Dynamic State Estimation of New Energy Power Systems Considering Multi-Level False Data Identification Based on LSTM-CNN

Abstract

With the increase of new energy integration, it is difficult to identify the measured data and false data in power system when they are mixed into cyber network. If false data with error information is utilized in the power system state estimation, the accuracy of state estimation will be reduced. The inaccurate estimation results will lead to wrong control decisions by operators. This paper proposes an improved dynamic state estimation method based on multi-level false data identification. This method uses innovation vector for the first-level identification, long-short term memory neural network for the second-level temporal identification, and convolution neural network for the third-level spatial identification. Through the identification, the mutation data are distinguished as fluctuant real data and false data. The identification results can provide precise operation information for power system, dynamically correct the filtering direction of state estimation and improve the accuracy of state estimation. The method is verified by IEEE-57 power system with actual operating data. The results show that the improved method can not only resist false data injection attacks, but also maintain high estimation accuracy in new energy power systems with strong data volatility.

Keywords

• New energy power system
• false data identification
• dynamic state estimation
• long-short term memory neural network (LSTM)
• convolution neural network (CNN)