Large-scale%20wind%20power%20integration%20brings%20great%20challenges to the%20safe%20operation%20of%20power%20grid,%20and%20wind%20power%20forecasting%20is%20one%20of%20the%20main%20solutions.%20Considering%20the%20non-linearity%20and%20non-stationarity%20of%20wind%20power%20signals,%20a%20combined%20forecasting%20method%20of%20wind%20power%20signals%20based%20on%20improved%20empirical%20mode%20decomposition%20(IEMD)%20and%20support%20vector%20machine%20(SVM)%20is%20proposed.%20Firstly,%20the%20basic%20principle%20and%20advantages%20and%20disadvantages%20of%20EMD%20are%20described.%20Aiming%20at%20the%20problem%20of%20sampling%20rate%20when%20EMD%20decomposes%20non-linear%20and%20non-stationary%20signals,%20an%20improved%20method to reduce%20under%20impulse%20is%20proposed.%20Secondly,%20taking%20wind%20power%20data%20provided%20by%20a%20wind%20farm%20in%20Liaoning%20Province%20as%20an%20example,%20wind%20power%20signals%20are%20decomposed%20into%20a%20set%20of%20relatively%20stable%20subsequence%20components%20using%20IEMD.%20Then,%20EMD-SVM%20and%20IEMD-SVM%20combined%20forecasting%20models%20of%20low-frequency%20and%20intermediate-frequency%20components%20of%20wind%20power%20signals%20are%20constructed%20respectively%20by%20using%20SVM%20theory,%20and%20the%20forecasting%20results%20of%20the%20two%20models%20are%20simulated%20and%20compared%20in%20MATLAB.%20The%20results%20show%20that%20IEMD-SVM%20combined%20forecasting%20model%20can%20effectively%20reduce%20the%20number%20of%20undershoots%20when%20decomposing%20wind%20power%20signals,%20and%20better%20represent%20the%20overall%20trend%20of%20the%20original%20signal.%20Compared%20with%20EMD-SVM%20forecasting%20method,%20it%20has%20higher%20accuracy%20and%20accuracy.