Short-term　generation　scheduling　for　autonomous　power　plants　(APP)　in　energy　intensive　enterprises　(EIE)　is　a　typical　problem　in　production　scheduling.　The　problem　is　generally　formulated　as　a　stochastic　one　since　the　load　demand　is　uncertain.　The　full　distribution　information　of　the　load　demand　is　usually　required　in　previous　literature　but　it　is　very　hard　to　be　obtained　in　practical　applications.　In　this　paper,　a　short-term　generation　scheduling　method　which　requires　only　limited　distribution　information　of　the　load　demand　is　presented　for　APP　in　EIE　to　get　the　minimum　expected　total　cost　of　electricity　consumption.　Since　the　probability　density　function　(PDF)　is　unavailable,　the　objective　function　corresponding　to　the　expected　cost　is　unknown.　Due　to　this,　Polynomial　interpolation　is　adopted　to　get　a　good　approximation　of　the　objective　utilizing　the　prediction　mean　and　interval　of　the　load　demand.　Numerical　tests　are　performed　with　the　actual　data　of　a　large　iron　and　steel　enterprise.　The　results　obtained　by　our　approach　based　on　limited　distribution　information　are　quite　closer　to　the　real　optimal　ones,　which　demonstrate　the　proposed　approach　is　practicable　and　effective.　The　method　can　be　extended　to　solve　other　related　generation　scheduling　problems　with　unknown　demands.