Method for Identifying PMU Bad Data Based on Long Short-Term Memory Network

Phasor measurement units (PMUs) have become one of the most effective tools for state awareness of power systems. However, the complex environment caused PMU data to have quality problems such as data loss, which seriously affected its applications in power systems. This paper proposes a method for identifying PMU bad data based on long short-term memory (LSTM) network. First, the advantages of LSTM in bad data identification are analyzed. Based on the advantages, a two-layer network is constructed, and a decomposition and reconstruction method for original data is proposed. On this basis, two objective functions are defined, and different error characteristics are obtained. A method for determining the threshold of bad data based on decision tree is proposed to realize the identification of bad data. The results by simulations and field data verified PMU data quality is improved, making it better applied to all aspects of the power system.