The　traditional　indoor　location　algorithm　based　on　distance-loss　model　mostly　turn　received　signal　strength　indicator　RSSI　into　distance,　and　then　through　the　location-distance　algorithm　to　achieve　positioning.　These　algorithms　need　fit　the　wireless　signal　propagation　model　parameters　A　and　N　through　experience　or　large　amounts　of　data,　so　they　are　dependent　on　experience　and　are　not　strong　universal　algorithms　for　location　of　the　different　environment,　also　low　accuracy.　After　lots　of　research　and　analysis　of　radio　signal　propagation　model　and　the　traditional　indoor　location　algorithm,　a　new　indoor　location　algorithm　uses　BP　neural　network　to　fit　the　distance-loss　model　is　proposed.　From　a　number　of　distances　between　reference　nodes　and　blind　node,　Taylor　series　expansion　algorithm　is　used　to　determine　the　coordinates　of　the　blind　node.　Finally,　the　experiment　result　shows　that　the　new　algorithm　improves　the　positioning　accuracy　and　universality,　compared　with　the　traditional　positioning　algorithms.