Oct-2022

Digitalisation on cloud

Most refineries are adopting cloud-enabled digitalisation to reap the benefits of technology in terms of scale-up, security, and availability of the applications and services to compress the decision-making cycle and empowerment of employees.

Jagadesh Donepudi, Ashok Pathak and Mike Aylott
KBC (A Yokogawa Company), Mumbai

Article Summary

Traditionally all these applications are used in silos, and the data sources are different; hence the interpretations are limited to specific tasks without a holistic approach.

With the advent of the Internet of Things, obtaining data from remote locations became easy by deploying sensors and wireless technologies. For refineries today, the data is centralised in the distributed control system (DCS) control rooms, and the same data is being used for controls, automation, and analytics. Industrial Internet of Things (IIoT) sensors allow refiners to add to the conventional data with new information, enabling improved pictures of equipment health.

With the application of linear models, like linear programming (LP), the supply chain cycle demand forecast, planning and scheduling have made it possible to procure the right crudes, process them in a given refinery configuration, and meet the demand scenarios.

Many times at the end of the month after backcasting, we observe gaps between plan and actual. The yields and qualities do not match with the actuals. The differences may be due to feed qualities, measurements, catalyst performance, and so on. Process simulation tools, such as Petro-SIM, help analyse these deviations, accounting for the non-linearities, mass balances, and measurements.

One of the solutions KBC is architecting is digital twins of the entire plant, including the major equipment, connected with real-time data and running the applications on a cloud-based platform such Yokogawa Cloud, hosted on a commercial/private cloud platform like Azure or AWS.

The engineering applications such as process simulation, in our case Petro-SIM and Visual MESA Energy Real Time Optimisation, reside in the containers where the data is received. The plant and calculations are carried out in the cloud, and the results are published as a key performance indicator (KPI) on the visualisation layer through a browser.

Digitalisation on Cloud
KBC’s digitalisation architecture is both flexible and agnostic of the cloud platform selected. In collaboration with Yokogawa, we are redeveloping our applications to work on Yokogawa’s standard cloud platform (see Figure 1), as well as enabling the component parts to plug into other cloud platforms such as Microsoft Azure or AWS using industry-standard containerisation and integration techniques:
• Standard Docker images for engines such as Petro-SIM
• REST application programming interface (APIs), enabling largely automatic integration with platform components such as asset models, time series databases, and relational datastores and enabling flexible data import and export
• Web-based user interfaces (UI)

This approach allows considerable flexibility in architecting solutions. Taking Petro-SIM process digital twins as an example, they can be implemented:
• On-premise, tying into site historian systems like IP-21 and LIMS
• Natively on a digital platform such as Azure tenant as Docker images that work with Petro-SIM models built on the desktop today and with fully web-based Petro-SIM in the future
• Natively on the Yokogawa platform deployed on Azure tenant and integrated with your systems of record
• Natively on the Yokogawa platform as software as a service (SaaS) application

Digital Twins Architectures
Refinery planning, scheduling, and control operations involve multiple models that need maintenance and, too often, existing processes are highly manual. Implementing digital twins enables running these applications automatically on digital platforms, as shown in Figure 2. Some digital twin calculations can run on fixed timers, such as hourly or daily intervals, while others will have to respond to triggers based on new data availability and other events.

The big picture of digital twins integrated with planning, scheduling, and recover time objective (RTO) on Lookahead mode enables obtaining the right set points to the APCs. Lookback mode provides reconciled data by comparing actual results with calculated parameters using data reconciliation and correction to get the right data to simulation models and updates of LP vectors in the planning models through recalibration of the models as required.

Digital Twin Workflows
The above digital twin applications are developed using KBC’s proprietary Petro-SIM flowsheet embedded with SIM reactor models. These digital twins are linked to data historians allowing them to gather and process data from the refinery data historian, enabling unit monitoring, visualisation, what-if scenarios, and other automation of the model upkeep.

Digital twins perform data reconciliation and calculate KPIs on a process unit, such as feed quality summaries, unit yields, process conditions, and reactor severities (such as conversion) for analysis by process engineers, and also the economic value of the streams with all the prices in place.

A digital twin will perform the following key tasks using a standard automated methodology:
• Retrieve and screen process and lab data
• Reconcile unit material balances
• Calculate key, unmeasured process variables
• Calculate a set of KPIs
• Provide the ability to trend actual versus simulation versus plan
• Summarise deviation of trends in the status report of KBC Explorer
• Generate a set of reports – material balance (raw and reconciled), data quality indicators, and unit KPIs
• Integration of digital twins with artificial intelligence/machine learning (AI/ML) to provide auto-calibration of the models and updating of the LP vectors

Use Cases
The Petro-SIM Explorer has been used for graphical trending of actual versus simulation versus LP sub-model values for given process variables, such as unit operation variables, yields, and product properties. The SIM model predictions are tracked and compared against the actual unit operations (see Figure 3). Cases are stored in an associated database to allow historical trending and monitoring of selected variables.

Value Realisation
Solutions like profit improvement and energy optimisation are taking advantage of all the technological advances already covered to turn what was originally a single engagement into a continuous process. Let us illustrate this through one aspect of our traditional refinery performance improvement programme: benchmarking, identification of gaps, and the opportunity roadmap for implementation. The value of identified opportunities is traditionally assessed using a base case operating scenario and pricing. Using a cloud platform, we can automate running the individual opportunity calculations against current conditions and pricing, letting you have a dynamic projection of value and making it easier to validate and select opportunities for implementation. We can also make it easy to modify the list of opportunities on the roadmap, bringing in new ideas or operating modes for ongoing analysis. This makes much of the consulting analysis evergreen.

Figure 4 shows some examples of the benefits obtained by different KBC clients from digital twin implementation.

Summary
ΠCloud-based architectures provide flexible storage, workflow, and compute methods, which allow for more flexible/changeable integration and automation
    Assembly of information in data lakes allows for insight from big data analysis benefiting from one source of the truth
Ž Maturity of AI engines allows us to use AI for a wide range of use cases, such as early identification of potential faults, to improve data quality during data normalisation, and to use ML to supplement first-principle models that drive much of planning and scheduling.

This short article appeared in the 2022 Refining India Newspaper, which you can view HERE