The Research On Substation Load Clustering And Aggregate Load Modeling

Load modeling plays an important role on the operation, planning and control for power system. Based on the preliminary theory research, load clustering and model identification using the comprehensive measuring method are deeply studied in this paper. And it provides qualitative correctly load model and quantitative accurately simulation results. The specific work is performed as follows:In this paper, the practical research on fuzzy clustering is carried out considering the load characteristics. On the basic of the characteristics of two kinds of fuzzy clustering method, an improved fuzzy C-means method is proposed and is applied for load characteristics clustering of 220kV substation in Hunan. The clustering center obtained using the square sum of deviations is used as the initial parameters of fuzzy C-means method to achieve the second clustering. The clustering results verity the feasibility and effectiveness of the improved method, and the method overcomes the shortcoming of initial parameters sensitivity of traditional method. The cluster results are used as the choice basis for the installation of the measurement device in typical substation, and provide accurate data for measurement-based load modeling.Furthermore, an aggregate load model is adopted in this paper. Considering the shortcomings of Gauss-Newton least square method, the damped Gauss-Newton least square method with a damping factor is applied for the synthetic static load model identification. The simulation results are compared between traditional Gauss-Newton method and the new method. Based on this, the genetic algorithm - pattern search method is proposed for dynamic load modeling comparing with traditional search method. The simulation results indicate that the methods employed in this paper have fast learning speed and high identification precision and the model have good interpolated and extrapolated ability. Therefore, the methods can describe nonlinear characteristics of the integrated power system load model more accurate.The research on the substation load characteristics clustering and model identification provide some reference value for further development of practical synthetic load modeling.