Apr-2008

Applying relief load calculation methods

Case study compares dynamic simulation vs conventional unbalanced heat method for relief load calculations in a naphtha hydrotreater, focusing on the strengths and weaknesses of each method and providing clarity where dynamic simulation should be pursued

Allan Arbo, Irving Oil Limited
Deon Van Der Merwe, Michelle Danielson and Mohamed Abouelhassan
Fluor Canada

Article Summary

Relief load calculations are re-evaluated when a process unit is debottlenecked or running above nameplate. However, traditional relief load calculations may be overly conservative. With the development of dynamic simulation, process engineers have a new tool with which to predict relief 
load calculations.

Given the current shortfall of refinery capacity in areas such as North America, refineries are being called upon to debottleneck and increase throughputs beyond initial design nameplates. As units are debottlenecked and capacity increases, relief loads are recalculated to ensure protection of the unit during relief scenarios. With revamped units, adding relief valves or modifying relief systems can be costly and construction-wise difficult and risky for the site. Ensuring an accurate calculation of the relief load is essential for both equipment protection and minimising unnecessary rework. API 521 states that “conventional methods for calculating relief loads are generally conservative and can lead to overly sized relief- and flare-system designs. Dynamic simulation provides an alternative method to better define the relief load and improves the understanding of what happens 
during relief.”¹

Dynamic simulation provides a rigorous method for calculating the relief load, as it takes into account system volumes and the changes in compositions, conditions and properties of fluids over time during the relief scenario. For new units, relief loads using conventional methods are generally accepted as the norm, while dynamic simulation is commonly used only when conventional methods result in very large relief loads and require correspond-ingly large header and flare systems.

Selecting proper relief load calculation method
Two methods of relief load calculation for various existing columns are discussed and compared in this article; namely, the unbalanced heat method and the dynamic simulation method. The intent is to identify the strengths and weaknesses of each method, 
and to provide some guidance in selecting the proper method for their specific systems. Relief loads are predicted using the two methods for the following existing columns at Irving Oil’s Saint John, New Brunswick, Canada, refinery: naphtha splitter, ethane stripper and debutaniser.

The three columns are in an existing naphtha hydrotreater (NHT), where there is desire to increase the 
overall nameplate capacity of the NHT unit and check if the existing relief valves are adequate for the revised nameplate.

Naphtha splitter process
The naphtha splitter separates the hydrotreated naphtha into light and heavy fractions. Heavy naphtha feeds the catalytic reformers, while light naphtha (C5s and C6s) is directed to the isomerisation unit after being deethanised and debutanised. The splitter operates at 100 psig with a PSV set pressure of 138 psig. The feed comes in the middle of the column. The reboiler is a forced flow-fired heater with a steam-driven pump, and the condenser is an air-cooled heat exchanger with a steam and spare motor-driven reflux/product pump (Figure 1).

Naphtha splitter relief scenario The total power failure case was identified as the governing case for the relief valve sizing, as detailed in Figure 1.

Naphtha splitter relief load The relief load for the naphtha splitter column was calculated using both the conventional unbalanced heat method and a detailed dynamic model of the column. The conventional relief load calculation method or unbalanced heat method assumes a relief load equal to the unbalanced heat input to the system divided by the latent heat of vapourisation of the reflux liquid (or, alternatively, top tray liquid). This approach simplifies the calculations of the relief load and is designed to give conservative results. The latent heat of vapourisation was calculated using a conservative method (Method A, as described in the latent heat discussion section that follows) and the value reported by the steady-state simulation (Method B, as described in the following latent heat discussion).

The other approach to conventional methods recognised by API 521 is dynamic simulation. Building a dynamic simulation model of the system allows us to enter the equipment’s physical dimensions and elevations, then predict the system’s behaviour over time using the identified relief scenario. Dynamic simulation is one extra step closer to the real plant, as it simulates the behaviour of equipment based on sizes, hydraulics and control system responses.

Figure 2 shows the results using the previously described methods as a function of time. Note that the conventional method is not time dependent, so the relief load appears as a constant relief load over time, while dynamic simulation shows the transient behaviour of the system.

Figure 2 shows the required relief valve area calculated using:
- Latent heat of vapourisation of reflux liquid composition at dew-point and relieving pressure
- Latent heat of vapourisation of reflux liquid composition, as reported by steady-state simulation at relieving pressure
- Dynamic model of the column.

Note that using the latent heat of vapourisation as calculated by the steady-state simulation is normally considered incorrect, but is shown only as a comparison for latent heat that includes sensible heat.

There is significant variation between the three methods. In this specific case, the wide composition range within the column resulted in a conventional method area that is twice that predicted by dynamic simulation. The steady-state simulation latent heat resulted in a relief area that is closer to the dynamic simulation results. However, the area is too small, and using these results would not adequately protect the equipment and may yield an under-designed 
flare system.

The following three sections show the variations in relief load predictions using the same dynamic model while varying some process parameters. Notice that the maximum relief load is consistently the same. A good part of this consistency is attributed to the flat relief load (Figure 2) between 12 and 18 minutes. Possible system variations (as explored in the following discussion) change the system hold-up and shift the relief load only slightly compared to the overall relieved quantity (the area under the relief load curve). Other systems that have a single peak in the relief load will be affected differently by these same process variations.

Effect of overhead drum initial level
Initial level in the overhead drum is a key parameter. In conventional methods, it is common practice to assume the overhead drum and condenser will flood as a result of losing the reflux and product pumps (unless there is significant hold-up volume allowing adequate operator intervention time in the overhead drum). In dynamic simulation, the pump failure scenario can be simulated and credit taken for the time it takes the overhead drum to flood. This time is a function of the level in the drum at the beginning of the relief scenario. Figure 3 shows the results with using initial levels of 75%, 50% (base case) and 25% by volume. In the 75% and 50% cases, the drum flooded before the relief valves started opening. The two trends are very similar and only time to flood was shifted. In the 25% case, the column starts relieving before the drum is flooded, and the cooling effect is noticed as the relief load starts to level around 350 000 lb/hr. When the drum actually floods, the relief load starts increasing again to the 
500 000 lb/hr range, following a similar trend to the other two cases.

Effect of off-gas valve 
initial closure
Loss of cooling is not solely a function of overhead drum flooding, and closure of the off-gas valve can have a similar effect as non-condensables build up in the overhead drum and reduce the overall flow through the cooler, 
and hence stop the condenser duty altogether.