Given that most oil and gas wells must go on artificial lift at some point during their production lifecycles, and with electric submersible pumps (ESPs) being one of the most efficient ways of doing so, it is no surprise that some 50% of offshore wells use them to maximize output as much as possible.

But even though ESPs are designed, engineered and built for rugged reliability in the harsh conditions of corrosive seawater and extreme deepwater pressures, they can fail. And when they do, the costs to repair or replace them are extreme, but usually dwarfed by the costs of lost production.

Actionable insights

That is why Siemens developed a predictive maintenance solution for remotely monitoring ESP performance by applying artificial intelligence (AI) technology.

What is AI? It is the use of computers to do work that typically requires human cognition and intelligence, but at a far greater scale than humans are capable of. Among many examples are pattern recognition—either in data, images, speech or music.

Compared to conventional approaches of ESP monitoring, AI-assisted monitoring can be transformational. That is because large amounts of data—many data points every second—can be processed with almost unlimited scalability. Taken together, all this data can provide a digital map of ESP operations, effectively creating smart pumps at the heart of a digital oilfield. And since it is vendor-agnostic and standards-based, this concept can incorporate other vendors’ equipment, too. It can also apply to all types of ESP applications.

Successful field test

Recently, for a large onshore E&P customer in Germany, Siemens conducted a successful test of a cloud-based, ESP monitoring solution that uses AI and Industrial Internet of Things (IIoT) connectivity. Currently, Siemens and a major offshore E&P are planning a similar proof of concept.

Today, an ESP’s sensing fabric draws from its automation and electrification systems while its SCADA system logs data into historian databases, mostly used for troubleshooting or forensics. Although deviations can alert operators to performance issues, this now happens only after an event occurs—when a potential production impact may already be underway.

In contrast, the Siemens ESP predictive maintenance solution brings together AI and cloud-based IIoT technology via the highly secure Siemens MindSphere IoT operating system. It uses an ESP’s streaming data as “fuel” to build an ever-richer ESP operating profile in these ways:

1. Anomaly detection. As ESP data streams 24/7 from the wellsite into a MindSphere database, advanced analytics and AI algorithms seek variances from expected behaviors of various parameters. Deviations are flagged and alerts sent to operators before a performance event occurs. Shown below is a graphical representation of the different types of ESP data being processed. An anomaly in the data source as indicated can reveal a potential failure 12 days before the actual failure of the ESP mechanism.

2. Behavior labeling. As data keeps streaming into the database that holds the ESP’s ever-more precise operating model, machine learning occurs as the pattern recognition and statistical algorithms get smarter over time. Here, the Siemens Artificial Lift Suite software and the MindSphere cloud model’s advantages kick in: operating data from ESPs worldwide can be aggregated and analyzed to label ESP behavior profiles, specific to their applications and environments. These not only can flag behavior anomalies in one ESP, but also alert operators of ESPs in similar applications and environments, delivering even more advanced notice of an emerging issue to be aware of.

3. Predictive maintenance. Given the real-time feedback loop between an ESP and its cloud-based operating profile (i.e., its “digital twin”), ESP operators can deploy predictive maintenance models that use proactive, condition-monitoring to provide them with decision support about how to address impending issues. This can ensure greater ESP availability and uptime while saving spare parts and labor. Costly disruptions can be avoided.

AI’s potential is just starting, with many new applications expected in the future to help optimize asset utilization and lower production costs for greater profitability across the oil and gas industry. The ultimate goal of applying AI in the digital oilfield is to improve decision support, so ESP operators can know what to prevent production disruptions.