In　order　to　improve　the　temperature　measurement　accuracy　of　microwave　radiometer,　the　inversion　algorithm　of　interferometric　microwave　radiometer　is　improved.　In　the　paper　propose　an　improved　PSO-LM　algorithm　based　on　BP　network,　which　combines　Particle　Swarm　optimization(PSO)　and　Levenberg-Marquardt(LM).　By　analyzing　the　influencing　factors　of　the　temperature　measurement　process　of　microwave　radiometer,　the　influences　of　radiometer　output　voltage,　transmission　line　temperature,　aluminum　tube　temperature　and　antenna　temperature　on　the　temperature　measurement　results　are　considered　in　the　inversion　algorithm.　In　order　to　confirm　the　effectiveness　and　superiority　of　PSO-LM　optimized　BP　network　in　water　temperature　inversion　experiment,　this　paper　selects　PSO　algorithm,　LM　algorithm　and　PSO-LM　algorithm　to　optimize　BP　network　for　comparative　experiment　and　cross-validation.　The　experimental　results　show　that　using　PSO-LM　algorithm　to　optimize　BP　network　inversion　has　higher　accuracy　and　faster　convergence　rate　than　traditional　PSO　algorithm,　and　is　more　stable　than　LM　algorithm.　PSO-LM-BP　algorithm　can　improve　the　temperature　measurement　accuracy　of　microwave　radiometer　to　0.207°C,　which　makes　the　temperature　measurement　accuracy　of　microwave　radiometer　significantly　improved,　and　has　certain　practical　value　and　social　significance.　©　2021　IEEE.