Oil holding

4 of 3

Virtual pump sensor

Signal without lifting the equipment.

Machine learningTime seriesIndustrial telemetry

4 of 3

signal without lifting the equipment

The model reconstructs the signal of the fourth sensor from the three working ones — the equipment does not have to be lifted out of the well.

Client

Oil holding

Industry

Oil production

Areas

Machine learning, telemetry

ESTESIS role

Model development

From task to result

4 chapters

Task

Keep control over a downhole pump when one of its sensors has failed.

Replacing the sensor requires lifting the equipment out of the well — that means stopping production and costly work.

Client

An oil holding. Name under NDA.

What we did and how

We developed a virtual sensor — a model that calculates the readings of a failed sensor from the data of the three working ones.

  1. 01Analysis of historical pump telemetry
  2. 02Training and validating the model on actual readings
  3. 03Real-time signal calculation
  4. 04Monitoring model accuracy in operation

Result

4 of 3

signal without lifting the equipment

The pump keeps running under control, and there is no need to lift the equipment to replace the sensor.

Stack and competencies

Machine learningTime seriesIndustrial telemetry

More stories

Your task —next story

Tell us about the project — within 5 working days we will send an architecture and an estimate.

Accredited IT company · Russian Ministry of Digital Development registry No. 20142