Research On Wind Farm Equivalent Modeling Method Based On MMPSO And BP Neural Network

In recent years,the total installed capacity and power generation of wind power continue to grow rapidly.It is particularly important to study the impact of large-scale wind farm access on the dynamic stability of power systems,witch requires accurate wind farm equivalent models to characterize the overall dynamics.How to establish an accurate and effective wind farm equivalent model has always been a hot and difficult problem.Equivalent models need not only to achieve the simplicity,but also to meet the accuracy and value requirements of power system analysis and calculation.Different wind farm equivalent modeling methods are studied here.First of all,according to the traditional mechanism modeling method firstly,in order to study the equivalence modeling of wind farm composed of doubly-fed wind turbines,a detailed model of doubly-fed wind farm is built in MATLAB/Simulink to simulate the operating characteristics of actual wind farm.Firstly,the wake coefficient is introduced as the clustering index.The 4-machine equivalent model of the wind farm is obtained by volume-weighted aggregation of wind turbine parameters.Secondly,in order to achieve higher standards of equivalent accuracy,this thesis applies the parameter identification to the model.The grid-side fault is used as the incentive.Key parameters and general parameters are distinguished according to trajectory sensitivity of each equivalent unit parameter.The relationship and variation between the trajectory sensitivity and the output power of different units are analyzed by simulation.Thirdly,the MMPSO is used to identify the key parameters of the electrical and controller parts.The equivalent model with better dynamic performance after parameter correction is obtained.Then,Because of the expansion of the scale of wind farms,the differences in types,parameters,control methods of wind turbines in the same wind farm is increasing significantly,witch makes the difficulty in defining clustering,worsening effect of reducing and reduced applicability.In order to study the equivalent modeling of hybrid wind farms,this thesis applies the idea of the "black box" system identification and wind power uncertainty to equivalence modeling.Finally,after the simulation data acquisition of the input and output external characteristics under the transient and steady state conditions of the hybrid wind farm,the complex structure and coupling relationship inside the wind turbine are ignored.Equivalent nodal models of wind farm are established under the condition of electromechanical transient faults and disturbances of wind direction and wind speed fluctuation based on neural network.The feasibility and practicability of the equivalent modeling method are verified by the historical operation data of actual wind farms in a certain place.