Research On Transient Stability Assessment Of Electric Power System Using Deep Learning Approach

Transient Stability Assessment(TSA)is of great significance for the stable and safe operation of power system.TSA information is used to make control decisions during emergencies and to support the scheduling operation of the power grid.At present,in order to meet the growing demand for electricity and the wide application of power electronic equipment,the interconnection of the system,the requirements of transient stability,it is urgent to carry out in-depth research and analysis of TSA.Failure to evaluate transient stability can result in severe failures,such as power outages and some other limitations in the power system.The deployment of different sensors,PMUs and other measuring devices has been increased due to its advanced features and time-synchronized performance.The conventional method of TSA cannot handle large amounts of data from these devices.Deep learning methods can be evaluated more accurately and faster at early or several emergencies to make quick control decisions to protect the system from severe conditions and downtime.The traditional time domain simulation and direct method are mostly used for the analysis based on trajectory and energy function,the methods are time-consuming and cannot meet the current large-scale calculation requirements of high-power system.Therefore,deep learning methods are needed for data mining and feature extraction for transient stability assessment.At present,the relevant deep learning methods improve the calculation speed and evaluation accuracy of TSA,but there are limitations in solving large amounts of computational data,and there is a lack of quantitative evaluation and robustness.Therefore,in order to improve the TSA calculation method,two different models and algorithms are proposed.Firstly,a hybrid CNN-LSTM model is proposed,which improves the accuracy of nonlinear relation altogether mapping between input and output data,and verifies its accuracy by extracting voltage phase characteristics and realizing transient stability evaluation.Secondly,the Bi-LSTM attention mechanism model based on voltage phasor quantity is proposed,the Bi-LSTM attention mechanism is used to map the relationship between voltage phasor and system transient stability,and by establishing a sample matrix of transient stability information in the initial stage of the perturbation level,the extracted features are more robust,effectively reducing the false and missing samples,thus improving the generalization ability and evaluation performance of the model.Furthermore,by adjusting the network structure parameters of the best evaluation indicator.The mapping model between input features and transient stability is established to further reduce false positives and sample loss,and to improve the accuracy of network model evaluation.The improved model combines Bi-LSTM feature extraction layer and attention mechanism to form a hybrid mode for a transient stability classification model,and the IEEE-39 bus New England test system is used to verify the accuracy of the model,and wide-area noise is introduced into the generated data to evaluate the robustness of the system.Finally,the method is used to realize the transient stability evaluation of the power system based on voltage phasor,and the validity of the proposed model is proved by comparing the proposed hybrid model with another model.