With　the　proliferation　of　non-linear　loads　and　distributed　power　resources　that　connect　through　power　electronic　equipment,　decentralized　harmonic　pollution　in　three-phase　symmetric　power　system　becomes　increasingly　serious.　Despite　this,　monitoring　the　current　harmonic　of　all　substation　feeders　regardless　of　cost　is　unrealistic.　To　this　end,　a　multi-source　data　driven　method　is　proposed　in　this　paper,　utilizing　data　from　the　power　quality　monitoring　system,　dispatching　system,　and　market　business　system,　to　estimate　the　harmonic　spectrum　of　substation　feeder　currents　and　estimate　the　harmonic　current　components　of　each　feeder　with　monitoring　data　of　bus　voltage　harmonics　and　line　harmonic　currents.　The　objective　function　of　the　harmonic　spectrum　estimation　model　is　constructed　based　on　the　analysis　of　current　harmonic　phasor　and　power　relationship　on　the　feeders.　An　improved　particle　swarm　optimization　(PSO)　algorithm　is　proposed　to　solve　the　proposed　estimation　model　thereby　obtaining　the　current　harmonic　content　of　each　feeder　and　the　phase　angle　difference　between　the　current　harmonics　of　the　feeders.　The　performance　of　the　proposed　model　and　algorithm　is　validated　by　numerical　simulation　with　a　test　system　and　an　actual　system.