July 2020

Process Engineering

Flare system design for a refinery mega-complex—front end and beyond

Estimating flare loads for a greenfield facility is often a challenging activity in the front end design phase.

Chaudhuri, S., Singh, R., Bechtel India

Estimating flare loads for a greenfield facility is often a challenging activity in the front end design phase. Flare system design evolves over the project phases, starting with configuration; feasibility study; licensor selection; basic engineering; front end engineering design (FEED); engineering, procurement and construction (EPC); and commissioning. Relief loads firm up at the EPC stage, once all the supplier data impacting the load estimation is available.

The implementation of the flare system involves a significant amount of investment in terms of capital cost and is an important consideration in the plot plan. If a significant change to the flare load is identified at a late stage of project development, the cost involved in design modification can be considerable. The challenge is to make early decisions with limited availability of project-specific data, while avoiding overly conservative design that adds costs and simultaneously limiting changes that may impact the overall project execution (FIG. 1).

FIG. 1. Key flare design parameters.

This article suggests a method for flare system design that is based on a mix of first principles and a database of past project relief load and flare system design best practices. A good first estimate, backed by smart strategies to mitigate changes, is key to generating a solid design. In the following sections, an overview of the typical refinery flare system is followed by a step-by-step approach for flare system design at each project execution stage. Several flare load management strategies that have proven effective in mega-scale refinery projects are given.

Overview of typical refinery flare system

The flare system is designed to collect and safely dispose of a wide range of fluids at various conditions during normal or upset operation of processing plants. These relief fluids are typically flammable and, in some cases, corrosive and/or toxic.

Typical components of a flare system include:

  • Pressure safety valve (PSV) tailpipes, subheaders and headers inside battery limits (ISBL)
  • ISBL flare knockout drums (KODs) and pumps
  • Outside battery limits (OSBL) main flare headers
  • Flare area KODs and pumps
  • Liquid seal drums and flare seal
  • End and emergency purge
  • Flare riser, tips and associated hardware (fuel gas, ignition, steam)
  • Associated monitoring and safety systems.

Each of these subsystems requires a different level of maturity in relief load data before it can be considered finalized for proceeding with procurement and construction. This aspect will be discussed in subsequent sections. A schematic of a typical refinery acid gas flare is depicted in FIG. 2.

FIG. 2. Schematic of a typical refinery acid gas flare.

Stepwise approach tailored for project execution needs

Refinery mega-project execution approach is similar around the world, with relatively minor differences based on contracting strategies. Flare system definition requirements and good practices for greenfield flare system design are described in the following subsections (FIG. 3).

FIG. 3. Project development roadmap.

Configuration and feasibility study. The requirement for typical greenfield projects is to have a preliminary plot plan and to achieve a Class 4 (+/– 30%) estimate. At this stage, the estimate of flare system capacity and configuration is based on data from similar past projects and any local regulatory requirements. High-level flare system configuration decisions should be taken at this stage. Typically, such decisions include:

  • Ground vs. elevated flare
  • Segregation and number of flares [e.g., high pressure (HP), low pressure (LP), acid gas flare, low low pressure (LLP), low temperature (LT)]• Sparing and maintenance requirements
  • Mounting flares on a common derrick vs. a mix of derrick and guy-wired flares, etc.

It is critical to ensure that plot plan requirements are adequately considered to avoid the need for additional land at a later stage. It is not uncommon for projects to underestimate plot requirements for flares at the configuration and feasibility stage and end up compromising on other owner requirements to fit the facility inside the available plot space.

A database of major individual relief loads for similar technologies/facilities with comparable capacities is essential for facilitating early decisions in this phase. Significant relief loads for any unit come from a few specific pieces of equipment; these loads remain fairly consistent. It is important that the reference unit capacity is of a similar order of magnitude as the one to be estimated. A relief load from a 200-bpsd diesel hydrotreater will not be 10 times that from a 20-bpsd diesel hydrotreater. The licensor or technology provider selected for the flare loads is another critical consideration. Flare load from two leading technology providers for the same unit might differ completely.

When a reference unit for the selected licensor has been selected, along with a rough order-of-magnitude capacity and feedstock type, an estimate can be derived with a simple arithmetic pro-ration. Note: A risk of oversimplification exists for this method. It is a common benchmarking principal used by licensors or open-art FEED contractors across the industry. TABLE 1 represents a comparison of gas oil hydrotreater FEED level relief loads for a general power failure case from a single licensor.

