A Study of Frequency Estimation and Bad Data Detection of Power Systems
| Autor: | Chien-Hung Huang, 黃建宏 |
|---|---|
| Rok vydání: | 2010 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 98 Fundamental frequency estimation of distorted power system signals and bad data detection are two important topics in the energy control center of a power system. This dissertation proposes a robust technique consisted of a robust extended complex Kalman filter and a sliding surface-enhanced fuzzy adaptive controller (RECKF-FAC) to enhance the fitering performance and to promote the accuracy in fundamental frequency estimation. As for bad data detection, this dissertation presents an extended complex Kalman filter (ECKF) to train the link weights of the complex artificial neural network (CANN) to gain better immunity from noise of training data, to enhance the detection accuracy and to advance the ability of bad data detection. A robust technique incorporated a robust algorithm and fuzzy adaptive controller embedded into the extended complex Kalman filter is presented for fundamental frequency and amplitude estimations of distorted signals in a power system. With the aid of the robust algorithm and the fuzzy theory, the proposed approach is more effective for solving the uncertainty of fundamental frequency estimation. It is employed to suppress the measured signal contains noises or harmonics for promoting the efficiency in fundamental frequency estimation. Moreover, it not only can perform the ECKF without changing any form of the ECKF, but also can enhance the estimation accuracy and reduce the computation time. To validate the feasibility of the proposed approach, it is examined on different conditions of the simulation power system and the practice power system. Additionally, the comparison of the proposed approach and frequency relay is used in the substation to verify its effectiveness. As for the bad data detection, this dissertation presents a complex artificial neural network based on an extended complex Kalman filter which is used as a measurement estimator for diagnosis of bad-data in state estimation of a power system. Due to complex-type variables used in the proposed method, advantages of the proposed method are: (1) the size of the CANN is trimmed to result in a better convergence behavior; (2) the nonlinear mapping function of the CANN and the complex measurements are coordinated well to improve the detecting performance; (3) the noise immunity from the training data is enhanced since the link weighting can be adjusted automatically with the ECKF; (4) the ECKF-CANN does not need the value of the learning rate. Finally, a 6-bus, IEEE 30-bus power systems and a practical system are used to verify the feasibility of the proposed method. The results show the performance of the measurement estimator used in the proposed method is better than the conventional method. Moreover, the proposed approach can detect and identify the bad data simultaneously. As a result, the proposed method can improve correctness of the bad data detection in a power system. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|