In　order　to　highlight　the　impact　of　quality　variables　on　stage　division　and　improve　the　quality　prediction　accuracy,　a　quality　prediction　method　for　multi-stage　batch　processes　based　on　extended　score　matrix　was　proposed.　Original　three-dimensional　data　were　first　unfolded　along　the　batch　direction,　and　score　matrices　representing　process　variables　and　quality　variables　were　obtained　by　PLS　(partial　least　squares)　analysis　of　each　slice　matrix.　The　extended　scoring　matrix　was　obtained　by　combining　the　two　scoring　matrices,　and　the　similarity　of　the　two　adjacent　extended　scoring　matrices　was　calculated　by　CS　(Cauchy-Schwarz)　statistics　to　divide　the　stages.　MPLS　(multiway　PLS)　quality　prediction　models　were　then　established　in　the　transition　stage　and　the　stable　stage.　Finally,　the　effectiveness　and　utility　of　the　proposed　method　were　validated　through　a　fed-batch　penicillin　fermentation　simulation　platform　and　E.　coli　production　of　interleukin-2.　The　results　demonstrate　the　feasibility　and　effectiveness　of　the　proposed　method.　©　2019,　Editorial　Board　of　＇Journal　of　Chemical　Engineering　of　Chinese　Universities＇.　All　right　reserved.