Traditional wavelet thresholding methods require prior knowledge of either the characteristics of the clean signal or the noise sequence to determine the wavelet basis and decomposition level, making them less suitable for real-time acquisition of unknown and noisy pipeline leakage signals. To address this issue, an adaptive wavelet thresholding algorithm is proposed to determine the wavelet basis and decomposition level automatically. Firstly, a wavelet basis library is established, and the noisy signal is decomposed at the first level using different wavelet bases. The wavelet basis with the maximum ratio of energy entropy to energy of the approximation coefficients is selected. Secondly, the signal is subjected to multi-level wavelet decomposition, and the energy of each detail coefficient is calculated. The optimal decomposition level is determined by identifying the first level where the energy decay criterion is violated. Finally, the detail coefficients are thresholded and the signal is reconstructed, resulting in a denoised signal. Experimental simulations demonstrate that the proposed algorithm outperforms other methods in terms of denoising performance metrics such as signal-to-noise ratio (SNR) and root mean square error (RMSE). It exhibits favorable denoising effects on real leakage acoustic and negative pressure wave signals, thereby possessing significant engineering value.