Research On Evaluation Method Of Power System Transient Stability Based On Deep Learning

Along with the good trend of steady improvement of the national economy year by year,electricity has become an important foundation support for social development.How to ensure the provision of stable electrical energy support for the development of society has always been a topic of focus in the field of electric power.This paper studies the basic theory of neural network,the structure of the model and the training method of the model,as well as the strategy of optimizing the model parameters,and analyzes the relationship between the voltage amplitude change and the system transient,so as to draw the voltage amplitude information to the system The transient stability is evaluated and predicted.This article first introduces the basic theory,composition structure and training methods of neural networks.In order to enable the evaluation model to complete the parameter optimization quickly and efficiently,this article also introduces several methods of deep learning optimization to improve the theoretical basis for subsequent model optimization links in this article.This paper studies the basic theory and training methods of deep confidence networks,and builds a transient stability evaluation model to evaluate the transient stability of the power system.The deep confidence network model has great advantages in mining potential rules of data,and can quickly complete the extraction of transient stable features.In the process of model training,in order to improve the generalization ability of the model,a combined training method of supervised and unsupervised is adopted.The new England 10-machine 39-node system is used to obtain all required training data and the adjustment and optimization of model parameters are completed,and the performance of the model is verified using the IEEE-16 machine system and IEEE-50 machine system.The results prove that the transient evaluation model based on the deep confidence network has superior evaluation accuracy and fault tolerance.The above evaluation model is excellent in data mining and evaluation speed,but there is a disadvantage that it requires human participation in the selection of feature vectors.This process may lead to the loss of some important parameters.Based on the relationship between voltage fluctuations and transients,this paper analyzes the relationship between voltage amplitude changes and transient stability.Using the convolutional neural network as a deep learning framework,a transient stability evaluation model is constructed,which is directly Collect the voltage information of the system after the fault as the direct input information of the model.The advantage of this method is that it combines the feature extraction link and the assessment classification link,avoids the participation of human factors,and realizes the "end-to-end" intelligent assessment.The new England 10-machine 39-node system was used to obtain all the required training data and the model parameters were adjusted and optimized.The performance of the model was verified using the IEEE-16 machine system and IEEE-50 machine system.The test results showed that the comparison Good progress has been made in the previous evaluation model.