Leakages from subsea oil and gas equipment cause substantial economic losses and damage to marine ecosystem, so it is essential to locate the source of the leak. However, due to the complexity and variability of the marine environment, the signals collected by hydrophone contain a variety of noises, which makes it challenging to extract useful signals for localization. To solve this problem, a hydrophone denoising algorithm is proposed based on variational modal decomposition (VMD) with grey wolf optimization. First, the average envelope entropy is used as the fitness function of the grey wolf optimizer to find the optimal solution for the parameters K and α. Afterward, the VMD algorithm decomposes the original signal parameters to obtain the intrinsic mode functions (IMFs). Subsequently, the number of interrelationships between each IMF and the original signal was calculated, the threshold value was set, and the noise signal was removed to calculate the time difference using the valid signal obtained by reconstruction. Finally, the arrival time difference is used to locate the origin of the leak. The localization accuracy of the method in finding leaks is investigated experimentally by constructing a simulated leak test rig, and the effectiveness and feasibility of the method are verified. © Chinese Ocean Engineering Society and Springer-Verlag GmbH Germany, part of Springer Nature 2024.