The Parameter Identification Of The Doubly-Fed Induction Wind Generator

Facing with the ever-increasing shortage of traditional coal, oil resources as well as the daily improved requirements for environmental nowadays, the wind energy has turned into a new clean energy with the highest level of development, utilization and commercialization among the renewable energy field. With the continuous expansion of wind turbine capacity, the influence of wind tubine operation state on the stable operation and stability of power system has growing more and more serious, which not only interferes the wavleform of system frequency and voltage, but also affects the power system scheduling directly. Therefore, it is of great importance to obtain accurate parameters thus to help power dispatching departments to coordinate arrangements for scheduling plans for studying the impact of large-scale wind power connected into grid, the calculation of real-time power system load and etc..As the parameters of doubly-fed induction generator can not be measured directly during its operation process, a 1.5 MW doubly-fed induction generator of a local wind farm in Gansu is utilized as the research obeject in this thesis, for the parameters of which is identified via forgetting factor based recursive least squares method, Model reference adaptive approach and simulated annealing particle swarm optimization algorithm, respectively. Then the algorithm with the highest identification accuracy is used as the basis for generator parameter identification. The identified parameters are used for estimating with the mathematical model based approaches utilized.Firstly, the doubly-fed induction wind generator mathematical model is set up and the parameter identifiability is analyzed on the basis of an analysis of its operation principle. Then, two time domain identification algorithms are proposed, namely the forgetting factor based recursive least squares approach and the model reference adaptive approach. The forgetting factor based recursive least squares approach overcomes the shortcomings where the parameters can not be updated timely which result in a saturation of the identification of values as the new collect parameters accumulated, thus improves the identification accuracy to a great extent. In the model reference adaptive approach, the observation and error model of the doubly-fed induction wind generator are set up with using the stator current as the state variables, where the self-adaptive rules of the parameters are designed with Lyapunov stability used. The identification results are obtained by simulating the set pu model in MATLAB platform, compared with the forgetting factor based recursive least squares approach, the model reference adaptive approach can get a higher accuracy.Secondly, the partical swarm algorithm is widely applied, as its algorithm is with the characteristics like easily calculation and good convergence, in dealing with actual engineering problems with the development of intelligent control. However, the particle swarm optimization algorithm is easily falling into local optimal solution in optimization process because of prematurity, which will result in insufficient of diversity of later found particles and the global optimal value is unable to obtain. To enhance the ability of global optimization objective, the chracteristics of the simulated annealing algorithm can effectively jump out of local optimal solution is combined with the rapid optimization of particle swarm optimization algorithm and a parameter identification approach based on simulated annealing particle swarm optimization algorithm is proposed, which exhibited with the highest accuracy when simulated in MATLAB environment.Finally, in order to help power dispatch system assess the state of generators, the power flow calculation and the impact of grid connection with large-scale wind power provide the basis. By establishing the power prediction model of doubly-fed induction wind generator and establishing the credibility index, the similarity of the actual power curve and calculated one is compared using simulation, which finally achieves the purpose for parameter identification.