The cable elbow connector is an important component of power systems. Due to its long-term exposure to low-lying and humid environments, it is prone to water accumulation, resulting in reduced insulation performance and even failures. Traditional water accumulation detection methods rely on manual inspection and complex equipment, which are costly and lack a well-developed automated detection process. Acoustic fingerprint recognition technology based on tapping sound signals offers advantages such as simplicity, low cost, and environmental friendliness; however, its discriminability under different water accumulation conditions and the underlying acoustic mechanisms have not been systematically studied. To address this issue, this paper establishes an acoustic–structural coupled model of a cable elbow connector based on the finite element method to simulate the propagation characteristics of tapping sounds under various water accumulation conditions. Through time-domain feature extraction and continuous wavelet transform analysis, the study reveals the effects of water accumulation on sound wave propagation, including signal amplitude attenuation, response delay, frequency shift, and energy dispersion. The results indicate that acoustic modeling can effectively characterize the acoustic fingerprint differences under different water accumulation conditions, providing theoretical foundations and technical support for developing water accumulation state recognition algorithms for cable terminations based on acoustic fingerprint features.