The Research Of Grid Multiple Disturbances Online-detection Method Based On WAMS

The stable and safe operation of electric power system is closely related to the economic development of the society and the life of the people. With the increasing scale of power grid, and the operation of the power grid becoming more and more complex, there are various kinds of disturbances in power grid. In the actual power system, the disturbance is often not as a simple single form, and often occurs in the form of multiple disturbances. If the disturbances in the power grid can not be detected in time, the system condition will gradually deteriorate; it is likely to cause a large area of power outage. The traditional detection methods are mostly for single disturbance, ignoring the actual situation of the power grid. And the data source is mostly based on the traditional data acquisition and monitoring system SCADA. Because of the SCADA uploads relevant information to the dispatch center in a certain time interval(3~5s), can not detect the multiple disturbances on line. The construction of Wide Area Measurement System provides a good platform for the on-line detection of multiple disturbances. The data which uploaded by WAMS are in high precision, fast transmission speed and with a unified standard, can realize the dynamic observation of power grid. Therefore, it is significant to propose a method of using WAMS to detect the multiple disturbances of power system on line for the stable operation of the power grid.This paper first analyzes on the theory that the current and voltage of the system will be changed due to the redistribution of power flow when the disturbance occurs in the power grid. When multiple disturbances occur, the variation law of the multiple disturbances presents the superposition. Observe the change rule of electrical parameters by making a large number of simulations in different sequences of multiple disturbance in 3 machines 9 buses system. Through a large number of simulation and data analysis, the relative change of the voltage and current are chosen as the characteristic values of the multiple disturbance detection. By using the idea of pattern recognition, the multiple disturbance detection can be regarded as a criterion of the operation state of the power network. Choose the Extreme Learning Machine which has the advantage in the classification algorithm as the detection model of multiple disturbances. The relative variation of the voltage and current are divided into the normal operation data and the disturbed mutation data to train the ELM model. We can get a good performance testing model by selecting different activation function and the number of hidden layer neurons. By using this multiple disturbance detection model, it can be determined whether the relative input voltage and current variation within the scope of the normal operation of power grid. Two signals are detected simultaneously, and when the two characteristic values are simultaneously identified as the disturbance mutation data, the system will send out the alarm and display the specific time of the disturbance. To built BP neural network and SVM model with the same data. Considering the training speed, the training accuracy, testing speed and testing accuracy of these three models, it is verified the feasibility that the ELM is applied to multiple disturbances online detection. This paper proposes the processes of establishing multiple disturbances model based on WAMS and online-detecting the disturbances. And a multiple disturbance detection program is developed. In order to verify the feasibility of the method to detect the different sequences of multiple disturbances in 3 machines 9 buses systems, such as the double disturbance of the cutting machine and the load. The results show that the detection system can accurately judge the running state of the system, send out the alarm, and the output time is accurate. In order not to lose the general, applying this method to the New England 10 machines 39 buses system for verification. The feasibility of the proposed method is verified by the results.