How to Improve Production and Flow Assurance Management with a Digital Twin

1 hour

The formation of hydrates and other solids like waxes, asphaltenes and scales, is a major concern in the oil and gas industry, that could result in decreased production levels, safety issues, and even plant shutdowns. To prevent or mitigate the risks associated with flow assurance, operators often inject chemical inhibitors and retardant into the flow. Whereas for hydrates inhibitors, these chemicals are fairly well established and known, for waxes, asphaltenes or scales, the use of chemicals may trigger or enhance dangerous side effects.

This webinar will cover:

  • how data driven production and flow assurance can achieve significant savings through adaptive smart production management
  • optimal chemical injection with flow monitoring
  • dynamic adaptation of operating strategy to optimize production
  • the challenges posed by the changing conditions in operating assets
  • the integration of well-established flow assurance modelling packages with supervisory control & data acquisition systems, to provide enhanced monitoring and management capabilities over the occurrence and mitigation of flow assurance risks


We didn't have time to answer all the questions on our webinar so we've answered them below.

How do you distinguish wax deposition and emulsion occurrence in wells with a high water cut?

From the modeling perspective, the phenomena are very well distinguished. So, even if the phenomena overlap, their contribution should be obvious. From a predictions’ prospective at least.

From the operational point of view, if there is limited data, for instance, only pressure drop increasing over time, there might be ambiguity, since both phenomenal would presumably lead to an increase in the pressure drop.

To break the ambiguity, when you are using the runtime model as an exception-based surveillance monitoring system you need to find some other correlated evidence to distinguish the two.

If you see an increasing pressure drop, while the model is predicting small or no-increase due to wax build-up, you may conclude one of two things

  1. The model is going off-tuning.
  2. Something else is going on such as increased viscosity due to emulsions or foams formation, scales etc.

Either way, the system helps you troubleshoot the case.

You can also include the formation of a stabilized emulsion in the model if necessary. Again, by overlapping the predicted effect of wax and emulsions, you may be able to narrow down the choice even further.

There are several areas that you might need to look at in the operating asset and many of them will have similar effects on a macro-level. The simulator can help address some of them to narrow down the options. Then correlating with other factors, like compositional variations, changes in GOR, or water cut, you might be able to pinpoint the problems. For more complex cases, integration with data-driven approaches, like regression and AI could also help. Insights from simulations will provide further details that can help making sense of the data.

Other than wax deposition/hydrate formation, how will scale deposition affect the modeling?

Scales formation is another item on our radar and roadmap for further developments. For the moment, we can add the risk of erosion/corrosion and we have some limited capabilities for monitoring the precipitation of halide scales. We are looking into the possibility of extending our capabilities further and through collaboration with OLI Systems, we are addressing modeling scales as a start. That capability is already included in Petro-SIM® for process equipment.

Is production chemical (SI, PPDs, asphaltene inhibitors, etc.) injections taken into consideration with modeling?

We have been in conversations with several chemical companies to model, for instance, the effect of wax and asphaltenes retardant. For wax, we do have some ways to take the effect into account, by reducing the wax deposition rate as an effect of a wax inhibitor. This is not yet available for asphaltenes. Unfortunately, the modeling of such substances is not as well established as for the hydrate inhibitor. There’s a lot of customization in the formulation for each crude. It’s therefore difficult to use a common modeling approach to predict the effect of an entire product class. We are definitely looking into it however.

Could this approach of optimization be carried out without using a digital twin?

A lot of people already use offline simulators for optimization to an extent. Maybe not as often as necessary and surely not in a simple way. It requires downloading data, setting a model up, tuning and comparing predictions, or making new predictions. Obtaining information this way may be hard to share and end up being unusable for operators on site. The workflow that we implemented could indeed be done manually and it’s on purpose mimicking operations that are familiar to the flow assurance and production engineers. But there’s a huge value in having the tools available and running alongside the asset management software. It can automate some processes and warn operators in case things deviate from initial expectations, helping operators making informed choices and adapting their strategy as things change.

How often do you consider the solution of direct parameterization instead of complex model application?

If the modifications aren’t too large, somewhat predictable, and of a certain nature, correlations and parameterization may work well enough. With no need to get into any advanced regression or ML technique, you could imagine correlating, for instance, the inhibitor injection to flow rates and be done with it. That’s true if the case is simple and modifications in the asset are small.

However, when you have different fluids with different compositions and properties, comingling with changing and varied flow rates, it becomes much more difficult as the problem becomes multi-dimensional. Add to that the possibility of changes in the plant layout along with produced fluid modifications, and the variability can easily become too large to predict the outcome with some simple correlations.

If the solution is simple enough to deploy, the use of a rigorous predictive model is advantageous, since it would cover for the possibility of a simple as well as complex problem to solve. It provides reliability and flexibility because you can cover a lot more potential issues than a simpler approach.