| China is forming a super-large-scale interconnected power system with the development of power industry.It's a big problem to ensure that such a large-scale power system can operate economically,safely and stably.Therefore,it is necessary to analyze the transient stability of the power system under large disturbances.It is best to be able to online respond to disturbances with measures of prevention and control by transient stability simulation.However,the existing transient stability simulation methods,especially the time domain simulation method,are difficult to achieve real-time application in the actual power system due to the slow calculation speed.Two different methods are proposed in this paper to accelerate transient stability simulation.At first,a fast algorithm for transient stability which based on predicting virtual injection current with artificial neural network is proposed in this paper and parallel algorithm is used to improve efficiency of the method.The reliability and effectiveness are proved with 10-generator New England test system and example which consists of 3 generators and 9 buses.'The number of iterations of the network equation and differential equation can be significantly reduced.As a result,the time cost of transient stability calculation is reduced.The training of the artificial neural networks can be processed off-line,so the method proposed is useful for the application of online calculation of transient stability.Then a method which is useful for the evaluation of transient stability is proposed according to the type of data that is suitable for the convolutional neural network to process.Because it's suitable for training convolutional neural networks with the power angle data of all generators in the power system for a period of time after the fault occurs to the time the fault is resected.The feasibility and accuracy of the method in evaluation is proved through the test on the 10-generator New England test system and the IEEE118-node system.A layer-by-layer filter of the test samples according to the output of the output layer of the convolutional neural network can pick out most of the samples that cannot be accurately determine.d.Also the method can improve the accuracy of the evaluation of transient stability.Because of the reduction of the input of power angel,the transient stability simulation can be terminated early,so the simulation time is greatly reduced,and the online application of transient stability simulation is realized. |
PDF Full Text Request