Different　wind　farms　close　in　geographical　location　possess　strong　correlativity　among　wind　speeds　and　it　influences　the　analysis　and　evaluation　results　of　probabilistic　load　flow　in　power　grid　integrated　with　wind　farms.　Firstly,　based　on　historical　data　of　wind　speed　of　multi　wind　farms　a　discretized　joint　probability　model　of　wind　speed,　in　which　the　correlativity　among　different　wind　farms　is　taken　into　account,　is　established;　on　this　basis,　combining　with　semi　invariant　and　series　expansion　method　and　considering　the　correlativity　among　different　wind　farms　a　model　to　analyze　and　evaluate　probabilistic　load　flow　in　power　grid　integrated　with　wind　farms　is　proposed　to　obtain　such　indices　as　the　distribution　of　nodal　voltages　and　the　out-of-limit　probability　of　branch　power　flow,　meanwhile　in　the　proposed　model　the　correlativity　among　active　and　reactive　outputs　of　wind　farms　can　be　considered.　Taking　a　certain　power　grid　as　the　case,　the　verification　results　show　that　the　evaluation　results　of　probabilistic　load　flow　by　the　proposed　method　are　much　more　accurate　than　the　evaluation　results,　in　which　the　correlativity　among　wind　farms　are　not　taken　into　account;　comparing　with　the　results　by　Monte　Carlo　method,　there　is　smaller　computation　error　in　the　obtained　results　while　the　computation　speed　is　increased　greatly.　Using　the　proposed　algorithm,　the　evaluation　indices　of　probabilistic　load　flow　in　power　grid　integrated　with　wind　farms　can　be　obtained　rapidly　and　effectively,　and　the　reactive　compensation　scheme　corresponding　to　the　obtained　out-of-limit　probability　can　be　put　forward.