HP flare load is much higher in the reference project than the project under design, whereas LP flare load is only 20% different than the actual relief load shared by the licensor. This difference illustrates the importance of critically analyzing reference project data with regard to the project under design to pinpoint significant configuration differences or preexisting mitigation measure incorporated in the licensor design.

Selection of flare type is often dependent on client preference (FIG. 4). Elevated flares are common in refineries because they are low in cost and sturdy in design. However, refineries close to human habitation often select ground or enclosed flares to control noise and flame exposure to the communities.

FIG. 4. Major factors influencing flare configuration.

Licensor selection and basic engineering package. An important part of a flare package is finalizing the relief and blowdown philosophy, including the segregation/number of flares and the estimated backpressures for each flare. Selected licensors can be asked to share relief summaries from their databases before they calculate project-specific relief information. Some useful pointers in this exercise are:

  • Look for the ratio between minimum and maximum set pressure. If the ratio is five times
    or higher (which is almost always the case for complex refineries), then segregated HP and LP flare systems are justifiable.
  • When combined single-flare header size is greater than 80 in.–100 in. with a backpressure of 1.75 barg (approximately 50% of 3.5 barg), it is generally economical to segregate flares in more than one system.
  • When LP and HP flare system segregation is practicable for design, there are significant savings in choosing the sweet spot for setpoint changeover. Generally, the setpoint cut-off between the HP and LP flares is 100 psig (i.e., approximately 6.9 barg). The breakpoint is a nominal value to start the flare system design, which must be reassessed by the end of FEED when reliable relief loads from various process units are available and LP/HP flare system capacity limits are nearly final. The ideal breakpoint for a refinery gives a higher backpressure allowance for sizing of the HP flare system, thereby aiding the selection of a slender HP flare header and a compact sonic flare tip.
  • Temperature is also an important criterion for flare segregation. Combining wet sources with LT sources must be evaluated carefully during design, as there is a chance of freezing or ice formation within the flare header. Temperature also has a bearing on the metallurgy selection for the flare system. The lowest temperature encountered from depressuring sources, normalized after ambient heat exchange, determines the design temperature. This is addressed by selecting compatible material of construction or providing for increased relief temperature before release into the main flare header.
  • Composition of relief gases affects the flare system selection in many ways, metallurgy being the primary one. The most common example is an acid gas or sour flare, which caters to hydrogen sulfide (H2S)-rich, wet relief sources; therefore, the metallurgy must be designed for wet sour service. The acid gas flare header is steam traced to prevent condensation and corrosion. In world-scale refineries, the acid gas flare is segregated from non-sour flare systems.
  • Composition is also significant in the selection of flare type. An aromatic or acid gas flare is always elevated. An elevated flare allows for atmospheric dispersion of toxic relief, preserving personnel safety in case of a flare burnout.

The segregated relief philosophy is shared with licensors to ensure that the project-specific relief summary is developed accordingly. The licensor basic engineering package does not include all relief cases for individual PSVs and will typically not have details of PSV tailpipes, headers/sub-headers, and KODs and pumps in the ISBL. Also, the loads provided are unmitigated relief loads. Common mode failure scenarios such as cooling water failure, general power failure, etc. are compiled at this stage, and OSBL flare header sizing and flare system equipment preliminary design are completed.

Front end engineering design

Flare system development in the FEED phase has two major components. In ISBL units, the individual relief loads are analyzed and scenarios not considered by the licensor (e.g., fire case, control valve failure, etc.) are estimated. The ISBL flare network, KODs and pumps are nearly finalized, with a few remaining holds for confirmation after future design development.

Governing case scenarios for OSBL flare network and equipment are now developed from large individual relief loads and common mode failure scenarios, such as general power failure, cooling water failure, etc. Unmitigated governing caseloads are analyzed using flare system analyzer software tools to identify hydraulic constraints in headers, KODs and flare risers/tips, based on practicable sizes and radiation requirements with regard to the selected height of the riser.

Sterile zone requirements are estimated and checked against the plot space considered. Flare risers more than 200 m in height are not considered practicable; at many locations, regulations limit to lower heights. Possible flare load mitigation measures are studied if the unmitigated loads lead to impracticable designs. Typically, at the end of this phase, a Class 3 (+/– 10%) estimate is developed. Several strategies for mitigation can be considered, as explained in the following subsections.

