Abstract: aiming　to 　solve 　satellite 　clock 　bias scbthat 　is 　difficult 　to 　formulate 　for 　predictionthispaper 　emplo ys　a　genetic　al gorithm　to　optimize 　the　network　parameters 　of 　a　wavelet　neural 　networksuitable 　for 　prediction　nonlinear 　time 　series．firstagenetic　wavelet　neural 　network gwnnwithbetter　prediction　performance　was　obtained．secondl ythe　scb　values　were　pre processed　based　ontheir 　characteristics 　and　a　gwnn　model　was　established 　to 　accuratel y　predict 　scb　in　near 　real 　time．finall ythe　paper 　carries 　out　scb　prediction 　tests　within　one　day　using　the　scb　data 　from　gps　satellites 　to　validiate 　the　proposed　model．the　results 　show　that 　the　proposed　prediction　model　outper－forms 　igs　ultra－ra pid 　predicted　scb　products