Optimal　scheduling　of　the　power　generation　system　is　of　great　importance　to　the　security　and　reliability　of　the　power　grid.　Microgrid　can　reduce　operating　costs　through　reasonable　economic　dispatch.　At　present,　related　research　mainly　uses　the　power　balance　constraint　based　on　the　discrete-time　model,　which　couldn＇t　exactly　model　the　integral　relationship　between　electrical　energy　and　power　output　rate　of　generators.　A　scheduling　formulation　is　established　with　consideration　of　integral　constraints　and　nonanticipativity　in　this　paper.　With　detailed　analyzing　the　original　model,　a　three-stage　numerical　solution　method　is　proposed.　By　solving　convex　programming,　in　most　instances,　the　global　optimal　solution　can　be　obtained　efficiently.　The　optimal　schedule　in　the　sense　of　expectation　can　be　obtained　with　the　real-time　load　demand　satisfied　and　nonanticipativity　guaranteed.　Numerical　testing　for　autonomous　power　plants　of　a　big　steel　enterprise　is　performed　and　the　effectiveness　of　the　method　is　verified.　©　2020　Technical　Committee　on　Control　Theory,　Chinese　Association　of　Automation.