An improved method to identify and locate the capacitor switching disturbances is presented. Other methods have drawbacks of lack of confidence when capacitors switch on at low voltage. The wavelet transform coefficient of the special layer is then adopted to enlarge useful signals and mitigate disturbing ones. The entropy eigenvalue extracted from coefficients can efficiently identify the capacitor switching disturbances at low voltage. An improved power and energy based method is introduced to locate capacitor switching disturbances. Simulations indicate that the method proposed can distinguish and locate the capacitor switching disturbances at low voltage with reliable results.