It　is　well　known　that　some　key　effluent　quality　parameters　are　difficult　to　measure　online　in　the　urban　sewage　treatment.　To　solve　this　problem,　this　paper　proposes　a　new　soft-measurement　model　using　empirical　mode　decomposition　and　modular　neural　network　(EMD-SMNN)　for　effluent　quality　parameters.　First,　a　task　decomposition　algorithm　based　on　EMD　is　proposed,　which　can　decompose　a　complex,　multi-frequency　time　series　of　effluent　quality　parameters　into　several　sub-time　series,　and　it　can　adaptively　adjust　subnetwork　modules　according　to　the　complexity　and　similarity　of　sub-time　series　calculating　by　the　sample　entropy　and　Euclidean　distance.　Then,　a　novel　self-organizing　algorithm　of　FNN　is　proposed　to　solve　the　problem　that　the　initiating　structure　of　subnetwork　is　difficult　to　given,　which　can　dynamically　adjust　the　structure　of　subnetworks　and　predict　subtasks　effectively.　Finally,　through　the　benchmark　time　series　prediction　and　the　actual　effluent　water　quality　parameter　detection　in　the　sewage　treatment　plant,　it　is　verified　that　the　proposed　EMD-SMNN　has　a　good　prediction　accuracy　and　self-adaptability.　©　2024　Materials　China.　All　rights　reserved.