Considering　the　component　based　modeling　approach＇s　shortage　of　requiring　the　typical　load　model　parameters　and　the　traditional　measurement　based　modeling　approach＇s　inadequacy　of　the　dependence　on　fault-disturbance　data,　the　recently　proposed　ambient　signal　based　load　modeling　approach　offers　an　important　idea　to　track　the　time-varying　characteristics　of　power　loads.　However,　it　suffers　the　impacts　of　phasor　measurement　unit　(PMU)　measurement　errors.　Consequently,　data　improvement　for　ambient　signal　based　load　modeling　is　both　important　and　necessary.　In　this　paper,　state　estimation　is　adopted　to　make　the　ambient　signal　based　method　usable　with　field　PMU　measurements.　To　be　specific,　a　weighted　least　square　method　is　applied　for　state　estimation　to　improve　the　data　quality.　On　this　basis,　the　improved　data　are　used　for　load　model　parameter　identification　to　evaluate　the　performance　of　the　state　estimation　method　from　the　perspective　of　data　application.　Case　studies　with　both　simulation　data　from　power　system　analysis　toolbox　and　field　PMU　measurements　from　China　southern　power　grid　have　verified　the　effectiveness　of　the　proposed　approach.