Simulation Research Of Set Membership Filter For Power Signal State Estimation

The state estimation for power signal is the common problem of fault diagnosis and power quality analysis, which is of great importance in parameters measuring for power system and ensuring the safety and reliability of the power grid. The power signal state estimation is to measure the power signal waveform, analysis the amplitude fluctuations, frequency offset and phase deviation of the anomalous power signals to track and analysis the parameter saltation signals.Traditional power signal state estimation methods are based on assumptions of random noise, the noise source is typically set to satisfy a certain probability distribution. However, the probability distribution characteristics of the noise are not available in actual systems, and hard to judge whether the hypothesis is consistent with the actual situation or not. Traditional methods lose its effectiveness and couldn't provide a result with 100% confidence. For this situation, set membership theory only need to know the boundaries of noise, reducing the need for a priori knowledge of the noise distribution, showing its advantages in power signal state estimation.In this paper, a study of the implementations of set-membership filter for power signal state estimation is carried out. The main works are as follows:(1) The power signal is typically nonlinear and the probability distribution characteristics of noise are not available. For this, an extended set membership filter is proposed in this paper. And the simulation with classical extended kalman filter validated the feasibility of our algorithm.(2) The parameters of power signal are varied with time. For this, a reset update ellipsoid extended set membership filter and strong tracking extended set membership filter are proposed in this thesis based on the study of the reason the conventional ways failed to track the mutation signal.(3) From the ellipsoid volume, tracking error and calculation time, the tracking results of the proposed three algorithms are compared. The real-time frequency is extracted according to the state estimation result. The results performed on simulated data demonstrate the precision and efficiency of the proposed algorithm.