As wavelet de-noising with a traditional threshold function is applied on microgrid detected signals, there are some problems such as the distortion of reconstructed signal and the thickness of waveform curve. Based on analysis of one kind of improved threshold function, a new expression of wavelet threshold function is proposed. To improve the applicability of that, Bayes shrink method is introduced to calculate the threshold of each decomposition layer, and Artificial Bee Colony (ABC) algorithm is used to select parameters of the new threshold function. De-noising effects of different threshold function are verificated with measured noisy signals. It is shown that the new threshold function and its de-noising process can improve the signal-to-noise ratio (SNR) and reduce the mean square error (MSE). And the targets such as noise suppress and waveform characteristics displays of detected signal are achieved.