A wide area measurement system（WAMS） based low frequency analysis method is proposed to make identifications and statistics of the small low frequency oscillations excited by daily random disturbances in power systems. By analyzing statistical data， the intrinsic oscillation mode of the power systems can be found from the daily field data. Moreover， the occurrence probabilities of the oscillation modes can be measured， and the modes which can be excited most easily are analyzed further to group the coherent buses， to identify the oscillation centers， to calculate the contribution factors， and to give dispatchers or analyzers an early-warning. While other oscillation analysis methods are based on mathematical models and specific operation conditions， the proposed method is independent of specific operation conditions and the accuracy or correctness of power system models. It will be a useful supplement to mathematical model based methods and experimental based methods. The application examples in Shandong Power Grid are presented to show the validity and the importance to apply this oscillation analysis method.