The　evaluation　of　water　quality　plays　a　very　important　part　in　water　resources　protection.　Compared　with　other　methods,　Hopfield　neural　network　can　evaluate　the　water　quality　effectively.　Aiming　at　the　problem　of　complicated　structure　of　Hopfield　neural　network,　a　simplified　model　is　proposed.　In　the　simplified　model,　connection　weights　are　designed　by　singular　value　decomposition　to　improve　running　speed.　And　the　simple　structure　is　got　by　deleting　unimportant　weights.　Finally,　the　effectiveness　and　feasibility　of　simplified　model　is　proved　for　water　quality　evaluation.　©　2015　Technical　Committee　on　Control　Theory,　Chinese　Association　of　Automation.