Research And Application On Wide Area Measurement Based Power System Dynamic Equivalent Modeling

Due to complex structure,high model order and strong nonlinearity of modern power system,the modeling process and the corresponded model simplification methodology are essentially to be investigated in respect to power system analysis and control.Nowadays,with the co-existing of interconnected large-scale power gird,highly penetrated renewable energy system and the distributed energy resources featured active distribution network,the complexities of power system modeling have been further intensified.In recent years,with the development of wide area measurement technology,actual measuring based dynamic equivalent modeling plays an important role in power system security and stability analysis.However,traditional method derived dynamic equivalent model faced to the problem of weak generalization ability,which would definitely reduce the accuracy of model and significantly limit the application in practical uses.Meanwhile,due to the large difference in dynamic characteristics of different system components,it is hard to find out a universal method of dynamic equivalent modeling when high proportion of renewable energy is penetrated into the grid.Therefore,in order to address these issues and better adapt to the different application in modern power system of China,the novel method on wide area measurement based power system dynamic equivalent modeling method,which has high accuracy,strong model generalization ability and wide adaptability,is required to be comprehensively developed.This dissertation focuses on the establishment and simplification of modern complex power system models on the basis of wide area data,and the main research contents and major contributions are briefly concluded as follows:(1)A systematic solution to improve the generalization ability of power system equivalent model is proposed.In order to mitigate the impacts of incoherent power generation units on the accuracy of equivalent model,coherency identification method is explored to cluster the equivalent grids,and the corresponding equivalent sub-models are established respectively;to solve the “multiple solution problem” of equivalent model,the key parameters,which have high sensitivity and can be uniquely identified,are screened out based on the method of parameter sensitivity and correlation analysis.Meanwhile,the key parameters of each equivalent sub-model are fully decoupled and identified step-by-step with the application of hybrid dynamic simulation technique;finally,in order to improve the characterization ability of equivalent model parameters to different system faults,the multi-faults featured multi-objective optimization identification of equivalent model parameters is performed.With the necessary modifications,this proposed genetic method can be widely adopted in different application scenarios of power system dynamic equivalent modeling.(2)A dynamic equivalent modeling method of small hydropower generator stacks considering model generalization ability is proposed.Firstly,the time series clustering method is used to divide the equivalent system into several coherent subgroups,for which the corresponding equivalent sub-models are established respectively;then,based on the sensitivity and correlation analysis method,the key parameters to be identified are screened out;finally,based on the proposed multi-objective optimization algorithm and hybrid dynamic simulation technology,the key parameters of each equivalent sub-model are decoupled identified step by step.The effectiveness of the proposed method is verified by a case study of dynamic equivalent modeling of small hydropower stacks in southwest China.(3)A dynamic equivalent modeling method of active distribution network considering model generalization ability is proposed.Firstly,the system characteristic database,which can characterize the different operation states of the active distribution network,is established.In the meantime,the operation states of the active distribution network are classified by two-step clustering method and Fisher discriminant method;secondly,the key parameter identification method is introduced to eliminate the "multi-solution problem" in parameter identification process;finally,through the Elman neural network,the nonlinear mapping features of different operation states and equivalent model parameters are obtained.In addition,an actual active distribution network system in the southeast coast of China is used to verify the effectiveness of the proposed method.(4)A dynamic equivalent modeling method for wind farms is proposed.Firstly,the wind turbines are grouped by time series clustering method,and each subgroup is aggregated into a single wind turbine sub-model;secondly,key parameters of the wind generator and control module of equivalent sub-model are screened out by the sensitivity and correlation analysis method;finally,based on multi-objective optimization algorithm and hybrid dynamic simulation technology,the key parameters of each wind turbine sub-model are fully decoupled and identified step by step in respect to wind speed fluctuations and external system faults.The effectiveness of the proposed method is verified by a real wind farm model in northwest China.(5)A probabilistic multi-machine represented dynamic equivalent modeling method for wind farms is proposed.Firstly,Weibull distribution featured input wind speed probability model of wind farm is established.By considering wake effect and time delay effect,the probability database of wind turbine input wind speed is developed;A practical 4-machine clustering method is used to cluster wind turbines,and the clustering results are reasonably combined based on Fisher discriminant analysis;Finally,a probabilistic multi-machine represented equivalent model of wind farm is established based on capacity weighted method.The validity of the probabilistic equivalent model is verified by a real wind farm in northwest China.