The Research Of Grid Low Frequency Oscillation On-line Detection Method Based On WAMS

With the development of National Smart Grid, multiple low frequency oscillation accident happened and it has brought adverse effects to the state and the people. When low frequency oscillation occurs in power grid, the district grid interconnection will result in a chain reaction. If it is not found and handled in time, it may cause a large area power grid blackout and bring serious consequences to the people’s daily life. Although the existing low frequency oscillation detection methods can improve the detection speed by using the recursive fast algorithm and the raised cosine filter, it is difficult to meet the requirements of online detection because of its complex operation process and long time consuming and other shortcomings. Therefore, SVM classification algorithm is proposed in this paper, which can solve the disadvantages of the above methods, and provide a new idea for the realization of online detection. For the acceleration of the detection of low frequency oscillations, the Scheduling workers have to take appropriate measures to eliminate or suppress the low frequency oscillation in power system, it has important practical significance to study the method of rapid detection of low frequency oscillation in power grid for the safe and reliable operation of power network.Using the wide area measurement system(WAMS) to obtain the advantage of data in real time, power and power angle are selected as the basis for the detection of low frequency oscillation of power grid in this paper, and a new method for online detection of low frequency oscillation of power network is proposed. This paper is divided into two parts, one is the calculation of the power of the wave frequency; the other one is to classify the frequency of the wave and power angle, and build a classification model. Putting the power and power angle data of PMU real-time by the calculated results, frequency and the rate of power angle change pretreated, into the detection model, and then realize the online detection of low frequency oscillation of power grid. For the calculation of power fluctuation, the Prony algorithm to simulate the fluctuation of the signal is used in this paper, at the same time,the improvement measures are added to achieve the power fluctuation frequency. In order to quickly detect fluctuation frequency is in the frequency range of the low frequency oscillation,and join angle to limit the power fluctuation amplitude, the algorithm of support vector machine was put forward, through the training sample data to obtain the optimal hyperplane,namely, training all vector can be the hyperplane correctly divide. Using this principle, the establishment of detection of low frequency oscillation in power system classification model.And by using the artificial fish swarm algorithm to the optimization of the SVM model, the algorithm mainly through simulation of artificial fish foraging behavior, tailgating, poly group behavior, the stochastic behavior of the training samples to train, looking for the best c,g, and the optimal fitness(accuracy). At the same time, the particle swarm optimization algorithm and genetic algorithm are introduced to optimize the parameters of the SVM model,and the three optimization algorithms are compared to determine the advantages of this method. SVM classification algorithm and artificial fish swarm algorithm are combined to improve the classification accuracy of the model to detect the low frequency oscillation of power grid. Finally, the WAMS measured in real time with standard of power and power angle data, by improved Prony method, frequency and the rate of power angle change pretreated data are input into the construction of SVM classification model, if power system in the low frequency oscillation, then output 1 is low frequency oscillation, also can according to the occurrence of low frequency oscillation of power and power angle with the standard to determine the specific time for low frequency oscillation occurs, for managers to provide convenient conditions; if there is no low frequency oscillation output 0(no low frequency oscillation). In the end, the program is written to realize the online detection of low frequency oscillation in power grid.The MATLAB simulation software is used in this paper, using the simplified system model of Daqing Electric Power Bureau and the initial data of four machine of two area interconnected system examples as measured by PMU to verify the feasibility of the proposed method.