For　the　difference　and　uncertainty　between　each　batch　of　fermentation　process,　and　currently　the　established　models　based　on　SVM　are　pre　or　off-line　model,　so　once　production　conditions　change,　existing　models　may　not　be　able　to　adapt　to　the　new　environment　inevitably.　And　generalization　capability　of　the　model　based　on　global　learning　support　vector　machine　is　not　strong,　so　according　to　local　learning　theory　the　method　of　establishing　the　fermentation　process　dynamic　model　is　proposed　in　this　paper.　The　dynamic　of　the　fermentation　process　model　is　realized　through　establishing　the　fermentation　process　dynamic　sample　sets.　Taking　the　process　of　Escherichia　coli　producing　interleukin-2　for　example,　experimental　results　verify　that　the　method　can　establish　a　more　accurate　prediction　model　in　the　case　of　a　smaller　number　of　samples.　Compared　with　the　static　SVM　model,　the　dynamic　model　has　a　higher　accuracy　and　a　better　dynamic　adaptability.