Hybrid GA-ANN-LSSVM algorithm for adaptive Brillouin frequency shift extraction in long-distance oil and gas pipeline monitoring

Long-distance oil and gas pipelines require high-precision safety monitoring systems to prevent failures caused by environmental changes and operational strains. Although Brillouin optical time-domain analysis (BOTDA) systems are widely used for distributed sensing, existing algorithms face challenges in handling multi-peak Brillouin gain spectra (BGS) and high signal-to-noise ratio (SNR) requirements. This paper proposes a hybrid intelligent algorithm (GA-ANN-LSSVM) combining a genetic algorithm-optimized artificial neural network (GA-ANN) and least squares support vector machine (LSSVM) to improve BOTDA performance. The algorithm adaptively identifies single/multi-peak BGS and dynamically invokes GA-ANN or LSSVM for Brillouin frequency shift (BFS) extraction. Experimental results demonstrate that the proposed method achieves temperature and strain measurement errors below 1 °C and 2 με, respectively, with 4.4% higher peak recognition accuracy than conventional algorithms. Additionally, the proposed method reduces the initialization SNR threshold by 5 dB and processing time by 3 seconds, significantly enhancing the practicality of BOTDA in complex pipeline environments.