With　the　depletion　of　coal　in　the　world,　coal　quality　fluctuates　and　deviates　greatly　from　the　designed　coal　in　many　large　scale　coal-fired　power　plants.　This　increases　the　coal　consumption　while　reduces　the　boiler　combustion　efficiency　and　stability.　Thus,　it　is　very　important　to　conduct　real-time　measurement　to　the　quality　of　the　coal　for　optimizing　the　operation.　The　calorific　value　analysis　is　a　significant　part　of　the　coal　quality　analysis,　and　regular　proximate　analysis　method　can＇t　meet　real-time　control　requirements.　In　this　paper,　an　artificial　neural　network　(ANN)　model　using　real　plant　data　for　prediction　of　net　calorific　value　of　coal　in　a　China　power　plant　is　reported.　A　3-layer　BP　neural　network　has　been　adopted.　The　input　parameters　selection　was　optimized　with　a　compromise　between　smaller　number　of　parameters　and　higher　level　of　accuracy　through　sensitivity　analysis.　The　activation　function　selection　was　also　discussed　in　details.　The　results　indicate　that　when　the　pureline　was　selected　as　the　activation　function　for　hidden　layer　and　logsig　was　selected　as　the　activation　function　for　output　layer,　the　prediction　is　most　accurate.　The　results　have　shown　good　potential　for　predicting　the　net　calorific　value　of　coal　using　the　real　time　data.　This　information　will　enhance　the　performance　of　the　combustion　control　system　for　power　utilities.　©　Tsinghua　University　Press,　Beijing　and　Springer-Verlag　Berlin　Heidelberg　2012.