Aiming at the problem of noise reduction of gas leakage acoustic signals, a method combining ensemble Wavelet Packet Analysis(WPA) and Variational Mode Decomposition(VMD) was proposed to de-noise the collected gas leakage acoustic signals. Firstly, the wavelet packet transform is employed to preprocess the noise of the signal. Then, the de-noised signal was decomposed by VMD to obtain all the Intrinsic Mode Function(IMF) components, and the effective IMF was judged according to the correlation coefficient criterion. Finally, the active components were extracted and the signal was reconstructed. The experimental results show that the above method can effectively eliminate all kinds of noises contained in the gas leakage signal. After noise reduction, the SNR is 15.485 1, and the root mean square error is 0.028, which reduces the interference for the subsequent signal analysis. The above method provides a new idea for the feature extraction and analysis of gas leakage acoustic signal.