Research On Transient Stability Prediction And Control Of Power System Based On CRMN And GR

With the continuous expansion of the scale of the internet power grid and the increasing proportion of renewable energy generation and power electronics equipment,the operating characteristics of the power system are becoming increasingly complex,and stability issues are becoming more prominent.In the face of the time-varying,complex coupling,and stochastic fluctuations of the power system,it has become increasingly difficult to complete power system stability prediction and control quickly and efficiently.With the widespread application of phasor measurement units(PMUs),information and communication technologies such as 5G,and artificial intelligence(AI)technologies such as Deep learning(DL),from a data-driven perspective,using big data and AI technology to extract features from the data collected by the power grid measurement system and predict the state of the power system may bring new solutions and technological paths for the safety analysis and control of the new generation of power systems in the context of transient stability prediction(TSP)and transient stability emergency control(TSEC)problems.Firstly,this paper introduces the research background and significance of the topic,outlines the current research status at home and abroad,and the main work of this paper.Then,the basic knowledge of TSP based on convolutional residual memory network(CRMN)and gated recurrent unit(GRU)is briefly introduced.The basic principles of CRMN,the construction and training methods of the model,as well as the evaluation metrics of the model are described.In addition,the focal loss function,which is introduced to address the problem of sample imbalance,is also discussed.Secondly,in order to improve the accuracy and reliability of TSP,a two-stage TSP method based on CRMN and GRU is proposed.This model improves the accuracy of TSP by fully mining the spatio-temporal big data of the power system.In addition,the reliability of TSP is improved by further predicting the generator rotor angle trajectory of unstable and critical stable samples.Simulation verification is conducted on a modified IEEE 10-machine 39-bus system and an IEEE50-machine145-bus system.The results show the effectiveness and superior predictive performance of the proposed method compared to traditional DL methods.Moreover,the case analysis also verifies the higher reliability of the proposed two-stage TSP results.Thirdly,to address the issues of model adaptability and sample imbalance in data-driven TSP,a mixed transfer learning TSP method that considers sample imbalance is proposed.This method uses Conditional Generative AdversarialNetworks(CGAN)to address sample imbalance issues and uses a mixed transfer of sample and model to address TSP model adaptability issues.Through a standard IEEE10-machine 39-bus system,an IEEE50-machine 145-bus system,and a Western Electricity Coordinating Council(WECC)system case,the results show that compared to traditional data augmentation algorithms,the sample generated by CGAN-based sample augmentation algorithms has higher quality and diversity.Additionally,numerical simulation results confirm that the proposed mixed transfer learning TSP method that considers sample imbalance effectively addresses model adaptability issues.Finally,in order to apply DL to TSEC,a real-time emergency control decision-making method for transient stability based on CRMN and honey badger algorithm(HBA)is proposed.The method uses focal loss function to solve the problem of sample imbalance and improve the accuracy of TSP.Additionally,through a large amount of offline simulation,sample generation,offline training,and optimization,the method achieves the online calculation of optimal emergency control strategy based on data-driven methods,which meets the requirements for real-time online control and enables the system to recover transient stability.Finally,simulations are conducted on a standard IEEE10-machine 39-bus system and an IEEE50-machine 145-bus system to validate the effectiveness of the proposed method.