Mar-2021

AI in refinery modelling

Hybrid modelling combines artificial intelligence and processing expertise to develop accurate models of refining processes.

RON BECK
Aspen Technology

Article Summary

Artificial intelligence (AI) and machine learning are rapidly emerging as tools that can greatly accelerate the ability to employ plant data both to calibrate first principles models and to create data based models of phenomena and processes quickly. With net zero carbon initiatives, the imperative to adopt circular economy principles, the rapid pace of energy transition, and ongoing economic turbulence, process companies are driving rapid innovation in refining and chemical plant processes. Many chemical manufacturers are pursuing concepts of process intensification. To support this rapid change, process modelling and optimisation must advance rapidly.

Hybrid models address this need. They combine AI, first principles, and domain expertise to deliver a comprehensive, accurate model of new and complex processes quickly. This approach lowers the bar towards applying AI in the process industries. The hybrid model approach harnesses AI, combining it with domain expertise to create the guard rails that make it work safely, reliably, and intuitively.

Machine learning is used to create the model, leveraging simulation, plant, or pilot plant data, while using domain knowledge including first principles and engineering constraints to build an enriched model without requiring the user to have deep process expertise or be an AI expert. This next generation of solutions democratises AI within hybrid models to optimally design, operate, and maintain assets online and at the edge.

AI and machine learning allow us to build a model analysing a broader set of data while leveraging advanced data science techniques for model prediction. When combined with engineering principles and domain expertise, models can be built and maintained more quickly than traditional methods without requiring significant user expertise.

With hybrid models, users can model processes and assets that cannot easily be modelled using first principles alone. Examples include:
- Batch processes, which can be too varied to model systematically
- Fluidised bed processes with complex chemical and fluid behaviour
- Bioprocess reactors and fermenters
- Complex refining units

Users gain the accuracy of empirical models and the strength of first principles models, leveraging the power of AI paired with domain expertise, to create a more predictive model faster and with less experience required.

Hybrid models provide a better representation of the plant, which keeps the model more relevant over a longer period of time. This reduces the barrier to entry for using modelling for asset optimisation by requiring less effort and expertise. With the models in place, the connected worker becomes free to perform higher value added and strategic work.

Today, we are seeing hybrid modelling capabilities increasingly deployed across software solutions through a model alliance approach, which synchronises fit for purpose models in different functional areas needed to safely, reliably, sustainably, and profitably operate an asset. An example of the model alliance is the use of reduced order unit models in planning, dynamic optimisation, and online equipment monitoring, all derived from the same root refining unit operating data set and simulation model, achieving closed-loop production optimisation.

The following summarises the business challenges this new technology solves, the three types of hybrid models we are seeing introduced to the market, and some of their key benefits.

Current business challenges
The process industry faces unprecedented uncertainties and macroeconomic threats. Process industry leaders face unparalleled volatility in all phases of their business. External factors including hydrocarbon price turbulence, changes in remote working needs, and supply chain interruptions are making change inevitable for process manufacturers, from the smallest to the most global.

Addressing challenges ranging from changes in feedstock price and demand to society’s drive towards sustainability, organisations must weigh complex trade-offs. Software technology, and in particular AI, is widely viewed as one of the primary tools available to equip organisations to thrive amid these challenges.

Market volatility and energy transition
A trio of external forces are forcing continued volatility and turbulence on energy and chemical companies globally. The global market supply and demand shock and economic recovery we are entering, the societal drive for energy transition and carbon zero industry, and the social contract driving for zero casualties and environmental incidents all have a massive impact on industry executive teams’ thinking.

Process companies are fixated on flexibility, strategies for resilience in producing at unpredictable utilisation factors and with extended maintenance intervals, yields, and operating margins. Faster models, solving key economic units or entire sites rapidly, tuned better to plant operating conditions, answer the crucial questions needed to achieve those goals. Hybrid modelling makes it possible to model and deploy quickly — even remotely — to address dynamic market forces and asset conditions. These models become key ingredients to transform operations through the future self-optimising plant.

The disappearing expert
As a generation of experts retires, process organisations face a gap in essential knowledge and a new generation of workers without the patience to develop that critical expertise. Hybrid models, embedded with AI, address those gaps, creating immediate value for organisations and assets. All but those enterprises with the deepest pockets need the ability to build and deploy these models without scarce and expensive experts.

Sustainability pressures
Formidable decarbonisation goals across industry will not disappear after the current economic cycle. The pressure to move towards a circular economy also creates many innovation challenges. Hybrid models provide the ability to optimise and evaluate optionality across a wide asset scope, to select the best strategies to meet these goals. Companies are challenged today to contend with the complexity that sustainability pressure imposes on their operating and strategy decisions.