Steam turbine drives for selected pumps/compressors. In this approach, steam system failure is delinked from general power failure, and some of the critical drives are driven by steam turbine. Examples of drives considered for steam turbine are boiler feedwater (BFW) pumps, boiler forced-draft (FD)/induced-draft (ID) fans, condensate pumps, demineralized water pumps, cooling water pumps, air compressors, reflux pumps, etc.

Dynamic simulation. Transient behavior of the flare system depends on process dynamics during the emergency scenario being analyzed. A common application is to use this tool to utilize credits due to line packing (time taken to pressurize the flare system, including inactive parts) and potential for staggered blowdowns. Theoretically, credit for equipment and control system response can also be estimated in specific cases, but is generally avoided in grassroots flare design.

High integrity pressure protection system (HIPPS). This approach identifies the largest common-mode relief loads from process units and then mitigates through HIPPS. The outcome of a typical flare load mitigation study utilizing HIPPS is presented as the reference in TABLE 2. This approach is generally undertaken in consultation with the owner, and it is common to consider failure of the largest HIPPS when accepting mitigated loads.

Segregated power islands. To achieve significant savings during the design of the power failure case, the power supply to the facility can be distributed at source, such that power failure does not occur in all sections simultaneously. This concept can be implemented by creating two segregated electrical islands for the various process and utility units to accomplish flare load reduction during a general power failure (FIG. 5). The key principles involved in this approach are:

FIG. 5. Flow diagram of power islands concept.
  • Both power islands will not fail simultaneously
  • The failure of one island will cause only part of the refinery to fail, and the design relief load is the only load associated with that part of the refinery
  • The plant associated with the operating (non-tripped) power island can be shut down safely with no additional flare load.

In power islanding philosophy, the refinery will not continue to operate on partial load. The intention is to facilitate safe shutdown with a low flare load.

To implement the philosophy, the double bus in the main intake substation can be designed and configured to function as two separate islands controlled by appropriate interlocks and permissive/supervision systems. Each island can be provided with the option to feed from the grid as backup, while the normal operation is sustained by 100% captive power. If the grid is to be connected in case of a generator failure, it is possible to connect the grid to only one island to prevent interconnecting the bus and avoid the possibility of common failure.

Process units in a refinery are allocated to achieve an approximate 50:50 split between the islands with respect to flare loads and power consumption, considering the shutdown grouping. For a specific case study of naphtha and diesel hydrotreaters, delayed cokers and auxiliary process units (i.e., amine regeneration unit, sour water stripper, hydrogen generation unit), the units were configured in two trains to allow flexibility of power allocation between the two power islands. This split electrical system resulted in the electrical loads at a 47:53 ratio between Island 1 and Island 2, respectively. Similarly, the flare loads were distributed at a 53:47 ratio between Island 1 and Island 2, respectively. The flare system was designed for the higher load of the two.

Distribution of the auxiliary process units, utilities and offsites between the islands were undertaken to ensure that during a general power failure at a single island, both of the additional flare loads arising from the impact on the operation of the running island are minimal. This was ensured by re-energization of the buses supporting the critical utility loads, either from the running island or from the emergency power, as appropriate via a power management system.

Engineering, procurement and construction. As progress is made on a project, the design focus turns to re-verifying individual loads based on the final layout and vendor inputs. Significant increases in loads are not expected when FEED is finished. Mitigation measures are applied and verified throughout the EPC phase to stay within the FEED design limits. Contractors and owners often rely on the use of dynamic simulation tools to estimate reduced individual PSV loads for design optimization. A typical example is column relief load reductions.


Flare system design for refinery mega-projects requires a stepwise approach that is aligned to the project execution plan and matches the data requirements at each stage. As industry has evolved, the capacity and complexity of projects has grown, and many tools and practices have been applied to optimize the design and overcome some of the constraints. The correct approach is to tune into the specific project requirements and focus on making timely decisions to avoid significant late changes that can lead to sub-optimal decisions. HP

The Authors

From the Archive



{{ error }}
{{ comment.comment.Name }} • {{ comment.timeAgo }}
{{ comment.comment.Text }}