To suppress excessive thermoacoustic instabilities in the gas turbine, it must be possible to predict pressure changes in the combustion chamber. The time-series data of acoustic pressure fluctuations in the Rijke type burner under external sound source interference were studied combined via nonlinear theory, and a new data-driven model for predicting internal sound pressure fluctuations under such conditions was established. An improved particle swarm optimization (PSO) algorithm was proposed to optimize the parameters of the support vector regression (SVR) model, and the parameter optimization time required for the improved PSO algorithm is only 3/5 of that before the improvement. The results show that at least 0.94 ms ahead, the improved data-driven model can accurately predict sound pressure oscillation signals. The improved PSO-SVR model proved to be more accurate than the Multilayer Perceptron (MLP) model and Gaussian process regression (GPR) model in predicting the fluctuation of sound pressure under variable conditions and can provide effective guidance for predicting and eliminating the thermoacoustic oscillations in the actual combustion chambers.

为了抑制燃气轮机中过度的热声不稳定性，必须有可能预测燃烧室中的压力变化。利用非线性理论，研究了Rijke型燃烧器在外部声源干扰下声压波动的时间序列数据，建立了一个新的数据驱动模型，用于预测这种情况下的内部声压波动。提出了一种改进的粒子群算法（PSO）对支持向量回归（SVR）模型的参数进行优化，改进的PSO算法所需的参数优化时间仅为改进前的3/5。结果表明，改进的数据驱动模型至少提前0.94 ms可以准确预测声压振荡信号。