| With the development of smart grid in our country and the improvement of demand for power supply quality by consumers, the electric power company has put forward higher requirements for real-time monitoring and on-line analysis and decision-making control of distribution network operation. Distribution network state estimation serving as an important foundation of distribution management system could provide other advanced application software with basic datas, it is of great significance to improve the distribution network safe and economic operation.State estimation method has been widely used and developed in power transmission systems, while less used in distribution network, especially in low voltage distribution network, the main problem lies in the differences of network structure and the lack of real time measurement. In recent years, a large number of smart meters of advanced measuring system have been installed and used in low voltage distribution network, it has been created favorable conditions for state estimation of low voltage distribution network. In order to improve the performance and result precision of state estimation methods for three-phase distribution network, it is very important to consider the power distribution network structure characteristic and types of the various measurements to establish the accurate three-phase state estimation model. Bad data detection and identification is the important content of power system state estimation, and is crucial for the impact on reliability of state estimation and qulity of the real time datas. Therefore, this paper makes a research on state estimation and bad data identification method for the three-phase four-wire distribution network system, some specific research work has been done as follows:① Robust state estimation method for three-phase four-wire distribution system using smart meters measurement data is proposed. According to the actual three-phase voltage and current as well as power measurements in real-time from smart meters, the exponential weighted function least squares estimation model is proposed in this paper, which is based on the current injection equation in three-phase four-wire distribution network. With no assumptions of zero potential of neutral voltage and known three-phase voltage balanced in the substation, the proposed model could enhance the convergence and robustness performance of state estimation significantly by utilizing the exponential weighted function of standardized residual and handling zero-injection equations as virtual measurement equations and equality constraints. The simulation results for modified IEEE13 node test system have indicated that the proposed method is efficient, robust and accurate.② Bad data identification method based on robust state estimation and k-means clustering algorithm is proposed. For the advantages of robust state estimation and k-means clustering algorithm in dealing with the bad data, the robust least squares estimation method based on the exponential weighted function as the error-suppressive aid is fisrt adopted to achieve the best estimate as well as the estimation error of measurements data, then the k-means clustering algorithm is utilized to improve the detection and identification of the abnormal data from the measurements. Meanwhile max-min distance method and gap statistics algorithm are embedded into k-means clustering algorithm to deal with unexpected influences caused by the initial cluster center and optimal clustering number, thus assisting in increasing the stability of clustering and bad data identification effect. Finally the proposed method has been simulated on the IEEE-13 nodes modified system and Chongqing jiangjin area network. The simulation results show that the method with high identification accuracy and strong practicability could be widely employed in solving the problem of abnormal data detection and identification in the aspect of system positioning, and has a good prospect of engineering application. |
PDF Full Text Request