Dynamic State Estimation For Synchronous Generator Based On Extended Set Membership Filter

Dynamic states of a synchronous generator are important reference for power system wide area control system. The dynamic states have been used in many controllers, such as power system stabilizers and automatic voltage regulators, as feedback signals to maintain or improve transient and small signal stability of a power system. Hence, accurate information of the dynamic states is important for reliable and efficient operation of a power system.Due hard to obtain real-time network topology during electromechanical transient process after system fault, the traditional network topology based state estimation was unable to apply. Therefore, using PMU to measure terminal electrical variables of a synchronous generator timely, the synchronous generator and the remaining system was able to be decoupled. In other words, it is not need to consider the impact of network topology. Then establishing a dynamic mathematical model of the generator that has considered the no mutation characteristic of power angle and rotor speed after system fault, the state variables(including power angles and rotor speeds) could be estimated by dynamic state estimation algorithm, rejecting PMU measurement errors at the same time. This is so called generator dynamic state estimation.Along with the derivation of generator dynamic model and kalman filter algorithm, the operating principle of kalman filter under traditional generator practical model and the problems facing practical application were detailed. By improving generator dynamic model and dynamic state estimation algorithm, a new generator dynamic state estimation method was proposed. Aiming at the unequal relationship between the power angle and the rotor angle during electromechanical transient process after system fault, a new generator dynamic model that considered the unequal relationship between the power angle and the rotor angle was brought up. And on this basis, aiming at the shortcomings of EKF, extended set membership filter(ESMF) based on unknown but bounded noise hypothesis was further put forward to solve the proposed state estimation model. However, because of the high computational burden, the computing time of ESMF is not satisfactory. Thus, a UD factorization-based ESMF algorithm was proposed to reduce the computing time and improve the numerical stability by UD factorization of error covariance matrix. In IEEE 9-bus system simulation analysis, the proposed method can estimate the real-time power angle trajectory of generator during power system electromechanical transient process effectively. What's more, it showed high accuracy and robustness against measurement and model noise.