The traditional method for locating gas leakage sources usually relies on manual inspection, which has problems such as small detection range, slow response speed, and low detection accuracy. Traditional methods for locating gas leakage sources cannot quickly and accurately pinpoint the location of the gas leakage source, which makes it difficult to address the issue promptly. The fireworks optimization algorithm is suitable for spatial gas source localization. However, the basic fireworks optimization algorithm has problems such as being prone to get stuck in local optima and having a slow convergence speed, which reduces the accuracy and real-time performance of gas source localization. Therefore, in this paper, we propose an inverse-optimization-based gas leakage source localization method using infrared spectroscopy, which combines an improved fireworks optimization algorithm (FUFWA) with a gas turbulence diffusion model. This method has been proven to provide high accuracy and real-time performance in gas source localization.

To improve the accuracy and real-time performance of the fireworks optimization algorithm in gas source positioning, a gravity search operator F and an adaptive coefficient U are introduced based on the basic fireworks algorithm. The gravity search operator F is introduced to apply gravity to each particle, which causes it to move toward the current optimal particle and thereby improves the algorithm’s global search capability to enhance localization accuracy. The adaptive coefficient U is introduced to improve the explosion operator and termination conditions of the fireworks algorithm, thereby enhancing the algorithm’s operational efficiency and real-time positioning performance. FUFWA is combined with the gas turbulence diffusion model to construct an inverse optimization model for gas source localization.

We conduct space gas source positioning simulation experiments using the FUFWA and the basic fireworks algorithm under different wind speeds, wind directions, and monitoring node numbers. We record the positioning errors and single run time of the two algorithms for each experiment. The simulation experiment results show that the proposed algorithm reduces the single run time by 90.94% compared to the basic fireworks algorithm. Under different wind speed conditions, the proposed algorithm reduces the average positioning error by 71.52% compared to the basic fireworks algorithm. Under different wind direction conditions, the proposed algorithm reduces the average positioning error by 57.02% compared to the basic fireworks algorithm. Under different monitoring node conditions, the proposed algorithm reduces the average positioning error by 72.24% compared to the basic fireworks algorithm. To further verify the positioning performance of the algorithm, four sets of carbon dioxide source positioning experiments are conducted on the campus of Jilin University using a carbon dioxide (CO₂) sensor based on non-dispersive infrared spectroscopy technology. The experimental results show that the FUFWA reduces the average positioning error by 81.87%, 84.50%, 85.94%, and 88.31%, respectively, compared to the basic fireworks algorithm. A methane (CH₄) sensor based on off-axis integrated cavity technology is used to conduct a three-dimensional space gas source positioning field experiment at the Agricultural Experimental Base of Jilin University. The positioning error based on the inverse optimization gas source positioning method is about 6.69 m. The results verify that the proposed inverse optimization method for gas source localization has good spatial gas source localization capability.

A gas source position inversion model based on mobile inspection is proposed by combining the FUFWA with a gas turbulence diffusion model. In the traditional fireworks algorithm, the gravity search operator F and adaptive coefficient U are introduced, which adds the influence of inertial mass to particles, thus improving the accuracy of positioning, enhancing the explosion operator and termination strategy, and boosting the efficiency of the algorithm. In terms of wind speed, wind direction, and the number of monitoring nodes, the FUFWA shows smaller errors and better robustness compared to the basic fireworks algorithm. The simulation results show that the FUFWA significantly improves overall positioning accuracy compared to the FWA. Meanwhile, introducing the adaptive coefficient U effectively reduces the single run time of the FUFWA and improves its real-time performance. We conduct an experiment on the localization of CO₂ gas sources in a two-dimensional plane and apply the FUFWA inverse model to estimate its performance. The experimental results show that the FUFWA improves positioning accuracy by 81.87%, 84.50%, 85.94%, and 88.31% compared to the basic fireworks algorithm, respectively. In addition, the three-dimensional spatial positioning ability of the FUFWA inverse model is verified through an experiment on airborne CH₄ gas source positioning in three-dimensional space.