Summary
The knowledge in real time of the concentration fields resulting from the accidental release of a hazardous substance would be extremely valuable information as support for emergency actions and for impact evaluation on the industrial site itself and its vicinity. For that purpose, a modeling platform is being developed and applied to simulate in real time the atmospheric dispersion of a hazardous substance at the scale of the industrial site and also of its surroundings. The industrial site of Lacq (France) has been chosen as a pilot, and the key hazardous substance considered in this study is hydrogen sulfide (H2S). A 3D computational-fluid-dynamic (CFD) model (Fluidyn-Panepr) has been chosen to simulate the 3D wind-field pattern on the industrial site, taking into account the details of the installations. This approach enables a simulation as close as possible of the turbulence and flow around the buildings that could not be achieved with a standard Gaussian approach. For that purpose, a detailed numerical model of the Lacq installation was built on the basis of a thorough review of the existing installations and an evaluation of their size and "porosity." Wind fields were calculated for a set of predefined boundary conditions based on the climatology of the site. Investigations were carried out to ensure that site information systems could deliver the information available from the H2S sensors and on-site meteorological station in real time. The real-time approach is made possible by the use of a complete wind-field precalculated database automatically selected in case of accidental release by comparison with real-time wind-direction and -speed measurements from the meteorological station located on the industrial site. The location and intensity of the source term are determined using a probabilistic approach (Bayesian inference), making use of both real-time measurements and precalculated concentration responses from unitary emissions (puffs) on sensors. This approach was validated successfully using a limited number of sensors and sources but with the complex structure and flow patterns expected on the site. The activation of the simulation platform is triggered by the detection of threshold concentrations at the sensors. The estimated source term is then used in forward dispersion mode to simulate the dispersion in (fast) Lagrangian puff mode. The modeling platform will be validated through measurement campaigns with a neutral species in 2010.
