Vapour Cloud Explosion Modeling – Setting the Right Limits for Oil and Gas Risks
Setting the right limit of liability (LOL) is one of the most important and complex decisions in designing an insurance program for large refineries and petrochemical plants. Buying up to the full Total Insurable Value (TIV) on a multi-billion dollar plant is often impractical due to lack of capacity – and unnecessary when there is no chance of total loss.
One might think the Estimated Maximum Loss (EML), or the largest potential loss of a risk, is the best way to determine the limit of liability. If an EML truly represents the worst case loss amount, setting the LOL at the same level as the EML seems reasonable – until you look a little further. The lack of a standard methodology of EML calculation means the results swing widely, depending on the modeling software as well as the people inputting the data.
In the world of oil and gas, EML is usually represented by Vapor Cloud Explosion (VCE). A VCE occurs when the leakage of a large volume of flammable gas or vapor mixes with air and finds its ignition source in a confined or congested area. Generally, the VCE EML calculation process involves modeling different explosion scenarios with defined parameters and assumptions, such as hold-up and released flammable gas or liquid, temperature, pressure, release period, congestion, etc.
Two commonly used softwares for VCE modeling are ExTool and SLAM. ExTool is based on the TNT equivalency method, which assumes a vapor cloud explosion is similar to a detonation of TNT. But VCE is a deflagration, not a detonation. SLAM, on the other hand, uses the Congestion Assessment Method (CAM), which calculates the maximum overpressure generated by a deflagration based on the extent of congestion. This method is limited to deflagration events.
As is the case with any modeling software, the modeled results are heavily influenced by the choice parameters, input data and assumptions made by the engineers running the tools, as well as the accuracy of the tools themselves. So how consistent could the results be?
In an interesting VCE modeling case study, researchers reviewed the VCE EML modeling results of a Swedish refinery that were conducted by two different insurance brokers using ExTool and SLAM. In one of the five different loss scenarios, the maximum property damages were estimated at SEK 6,430 million and SEK 2,390 million, respectively. However, another estimate based on the same scenario showed PD loss estimates of SEK 4,100 million and SEK 1,470 million, an almost 270% difference.
I took a different track and gathered the EMLs of 26 refineries worldwide, using each plant's risk engineering survey and divided Property Damage (PD) EML by TIV or Replacement Cost (RC) to determine how wide the PD EML swings were in proportion to its TIV. The highest EML against its TIV was 56% and the lowest was 6%. Considering the large EML/TIV spread, although every refinery is different in layout, spacing and asset concentration, these swings show it is worth double-checking the validity of EMLs when they are presented. While the point of an EML is to estimate the worst-case scenario, it is by no means a guarantee, and modeled results always need to be supplemented by further judgment.
In Gen Re, we have a team of experienced underwriters dedicated to the Oil and Gas industry. As a non-proportional underwriter who shares in the risk with you, we can help remove the unexpected volatility arising from getting the EML wrong. We are happy to share our experience or provide a second-opinion on VCE EML of your risk.
- “Evaluating EML Modeling Tools for Insurance Purposes: A Case Study.” International Journal of Chemical Engineering, 2010.