Research On Transient Stability Assessment Method Of Renewable Power System Based On Response Data

Power system transient stability assessment plays an important role in ensuring the safe and stable operation of power grid.Rapid and efficient assessment of transient power angle stability can help the operators to take effective measures in time to ensure the safe operation of power grid.In recent years,with a large number of new energy such as wind turbines and photovoltaic power plants connected to the power grid,the operation mode and dynamic characteristics of the power grid have changed significantly.In particular,wind power generation is mainly based on large-scale development and centralized integration,which has profound impact on the steady-state operating point of the power grid and the transient process after disturbance.The reason is that wind turbines have the characteristics of dynamic response output,low inertia and intermittent characteristics,which are different from the conventional synchronous generator.At the same time,it brings more uncertain factors to the security and stability of the power grid,and also poses new challenges to the transient power angle stability assessment.The traditional stability assessment method based on time-domain simulation depends on the accuracy of the model.As the control strategy of wind turbine is more complex,and there are many power electronic devices and different parameters,it needs a more detailed model.Besides,the wind power output and power system fault scenarios have various uncertainties,which is difficult to meet the real-time and accuracy requirements for largescale power grid online evaluation.In recent years,the data-driven power grid stability assessment method has developed rapidly,which provides a new way to solve such problems as the advantages of strong learning ability and fast assessment speed.Based on the state information before and after the fault obtained by the power grid measurement device,the thesis carries out the research on transient power angle stability assessment of the power grid with wind turbines.The main contribution are as follows:1.The input characteristic set of transient power angle stability assessment is established comprising of the wind power dynamic response characteristics.Firstly,the transient response process of synchronous generator is analyzed,the physical quantities strongly associated with power angle stability are found according to the rotor motion equation,and then the transient characteristic quantities of synchronous generators are constructed.Secondly,taking the parallel connection of doubly fed wind turbine to single machine infinite bus system as an example,the relationship between the transient output of wind turbine and the output electromagnetic power of synchronous generator as well as the system power angle stability are studied.Then the key physical quantities of the wind power that affecting the power angle stability of synchronous generator are calculated.Finally,the conventional generator characteristic quantity and wind power characteristic quantity are combined to obtain the transient stability assessment characteristic set containing wind power dynamic response.The rationality of the above characteristic set is compared and verified in a modified IEEE39 bus system by using the evaluation accuracy and other evaluation indicators for the machine learning model.2.A feature selection integration method based on improved minimum redundancy maximum relevance(mRMR)is proposed to solve the problems of high dimension of original feature set,large information redundancy and low model training efficiency.Firstly,the correlation and redundancy measures are introduced to generate a variety of improved mRMR algorithms and integrate them to determine the best evaluation criteria.Secondly,in order to enhance the search ability of the algorithm,some quantiles are introduced to adjust the correlation and redundancy weights in the subsequent incremental search process,and multiple initial features are used to expand the search space to obtain multiple groups of nested candidate feature subsets.Finally,support vector machine(SVM)is used to verify the features one by one to obtain the optimal feature subset,and the performance of the selected optimal subset is verified in the IEEE39 node test system and a provincial power grid.3.An integrated evaluation method of adaptive differential evolution optimization extreme learning machine(ELM)based on bagging strategy is proposed,which can be used for evaluating the transient power angle stability of the power system with high penetration of wind power.Firstly,aiming at the problems of low prediction accuracy and instability of the ELM,an improved adaptive differential evolution algorithm based on Logistic chaotic map is proposed to optimize network parameters,and the network structure is optimized by grey relation analysis.Then,the ensemble learning method is introduced,and the above model is used as the basic classifier,the accuracy and generalization performance of the model are further improved by improving the bagging strategy sampling method,using improved hybrid chaotic mapping and a variety of fitting rules to enhance the diversity of the model.Furthermore,aiming at the problems of unbalanced sample categories and different misclassification costs in different scenarios in practical applications,a cost-sensitive extreme learning machine model based on improved adaptive differential evolution optimization is proposed by introducing the cost-sensitive matrix,which can greatly reduces the missed judgment rate for unstable scenarios.Finally,IEEE39 nodes test system and a provincial power grid are used to verify the effectiveness of the proposed method from the aspects of optimal hidden layer node selection,classifier performance and the stability evaluation results,respectively.