Johan Sverdrup’s Digital Operations Drive Efficiency, Safety
The complete paper describes the development of the “digital field worker” at Johan Sverdrup, an initiative that has changed the approach toward not only construction and completion but also operations.
The sheer size of the Johan Sverdrup field development (expected recoverable reserves of 2.7 billion bbl, with projected operations of more than 50 years) has earned it designation as the operator’s digital flagship. The field has driven digital solutions and work flows that have the potential to be scaled up for the benefit of the operator’s entire portfolio. The complete paper describes the development of the “digital field worker” at Johan Sverdrup, an initiative that has changed the approach toward not only construction and completion but also operations.
The field is approximately 150 km west of Stavanger in the North Sea in a water depth of 110–120 m (Fig. 1). The top reservoir is approximately 1800 m below mean sea level. The reservoir is characterized by hydrostatic pressure and undersaturated oil with a low gas/oil ratio. A predrilling campaign included eight oil producers and 12 injectors. Additional wells will be drilled to achieve the expected Phase 1 oil production capacity of 70,000 m3/d.
The field has been developed in two phases. Phase 1 was approved by Norwegian authorities in 2015 and includes four bridge-linked platforms, all with jacket substructures. Primary power to the field is supplied from shore. The full field development includes 64 wells, both platform and subsea wells. The field came onstream in October 2019. Ramp-up of the predrilled wells proceeded smoothly to a field-production level of approximately 350,000 B/D. Phase 2 of the development was approved in 2019 and is expected to start production in Q4 2022.
At peak, this field will account for approximately one-third of all oil production in Norway with record low emissions. The expected full-field (Phase 1 and 2) production capacity is 105,000 m3/d at peak. Breakeven price for the full-field development is less than $20/bbl.
The Field’s Digital Solutions
Digital Twin. Echo is the operator’s digital twin, a 3D model used for planning, support, optimization, and followup of operations and maintenance. The first version was released in October 2018. In this version, life-cycle information, maintenance programs, and bolt-tension data were made available by highlighting equipment in the 3D model. The digital twin is modularized, with a shared Unity codebase across the devices, and extends from augmented reality to immersive virtual reality accessible on the current devices. The development includes sharing of the environment with other users to allow real-time cooperation.
Digital Field Worker (DFW). The device chosen for field use is a minitablet. It was selected after consideration of criteria such as security, weight, size, screen, user feedback, speed, battery lifetime, and functionality. A shoulder strap and small bag were developed for field safety. At the time of writing, field workers have access to numerous apps, websites, and designated software products to be used in the field. In March 2017, the new approach was introduced at building sites; in May 2018, it was introduced offshore during Phase 1 hookup and commissioning operations.
Safety. Examination of reports from major accidents and serious incidents reveal common characteristics. These contributing factors are some of the key elements that the operator is striving to improve through digitalization practices. Fast, efficient sharing of real-time information of high quality and relevance is key to the DFW process, as is knowledge availability and accountability.
Efficiency. High production efficiency is a high priority. This is ensured by reducing the time it takes to start up production by using a digital procedure. The coordination between the central control room and operators using an app on the tablet is much more time-efficient because all involved receive real-time information about status and the actions of colleagues, with a resulting assurance of procedural compliance.
Low Carbon. The DFW process has been developed as a digital training program for all disciplines in the field’s offshore operations and maintenance groups. Training and minor tests can be performed on tablets remotely, reducing the need for staff travel and thereby reducing carbon emissions. Similarly, the need for specialists and suppliers to travel offshore has been reduced by using digital devices to discuss issues and share videos and other information in real time.
Advanced Production Optimization (APO). APO applications range from simple, single variable controllers to true online optimization applications, thereby using the entire range of the automation hierarchy for improved operation. The application of automatic production optimization has the potential to increase production and lower the energy consumption typically by 2–5%. APO is also a key component in safer injection and production.
Automatic control and optimization enable more-efficient work processes because production targets can be set and monitored directly. This can reduce the level of workload for the central control room operators during startup and normal field operation. This process will be implemented in Phase 2 of the project.
Downhole Fiber Optics. Downhole fiber-optic sensing has been installed in the field’s wells. A fiber-optic cable is attached to the tubing to monitor temperature, acoustics, and pressure. The key data types considered are distributed temperature sensing and distributed acoustic sensing. The goal is to use distributed fiber data to improve health-, safety-, and environment-related surveillance and production and reservoir monitoring in real time.
Permanent Reservoir Monitoring (PRM). PRM provides frequent seismic data of higher quality for better well planning and reservoir management. Four-dimensional data, together with other relevant data, will be used as input to well planning, simulation models, increased-oil-recovery production optimization, and updates of geological and structural models. Ongoing digitalization initiatives target efficient use of PRM data with a focus on improved acquisition, processing, data access, and analytical methods. The PRM system strengthens reservoir- and well-monitoring capabilities by complementing downhole fiber optics to enable real-time reservoir surveillance.
Machine Learning. While the common approach in the industry is to use machine learning to predict specific failure modes earlier, Johan Sverdrup’s approach is different. It proposes that a higher value can be realized by leveraging the technology to allow for condition monitoring of equipment through the use of more sensors. To achieve this, an adequate solution had to be designed while taking into consideration the end-user and business-case, machine-learning-design, and information-technology perspectives.
From the first of these, the main challenge is to allow engineers to move from soft alarms based on thresholds and first-principle equations to alarms based on machine learning. As a result, far fewer false alarms are expected to be generated, which will allow for monitoring of more equipment without increasing the number of man-hours.
From a machine-learning-design perspective, an unsupervised approach was necessary to allow for the lack of historical data. A deep autoencoder was chosen that can be trained on the healthy state of the equipment and can learn the underlying relationships between the sensors. The autoencoder is now used in daily operations and is asked to calculate the difference between given input sensor readings and predicted outputs, called the reconstruction error. Once the error increases, the monitoring engineers see that the equipment is behaving differently from the trained behavior, and an anomaly is identified.
From an information-technology perspective, it was necessary to find a technology that allowed training and serving of thousands of models simultaneously. The operator’s data-science platform, which uses components from designated software suppliers, was used to implement the machine-learning pipeline.
With several successful proofs of concept on wells, rotating equipment, and virtual equipment (a collection of sensors defined as a piece of equipment), the field is planning to introduce machine-learning technology in 2020 to all equipment types subject to condition-based maintenance.
This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper OTC 30477, “Johan Sverdrup: The Digital Flagship,” by Paal Frode Larsen, Tor Tønnessen, and Florian Schuchert, Equinor, et al., prepared for the 2020 Offshore Technology Conference, originally scheduled to be held in Houston, 4–7 May. The paper has not been peer reviewed. Copyright 2020 Offshore Technology Conference. Reproduced by permission.