With　the　rapid　development　of　electric　vehicles　(EVs),　the　consequent　charging　demand　represents　a　significant　new　load　on　the　power　grids.　The　huge　number　of　high-rise　buildings　in　big　cities　and　modern　technological　advances　have　created　conditions　to　mount　on-site　wind　power　generators　on　the　buildings.　Since　modern　buildings　are　usually　equipped　with　large　parking　lots　for　EVs,　it　shows　vital　practical　significance　to　utilize　the　on-site　wind　power　generation　to　charge　EVs　parked　in　the　buildings.　In　this　paper,　we　first　use　a　case　study　in　Beijing　to　show　that　the　on-site　wind　power　generation　of　high-rise　buildings　can　potentially　support　all　the　EVs　in　the　city.　Considering　that　the　charging　demand　of　EVs　usually　does　not　align　with　the　uncertain　wind　power,　the　coordination　of　EV　charging　with　the　locally　generated　wind　power　in　a　microgrid　of　buildings　is　investigated　and　three　main　contributions　are　made.　First,　we　investigate　the　problem　and　formulate　it　as　a　Markov　decision　process,　which　incorporates　the　random　driving　requirements　of　EVs　among　the　buildings.　Second,　we　develop　a　distributed　simulation-based　policy　improvement　(DSBPI)　method,　which　can　improve　from　heuristic　and　experience-based　policies.　Third,　the　performance　of　the　distributed　policy　improvement　method　is　proved.　We　compare　DSBPI　with　a　central　version　method　on　two　case　studies.　The　DSBPI　method　demonstrates　good　performance　and　scalability.　©　2010-2012　IEEE.