Rapid source term estimation for multiple elevated leak sources using a Fast-AHPSOalgorithm: Application to industrial park emergency response

Authors:
Preprint
Publisher website
Google Scholar
Abstract
Conventional source term estimation algorithms for elevated point sources are limited to single-source scenarios, failing to provide rapid and accurate quantification of multiple simultaneous releases during industrial accidents. Moreover, the classical Gaussian plume model exhibits poor adaptability under dynamic meteorological conditions and multi-source coupling. To overcome these limitations, as well as the premature convergence and long runtime of traditional particle swarm optimization (PSO) and its improved variant AHPSO, this paper proposes a fast source term estimation method based on an improved adaptive heterogeneous particle swarm optimization (Fast-AHPSO) tailored for emergency response. A multi-source coupled Gaussian plume correction model is developed by incorporating wind field functions, superposition mechanisms, and temperature–humidity corrections to enhance dispersion simulation accuracy for accidental releases. The Fast-AHPSO algorithm reduces computational time through less frequent altruistic operations and an early stopping mechanism. Experimental results show that Fast-AHPSO achieves high-precision source strength estimation under both clustered and dispersed source layouts, with average relative errors of 1.29\% and 0.22\%, respectively. Its runtime is approximately one-tenth that of AHPSO, greatly improving efficiency for real-time decision-making. The spatial distribution of sources significantly affects inversion accuracy: dispersed configurations yield smoother concentration fields and smaller gradients, leading to more accurate estimates. This study provides an efficient and reliable tool for rapid source term estimation of multiple elevated releases in industrial zones, with important implications for process safety enhancement and emergency response planning.

This publication has 11 references indexed in Scilit: