The acoustic signals generated during the operation of high-voltage shunt reactors are a critical basis for accurately determining the reactor's operational status. However, collecting these acoustic signals on-site can be subject to interference from various external noises. The measuring instruments often fail to effectively pre-process these signals, resulting in an inaccurate assessment of the reactor's operating condition. This paper presents an enhanced ensemble empirical mode decomposition (EEMD) acoustic signal denoising approach for high-voltage shunt reactors, one which relies on multi-sensor data fusion and the selection of a minimum lower frequency limit for termination. Initially, a consistent data fusion algorithm is used to correlate and filter the fault sensors, discarding any invalid sensors and determining the active sensor group. Subsequently, the minimum lower limit frequency for each sensor signal is chosen as the screening termination criterion for the intrinsic mode function (IMF) through spectral analysis, and the EEMD decomposition is conducted. The correlation coefficient method is then employed to extract the effective IMF components. Finally, the extracted IMF components are superimposed and reconstructed to obtain the denoised signal. Experimental and measured signals demonstrate that the method can achieve signal denoising accurately. A comparison with the traditional empirical mode decomposition (EMD) method and the standard EEMD denoising technique verifies the practical application effectiveness and practicability of the proposed algorithm.