Jul-2017

Crude fingerprinting and predictive analytics

Predictive modelling technology has been developed to predict the behavioural characteristics of crude oils and their blends

CONRAD TERAN
GE Water & Process Technologies

Article Summary

It is well understood that crude oil is a complex mixture of a variety of hydrocarbons and impurities with varied polarity and polarisability, and is prone to cause fouling,^1,2 emulsion breaking and corrosion related issues during refinery processing. The issues arising from the processing of tight oils and other opportunity crudes will likely become more challenging as new crude oils are brought online due to improved and/or more aggressive production methods, transport and blending strategies.

The safe and profitable processing of crude oil is driven by the combined effects of best practices, optimal process conditions, reliable equipment, crude oil behavioural characteristics and effective chemical treatment programmes. Any uncontrolled variation resulting from our inability to predict it, detect it or adjust it will result in one or more costly processing issues.

Of all these factors, predicting, managing and controlling the impact of crude oil variability on behavioural characteristics and associated processing issues is perhaps the most difficult. This task becomes near impossible if we are restricted to the use of traditional crude assays and other database methods at a time when crude oil names are becoming less and less relevant in predicting actual behaviour. For example, Figure 1 shows samples from seven different crude deliveries to the same refinery on the same day, all labelled “Eagle Ford crude”.^{3, 4}

In every case, incremental variability and uncertainty will come from the commingling of incoming crudes with tank heels, slop and other crudes in the system. The actual characteristics of the final charge coming out of tankage are typically not known for certain in a time relevant fashion.

GE Water & Process Technologies (W&PT) has developed field methods and robust analytics to predict and respond to crude processing issues, irrespective of the source of the crude. These methods and predictive analytics have been integrated into a cohesive predictive modelling technology system called CrudePLUS.

Key drivers of crude processing issues
Crude oils can be characterised by their structural fractions, known as saturates (SAT), aromatics (ARO), resins (RES), and asphaltenes (ASP), also commonly known as SARA, which are determined based on their polarisability and polarity. Typical properties and nature of these fractions are summarised in Figure ².
The complex interactions between SARA fractions and the presence of solids and other contaminants drive a fluid’s behavioural characteristics, which in turn drive crude processing issues.^5,6,7
In the context of CrudePLUS, three crude oil behavioural 
characteristics are the key drivers of crude processing issues:
• Instability/incompatibility: the stability of a fluid is defined by its capacity to maintain asphaltenes and colloidal particles soluble or dispersed in the bulk fluid. Conversely, instability refers to the capacity of a fluid to destabilise itself or other fluids upon blending.
• Emulsification tendency: defines the tendency of a fluid to emulsify or to resist demulsification at typical conditions.
• Fouling potential: defines the tendency of a fluid to foul preheat exchangers and heaters at typical conditions.

Instability/incompatibility, solids and other contaminants magnify emulsification and fouling effects. Figure 3 illustrates the negative impacts of these key characteristics on a typical crude unit train.

Characterisation of crude oils and simulation of processing issues
From a practical standpoint, while the quantification of SARA fractions and contaminants does provide useful insight on the directional behavioural characteristics of a fluid, these parameters by themselves in most cases, even when used in conjunction with other physical properties, typically do not provide a consistent or expeditious way of predicting the key drivers of crude oil processing issues.

The path for the determination of fluid behavioural characteristics and the design of effective mitigation solutions starts with laboratory analytical testing and simulations that characterise the crude sample and realistically stress the fluid to emulate and quantify its potential processing issues.
The data generated from these lab procedures and the associated research and development, supported by field pilots and validation, are the developmental foundation of CrudePLUS technology.

Figure 4 illustrates analytical testing and application simulations and their relationship to specific fluid behavioural characteristics.

Crude oil characterisation includes procedures such as PIONA and SARA, API gravity, TAN, viscosity, sulphur, metals and other industry standards as well as W&PT’s proprietary procedures. On the simulation side, W&PT proprietary methods and techniques are implemented utilising a combination of proprietary devices such as the W&PT Static Desalter Simulator and commercially available devices such as the Turbiscan (Formulaction, Inc.) and the Hot Liquid Process Simulator (Alcor by PAC).

Over the years, these simulations have proven to be reproducible and reliable in emulating the processing issues observed in the field, significantly contributing to the understanding of the underlying mechanisms, the development of best practices and novel chemical mitigation solutions.

From lab simulations to predictive models
The time required to test crude samples and effectively predict and respond to processing issues makes laboratory simulations in most cases impractical and very time consuming for day-to-day proactive field use. More than five years ago, W&PT initiated a project to develop a field deployable predictive modelling solution to deal with the identified limitations.

Defining the Xs and Ys for the predictive modelling solution
Any predictive modelling approach requires a clear definition and understanding of each of the responses (the Ys) to be predicted and of the parameters (the Xs) to be used as predictors, where each Y is a function of the significant Xs and their transformations.

For every sample, the applicable outputs from the lab simulations (the Ys) are transformed into indexes representing the level of severity of the behavioural characteristics (instability, emulsification and fouling) of a given oil sample. Other Ys targeted for prediction include a selective number of physical properties (SARA fractions, API, TAN, viscosity and others).

A field deployable infrared spectroscopy device was selected as the generator of Xs for every fluid sample. Specific methods, techniques and algorithms were developed to process fingerprint spectrum, extract the relevant data and assure repeatability and reproducibility error at or below +/-5%. The thousands of processed Xs become the unique identifiers or “fingerprints” of the tested oil.