Transient Stability Assessment Of Power System Based On Artificial Intelligence

Nowadays,wide area monitoring system(WAMS)has been widely used in power system.The research on transient stability analysis of power system based on WAMS is very hot.At present,the research methods mainly include time domain simulation,direct method and artificial intelligence method.Due to the rapid development of big data in power system,transient stability assessment based on artificial intelligence is regarded as one of the most promising methods.Many intelligent algorithms(including artificial neural network,support vector machine,decision tree,etc.)have been proposed by researchers to improve the prediction accuracy.However,there are some problems such as low accuracy and poor interpretability.To solve these problems,the main content in this thesis are divided into the following parts:1)By analyzing the movement of generator rotor and variation of power after fault in a simple power system and they can reflect transient stability to a certain extent.Therefore,the speed of rotor and the relevant operation parameters can be used to predict transient stability of power systemThe phasor measurement unit(PMU)in real power grid is simulated by PSASP.The data collected includes rotor angles,speeds and so on.The stability of the system is judged by whether the relative rotor angle of any two generators is less than 360 degrees.2)XGBoost-based algorithm interpretation and application on post-fault transient stability status prediction of power system is proposed.Firstly,the important features of generator operating state during transient process have been extracted by analyzing the dynamics of generator.Meanwhile,the redundant features are removed based on the correlation filtering and the model-based feature selection.Then,the relationship between the features and transient stability has been explained by the decision rules and feature importance scores.At last,the XGBoost model constructed is used to predict the transient stability based on the selected features under the specific operating cases.3)Since the traditional machine learning algorithm used in transient analysis requires the manual construction of transient stability features based on PMU data.The quality of feature construction directly affects the prediction accuracy,and the process of feature construction is time-consuming and laborious.In order to solve these problem,a deep learning-based transient stability evaluation is proposed.Compared with traditional machine learning,deep learning has two advantages: first,it has stronger fitting ability when faced with large amount of data;second,it can automatically extract features.Therefore,convolutional neural network(CNN)and long-term and short-term memory network(LSTM)are respectively applied to the prediction of power system transient stability.4)After the judgement for system stability,the deep learning methods can be introduced to combine with power system to locate the fault when the power system suffers fault.There are 34 AC lines in the IEEE 39 system.The accuracy of transient stability prediction and fault location based on CNN model can reach more than 99%.