Abstract Accurate identification of leakage pressure signal is crucial for leakage detection and location of water supply pipelines. However, there is a lot of noise signals in the negative pressure waves generated during leakage accidents, making it difficult to identify leakage signals, and thus seriously affecting the accuracy of leakage detection. Therefore, in this study, we carried out leakage test of water transmission pipeline on our self-built test platform for transient flow with long-distance water transmission pipeline (pipeline length is 160 m, and diameter is 100 mm), and obtained the transient leakage pressure signals under various flow rates (15–55 m3/h). An improved wavelet filtering algorithm was proposed to filter the leakage pressure signal by optimizing the calculation of decomposition layers and the threshold processing strategy, and then compared with the traditional wavelet filtering method. The results show that the improved algorithm can denoise the leakage signals effectively, and the signal-to-noise ratio (SNR), mean square error (MSE) and mean absolute percentage error (MAPE) are all superior to those of the traditional wavelet algorithm. The results can provide support for leakage detection and location of water supply pipeline systems.