The　optimization　of　electric　field　distributions　in　the　optical　voltage　sensor　(OVS)　are　mostly　based　on　the　finite　element　calculations　and　the　exhaustive　search　method.　It　is　unable　to　map　complicated　and　nonlinear　relationships　between　the　electric　field　distribution　and　the　sensing　structure.　Moreover,　these　methods　turn　out　to　be　time-consuming,　inefficient,　and　easy　to　fall　into　local　traps.　A　hybrid　optimization　algorithm　based　on　particle　swarm　optimization　(PSO)　and　support　vector　machine　(SVM)　is　developed　to　fix　these　issues　in　which　the　PSO-SVM　establishes　an　electric　field　model　for　the　electro-optic　crystal　and　then　the　model　is　modulated　by　the　PSO　to　achieve　optimization.　The　medium　attaching　structure　is　taken　as　an　example　to　test　and　verify　the　job,　and　the　simulations　and　experiments　show　that　the　uniformity　of　electric　field　distributions　is　improved　by　58%　and　the　training　time　is　saved　by　87%.