Measuring, Analyzing and Improving ESPs

Darcy Partners, a technology scouting and innovation advisory firm serving the energy industry, recently featured NarrativeWave's work improving artificial lift and ESP production. Below is an excerpt from their series focusing on DMAIC methodology.

"We are live & pumping!"

Oftentimes during initial production, artificial lift is necessary to increase production rates. Within our member base, the two most common lift types used during primary production are electric submersible pumps (ESPs) and gas lift.

These two lift types tend to compete on several different categories, a few include high rate, reliability, OPEX, ability to deal with multiphase, and optimization capabilities (view ESP vs Gas Lift framework here).

What are some common reasons ESP operations experience downtime or operate inefficiently?

ESPs can fail for many reasons, some of the most common reasons we've heard include:

and last but definitely not least…

What can be measured, analyzed, and improved?

Recently, we caught up with NarrativeWave who provides a digital, knowledge management & analytics platform for automated anomaly detection and integrated operator diagnostic processes. NarrativeWave has been able to transfer some of the knowledge of working with ESPs on the geothermal side to Oil & Gas.

Below, they have shared categories that can be (and are actively) being tackled using digital tools.

Energy Efficiency:

It's no surprise this is the first category mentioned considering the amount of power issues ESPs have. Here, companies like NW can use analytics to establish the expected energy efficiency and detect deviations from that point. This approach can use pump curves at a reference standard, or it can also look for deviations from historical behavior. This is useful both to determine excess energy consumption, but can also be indicative of operating issues (For example, if an impeller is wearing out you will see an efficiency drop).

Failure Predictions:

Use known failures from past pumps (i.e. historical data & root cause analysis) to develop a data science-based analytic that can predict pump failures in the future.

Health Indexing:

Here, known behaviors like the number of starts/stops can be correlated to wear on pumps to establish a “health index”. SMEs in the space may use this type of index for other applications.

Anomaly Detection:

Here, operating anomalies can be detected across multiple signals (e.g. voltage, flow, current, pressure) to automatically identify behavior that is concerning. For ESPs, multiple methods can be used such as comparing pumps to each other, using a pump’s own historical performance to set the standard, and machine learning techniques where appropriate.

Read the whole article on, here.