Apr-2011

Optimising safety relief and flare systems

Understanding the behaviour of refinery units by dynamic modelling of emergencies enables the prediction of realistic relief loads

Alban Sirven, Julien Grosclaude and Guillaume Fenol, Technip France
Jeremy Saada, Invensys Operations Management

Article Summary

Over the past decades, the refining industry has continuously moved towards higher levels of crude conversion and more stringent product specifications. When they are combined with aromatics production in particular, refinery schemes have become more complex. Additionally, nameplate capacity limits for grassroots plants have been gradually pushed upwards.

At the same time, along with dieselisation of the European vehicle fleet and the strengthening of product specifications, the refining scheme of European refineries has become increasingly complex. The addition of new units to an existing refining scheme affects flare systems, leading to a revamp of the existing flare network or to the addition of a new flare network and the consequences of altering the network.

Flare systems are primarily sized with regard to common failure modes. General electrical power failure (GEPF) results in a simultaneous loss of condensation for most process systems, and all corresponding individual relief loads are summed up to determine the required capacity of the flare systems.

In the context of both an increased number of interconnected process units and higher processing capacities, flare systems approach critical sizes when industry-standard calculation methods for the determination of individual relief loads are applied. To overcome the related issues in terms of mechanical and structural design, supply and constructability, as well as to satisfy the requirements of refinery turnaround and scheduled maintenance, the configuration of the relief disposal system for grassroots designs should consist of several flare systems, the largest with a main header diameter as wide as 100in or more and a flare stack as high as 200m.

The cost of providing adequate protection systems for any refining complex is substantial. At this point, understanding and modelling the dynamic behaviour of refinery units in emergency 
situations becomes necessary to assess the required capacity of flare systems more accurately than conventional calculations methods, which tend to produce conservative results.

Modelling refinery equipment behaviour in emergency conditions
The sizing of refinery flare systems requires prediction of the behaviour of equipment 
(or refinery subsystems) Calculation methods used for the determination of relief loads have, since the beginning of the refining industry, required static and semi-dynamic calculations.

Conventional approach
The conventional approach is based on the use of process data shown in the unit heat and mass balance, corresponding to steady-state operating conditions.

Semi-dynamic methods
Semi-dynamic calculations require the enhanced use of static modelling tools such as SimSci-Esscor Pro/II to better model upset scenarios. This method will correct the results of static methods, accounting for basic equipment design data or the change of key fluid properties between operating and relief conditions. This type of analysis also aims to evaluate the relief start/end time for critical equipment and systems. The result is less conservative relief loads than with static calculations, but the analysis cannot account for complex phenomena related to the transient responses of process systems to major upsets.

Dynamic methods
With the increasing performance of calculation tools, dynamic process simulators are now able to work on desk computers with reasonable computation time. Process dynamic simulations are based on validated thermodynamic models as well as dynamic heat and material balance equations. They also consider the response times of control loops, thereby predicting more realistic relief scenarios. Figure 1 shows a comparison of the accumulated relief loads used for sizing a flare system, obtained by these different methods.

In addition to the reduction in relief loads, dynamic simulation offers the advantage of evaluating the impact of different system response times on accumulated relief loads.

However, dynamic modelling is time consuming compared to earlier calculation methods. This is due in particular to the amount and diversity of input data that must be defined for the description of any system; for instance:
• Definition of process streams
• Operating conditions based whenever applicable on licensor information
• Detailed equipment design and geometry data based on vendor information
• Instrumentation, automation and safeguarding data.

Technip developed a methodology for targeting opportunities for dynamic simulations and set up a list of criteria to select candidate systems for the use of dynamic simulation. This methodology for a refinery flare study (GEPF scenario) is detailed in 
Figure 2.

Dynamic simulation software
Computer simulations were made using the SimSci-Esscor Dynsim tool from Invensys as the dynamic process simulator. The program provides a series of capabilities that enable the modelling of rigorous transient processes and facilitate the development of dynamic simulations for applications from process studies through to operator training systems.

The equipment models are rigorous dynamic models. A library of unit operations, equipment types, control functions and other algorithms have been developed, enabling the specification of models for high-level concepts.