In　view　of　the　bad　operation　environment　of　large　motor,　which　often　suffers　from　various　strong　noise　interference,　the　partial　discharge　ultrasonic　signal　is　often　annihilated,　which　makes　it　difficult　to　detect　and　analyse.　A　de-noising　method　based　on　generalized　minimax　concavity　(GMC)　and　wavelet　for　partial　discharge　(PD)　ultrasonic　signal　is　proposed.　GMC　is　used　to　enhance　the　sparsity　of　PD　ultrasonic　signal　and　eliminate　the　high-frequency　noise　signal　at　the　same　time.　Then　the　residual　high-frequency　sparse　noise　and　low-frequency　noise　of　the　former　are　de-noised　again　combined　with　wavelet.　Finally,　the　signal　is　reconstructed　to　achieve　the　purpose　of　de-noising　the　original　PD　ultrasonic　signal　with　noise.　Compared　with　l(1)　-norm　method,　GMC　method,　wavelet　method　and　l(1)　-norm-wavelet　method,　the　simulation　results　show　that　based　on　time　domain　analysis,　the　de-noising　effect　of　the　proposed　method　is　obviously　better　than　the　other　four　methods.　The　SNR　and　MSE　of　the　former　are　better　than　those　of　the　latter.　In　addition,　the　insulation　bar　discharge　model　of　large　motor　is　constructed　to　obtain　the　actual　PD　ultrasonic　signal,　which　further　verifies　its　effectiveness,　and　its　de-noising　effect　is　also　better　than　the　four　methods.　This　method　can　not　only　enhance　the　sparsity　of　the　target　signal　and　improve　the　estimation　accuracy,　but　also　achieve　the　de-noising　effect,　while　retaining　the　effective　information　of　PD　ultrasonic　signal　characteristics.　This　method　can　provide　new　ideas　for　other　types　of　PD　signal　de-noising,　and　lay　the　foundation　for　later　feature　analysis.