According　the　problem　of　difficult　to　measure　online　water　quality　parameters　of　activated　sludge　process　wastewater　treatment　system,　this　paper　proposed　a　new　growth　Self-Organization　Neural　Network　This　network　can　dynamic　generate　network　nodes　and　grow　to　suitable　network　structure　rapidly　according　to　need　in　the　learning　process　no　need　to　advance　set　the　value　the　structure　and　scale.　The　water　quality　parameters　model　of　wastewater　treatment　system　based　on　this　network,　have　more　strong　adaptive　ability,　can　learning　online,　network　structure　is　simple,　learning　velocity　rapid,　prediction　effluent　water　COD　concentration　effectively　according　to　input,　which　proved　high　effectiveness　of　this　method.