An XLPE (cross linked polyethylene) cable partial discharge denoising method based on a chaotic annealing wavelet hard threshold method comprises steps as follows: S1, inputting a partial discharge signal of an XLPE cable; S2, selecting a db5 wavelet base function, decomposing a hierarchy J into five hierarchies, and performing wavelet decomposition to acquire a low-frequency coefficient C (j,n) and a high-frequency coefficient W (j,n); S3, using a generalized cross validation criterion as an assessment condition of threshold selection; S4, performing global coarse searching with a chaotic optimization method, and using a Logistic map as a chaotic optimization sequence; S5, performing fine searching by using a simulated annealing principle, and using a progressive contraction disturbance method for effectively promoting algorithm convergence; S6, selecting the optimum hard threshold according to a chaotic annealing algorithm, performing hard threshold processing on a wavelet coefficient to obtain a processed wavelet coefficient W (j,n), and performing wavelet reconstruction; S7, assessing the denoising effect, and using an SNR (signal to noise ratio) and an RMSE (root mean square error) after denoising as standards; S8, outputting a denoised partial discharge signal of the XLPE cable. Intelligent denoising can be realized with the method.