Intelligent Energy Management System for Oil Wells Developed in Perm
An AI-driven control system created in Perm is cutting electricity use at oil wells by double-digit percentages without reducing output, highlighting how digital transformation can deliver immediate cost and efficiency gains in upstream operations.
Driven by Neuro-Based Control
Researchers at Perm National Research Polytechnic University have developed an intelligent energy management system for oil wells that reduces electricity costs by 10–12% by optimizing equipment operation. The system is based on neural network algorithms and mathematical modeling that analyze well parameters and forecast energy consumption. Based on these forecasts, it automatically adjusts the operation of electric submersible pumps, selecting optimal operating modes in real time.
Energy savings per well can reach up to 1.7 kW, and in so-called “medium” and “heavy” operating modes, up to 5.5 kW. At fields with hundreds of wells, this translates into annual cost reductions amounting to tens of millions of rubles. High electricity consumption, which can account for up to half of production costs, remains one of the oil and gas industry’s most persistent challenges, particularly amid rising power tariffs. Lower operating costs for producers ultimately affect end fuel prices and strengthen the economic resilience of producing regions. More broadly, the technology can reduce the industry’s dependence on electricity-intensive production.
Getting to the Root of the Problem
Although electricity accounts for roughly half of oil and gas production costs, up to 25% of this energy is consumed inefficiently. Traditional and outdated control systems often fail to optimize pump performance. They either do not account for constantly changing reservoir conditions or require complex infrastructure, making them economically unattractive. Implementing automated solutions typically demands a dense network of expensive sensors to continuously monitor all critical well parameters. This not only raises operating costs but also complicates maintenance.
The distinguishing feature of the Perm-developed system lies in its ability to select optimal operating modes that reduce field-wide energy consumption. Two controllers operating in continuous and periodic modes automatically determine optimal pump parameters, ensuring maximum energy efficiency.
Further research by the Perm Polytechnic team is focused on scaling the system and integrating it into industrial-grade solutions.
From Model Validation to Deployment
To verify the accuracy of the mathematical model of a centrifugal pump, which comprehensively accounts for hydrodynamic processes in the well and the technical characteristics of the equipment, real operational data from an active field were used as a benchmark. Measured values for various parameters, including fluid level and pressure, were sequentially fed into the model so that it could calculate projected power consumption. These projections were then compared with actual readings from industrial electricity meters.
The result was an accuracy level exceeding 97%. The system can simulate thousands of operating scenarios under different parameter combinations. The neural network was trained using the results of these simulations. Production rates and electricity consumption under changing conditions were compared with outputs from the previously developed mathematical model. A specialized training algorithm was created to reduce forecasting error. This process was repeated across thousands of scenarios until the neural network’s results matched the mathematical model with the required precision. As a result, by entering just four parameters, operators can forecast changes in production mode within seconds.
A Step-by-Step Evolution
The emergence of such effective solutions is the result of years of multi-level research and development. In 2023, Russian researchers published findings confirming that neural network-based automation outperforms traditional production automation systems. Industry conferences have increasingly focused on digital models for improving energy efficiency at oilfields.
Internationally, a solution combining machine learning and IoT has been demonstrated to boost heavy-oil production through optimized steam injection. Unlike traditional modeling based on classical physics, this approach tested advanced machine learning methods. An optimization system was also developed to recommend optimal steam distribution plans, increasing production by 3%. The product additionally addresses adjacent tasks, such as predictive maintenance.
The Perm Polytechnic solution continues this trajectory but places a clear emphasis on energy conservation and cost optimization, which is especially critical under high electricity tariffs. It represents a balanced integration of artificial intelligence and industrial automation.
