ERROR IN FERC EIS DETERMINATION OF LNG VAPOR CLOUD EXCLUSION ZONES: FAILURE TO ACCOUNT FOR AIR MIXING IN VAPOR IMPOUNDMENTS

Transcription

1 Obstacles Vapor holdup in dikes ERROR IN FERC EIS DETERMINATION OF LNG VAPOR CLOUD EXCLUSION ZONES: FAILURE TO ACCOUNT FOR AIR MIXING IN VAPOR IMPOUNDMENTS obstacle vapor holdup in fence MKOC Symposium 2006 Texas A&M University Jerry Havens and Tom Spicer Chemical Engineering Department University of Arkansas

2 Presentation Agenda 1. Describe FERC s error in EIS s to determine vapor cloud exclusion zones required by 49 CFR Why is this error important? This paper describes a straightforward example for analysis of the effect of air/lng vapor mixing on exclusion zone determination. The analysis shows that the method approved by FERC is not conservative. 3. Conclusions

3 FERC S S ERROR Vapor cloud dispersion exclusion zones for specified (design) spills for land-based LNG containment and transfer systems are required by 49 CFR 193 and NFPA 59A (incorporated). If there is a question about the provisions of either, 49 CFR 193 prevails. As spills in a facility are typically into concrete impoundments located inside earthen-diked areas, there is a need to consider the effects on LNG vapor dispersion of vapor holdup in the impoundment/dike system. This is where the error has been made.

4 FERC S S ERROR FERC has allowed the DEGADIS model to be used, with input (pure LNG vapor source) from SOURCE5. This allowance is not in accordance with the requirements of 49 CFR 193: DEGADIS is limited to prediction of dispersion over flat, obstacle-free terrain, hence it cannot be used to consider the effects of dikes, obstacles, or terrain. FEM3A is currently the only option available to take into account the effects of dikes, obstacles, and terrain.

5 The question of LNG vapor holdup and the potential for flammable gas mixtures to extend beyond an impoundment/dike system has been studied in wind tunnel and field experiments as well as by using computational fluid dynamic models: FALCON Experiments (1980 s) intro slide CHRC Wind Tunnel Experiments (University of Arkansas) (1980 s s to present) FEM3A simulations by Lawrence Livermore Laboratories (1980 s) and the University of Arkansas (1980 s s to present) All of this experimental work and mathematical modeling contradicts the assumption made by FERC.

6 The SOURCE5 method, allowed by FERC for determining the input (LNG vapor source) for DEGADIS, estimates the rate of LNG evaporation from the spill, coupled with the assumption that LNG vapor fills the impoundment and overflows without any mixing with air or increase in temperature.. This method, which contradicts experimental measure- ments and CFD simulations, is in error: The method can result in the determination that the LNG vapor formed will not overflow the dike because the volume of the pure vapor (at the b.p.. temperature) does not exceed the volume of the largest dike surrounding the spill area, or If it does exceed that volume, it will overflow later and at a greatly diminished rate as a result of heat transfer effects that decrease the evaporation rate in the impoundment with time.

7 A simplified but straightforward argument shows that the method approved by FERC, which is disallowed by 49 CFR 193, is not conservative. We choose an example taken from a recent FEIS spillage of 0.8 m 3 /s LNG, for 10 minutes, into a concrete sump with dimensions 18 m by 18 m and depth 1.5 m. The concrete sump is located inside an earthen dike with dimensions 43 m by 43 m and dike wall height 1.5 m. The LNG tank is not located within the earthen dike.

8 The methodology allowed by FERC to determine the vapor cloud dispersion exclusion zone in the FEIS for this LNG spill scenario was a two step process: SOURCE5 was used to determine the time at which vapor overflows the sump, and, assuming that the vapor remained pure and at the b.p.. temperature, the time and rate at which the vapor subsequently overflows the surrounding dike. The rate at which the (assumed pure) LNG vapor overflows the dike was then input to DEGADIS. This method neglects the mixing of air into the sump as well as the surrounding dike.

9 FERC erred in using the (assumed) pure LNG vapor overflow from the dike as input to DEGADIS because this procedure attempts to consider the effect on dispersion of the dike surrounding the sump without using FEM3A - the only option provided by 49 CFR 193. This error is important because, aside its disallowance by 49 CFR 193, the procedure is not conservative the assumptions made result in a prediction of the vapor cloud exclusion zone considerably less than would result if correct accounting were made for the effect of air mixing with LNG vapor within the dike.

10 We repeated, as closely as we could, the analysis described in the FEIS, and extended it to account for the effects that would result from different degrees of air mixing that could occur with the LNG vapor in the dike volume. Using this method, we determined that the (assumed) pure LNG vapor would begin overflowing the 43 m by 43 m dike 338 seconds (~5.6 minutes) after the 0.8 m 3 /s spill commenced, and that the overflow rate at that time would be 8.3 kg/s.

11 We then calculated the dike overflow times and rates assuming that different amounts of air, specified to be at 0 o C, 1 atmosphere pressure, and 50% R.H., were mixed with the evolving vapor so as to result in volume averaged gas concentrations in the dike at the time of overflow ranging down to about 25% LNG vapor (75% air), the latter having been observed in CFD simulations of similar scenarios. To determine the volume of the gas/air mixtures in the dike at the time of overflow, we used the adiabatic mixing tools prescribed in DEGADIS.

12 We input these vapor cloud overflow rates into DEGADIS and assumed that the ground level source for that gas evolution rate was from the dike, which was represented for DEGADIS as a circular source of 24.3 m radius. The results are summarized below. Avg NG Mole Fraction Overflow Time, s NG Overflow Rate, kg/s Dist to 2.5%, m

13 CONCLUSIONS Calculations of vapor cloud exclusion distances, using DEGADIS, show that the vapor cloud exclusion distance increases monotonically for assumed air mixing down to at least 25% volume averaged LNG vapor gas concentration at the time of dike overflow. Analysis of the effects of air mixing with LNG vapor evolved from spills into impoundment/dike systems which cause holdup of LNG vapor demonstrates that the assumption of no air mixing inside the impoundment is not conservative in the safety sense.