Research On Wind Turbines Grouping Of Wind Farm Based On Measured Time Series

An accurate and suitable dynamic model of wind farm is demanded for wind power grid stability analysis,but usual single machine equivalent model is not applicable which ignored the difference of operating condition between generators.By using measured time series and similar operating conditions principle,a wind turbines groping method for wind farm is proposed based on measured time series.The specific content is as follows:(1)The mathematical model of wind turbine is establishied,which including wind speed model,wind turbine model(aerodynamics model,shaft system model,pitch angle control model),and generator model.(2)Based on the measured time series of the wind field and the usual clustering algorithm,this paper describes the principle of cluster partitioning based on time series clustering algorithm.Firstly,according to the definition and classification of time series,combined with the actual operation of the wind farm,the actual time series of wind farm are selected and constructed.Secondly,the paper introduces the common clustering algorithm in clustering analysis from the aspects of clustering algorithm classification and clustering similarity measure.Then,this paper selects measured active power time series of wind turbines as the clustering index.K-means clustering algorithm is used to cluster classification based on an actual wind farm,and the rationality of the grouping result is analyzed by using the outline values of WTGs as evaluating indicator.Finally,the dynamic equivalent model of wind farm is established,and the simulation analysis is carried out by taking the B4-39 CIGRE system of a certain actual wind farm as an example.(3)Aimed at the defects of the traditional K-means clustering algorithm and the irrationality of the cluster,a wind turbines groping method for wind farm is proposed based on simulated annealing optimized K-means clustering algorithm.Firstly,this paper selects measured active power time series of wind turbines as the clustering index,uses K-means algorithm for relatively better grouping,and then combines with the simulated annealing algorithm,makes a more accurately heuristic random searching in order to find the global optimal grouping result which has the minimum dispersion.Secondly,the paper compares and analyzes the outline values of the two algorithms,and the results show that the proposed algorithm is reasonable.Finally,a comparative analysis of the several equivalent models dynamic response is analyzed.Results show that the equivalent model proposed in this paper can reflect the dynamic responses of wind farm more accurately.