed the point of peak oil demand.

• The COVID-19 pandemic has dramatically reduced the demand for travel, and COVID-19 is still a global threat.

The Paris Agreement (also known as the Paris Accords) reflects the consensus of world leaders today that swift action is required to reduce greenhouse gas emissions, the larger goal being to achieve net zero emissions by the second half of this century.

• Investors, the media, and the general public collectively demand swift and effective improvement from the energy sector.

That is why decarbonization — the process of reducing greenhouse gas generation in the pursuit of a net zero world — is at the very top of the to-do list for energy sector leaders worldwide.

predictive maintenance by energy organizations

Fortunately, AI-powered predictive maintenance offers one of the most effective ways to decarbonize.

To understand why, begin with the fact that in the energy sector any form of production interruption (a stoppage or outage) runs the risk of increasing the amount of greenhouse gases released into the atmosphere.

Excessive flaring — intentionally burning natural gas, often to prevent damaging infrastructure during a work stoppage or maintenance event — has been estimated as contributing between five and ten percent of global greenhouse gas emissions annually.

So how can energy organizations keep production rolling along as continuously as possible?

One answer lies in predictive maintenance: predicting future asset failures or performance shortfalls and creating maintenance strategies on that basis, instead of replacing assets at fixed intervals or (much worse) replacing assets because they have failed.

Making accurate predictions requires the production infrastructure be digitized (equipped with sensors that transmit performance data) so assets can be continually monitored for changes in operating conditions or performance. It also requires data collected in previous years be taken into account, to determine when and why production stoppages have taken place in the past, and then incorporate that insight into the new maintenance strategies.

This tsunami of data is tremendous and ongoing and already exceeds the capacity of individual analysts, however smart or experienced, to handle it. AI is the most effective way to respond to this problem, but not all AI is created equal and in particular, not all AI is capable of scaling to the necessary level.

The SparkPredict® product meets and exceeds the challenge. It can learn on its own over time, becoming more insightful and accurate, as it ingests more data without requiring human expertise to adjust its cognitive models. For the SparkPredict product, a data tsunami isn’t a problem to solve; it’s a resource to draw on.

That, combined with our extensive history of successfully engaging with top-tier companies worldwide to solve similar challenges, is why SparkCognition was recently selected by an energy supermajor to improve the efficiency of its offshore operations and optimize predictive maintenance.

The supermajor had undergone a substantial digitization transformation and had both widespread sensorization and a cultural awareness that data-driven decisions yield the best long-term outcomes.

They had made initial efforts to implement their own AI, but ran into difficulties. Specifically, their in-house cognitive models had shortfalls in data streaming, alerting, model retraining, and model management that made it unsuitable for the production environment.

With our expertise and our technology, these issues were quickly resolved. First, to demonstrate what could be achieved, SparkCognition aggregated multiple years of historical data for each subsystem and created models that correctly predicted 75% of the historical failures with an average of nine days advance warning. This level of accuracy substantially exceeded the supermajor’s success criteria of 50% accuracy and five days advance warning.

The approved model was integrated into the SparkPredict product, which governs data ingestion, model building and execution, alerting strategy, SME input, and retraining. The SparkPredict product was then deployed onshore in the supermajor’s remote control center. Since deployment, it now provides alerting, 10-minute diagnostics, and a significant increase in overall operational visibility.

Our technology continues to deliver results. For instance, one alert generated by the SparkPredict product led to an investigation that found a temperature sensor was providing faulty values. This was easily addressed, and led to a far more effective outcome. Previously, maintenance would have taken up to two days while also generating substantial costs in lost production and downtime.

As a result, this engagement is considered one of the largest successful AI deployments in offshore oil and gas and the supermajor is expanding the project to additional assets.

As successful as it’s been, however, it represents only the barest beginning of what must be done on a global scale by energy supermajors to address the ongoing problem of greenhouse gas emissions and their contribution to the global warming trend. We welcome the opportunity to help companies around the world in their struggle to build a smarter, better, and more sustainable world.

SparkCognition is committed to compliance with applicable privacy laws, including GDPR, and we provide related assurances in our contractual commitments. Click here to review our Cookie & Privacy Policy.