Research On Clustering And Measurement-based Modeling Of Loads In A Large-scale Power Grid

Power system simulation calculation is the basis for the formulation of large-scale power grid control operation strategies,and the accuracy of the load model has an important impact on the simulation results.At present,large-scale power grid load modeling faces many difficulties.The generalization ability of the overall model is not strong,which makes the simulation results of some areas have a large gap with the actual situation,and the models for special loads are not mature enough.For example,my country has built a large number of electrolysis In aluminum plants,the proportion of electrolytic aluminum load in the grid capacity is increasing,but its model has not been fully analyzed and studied.For large power grid modeling,substations should be clustered to accurately identify model parameters.For special loads such as electrolytic aluminum,the unique load characteristics should be considered in the modeling.In view of the above problems,this paper has done the following work in the research of substation load clustering,parameter identification algorithm and electrolytic aluminum model structure:(1)For substation load clustering,this paper uses the proportions of different industries in substations as the clustering basis.Aiming at the problem of hierarchical dendrogram tilt that may occur in the hierarchical clustering algorithm,Gini coefficient and fairness index are introduced for improvement,and it is actually applied to a certain 220kV substation owned by the province.And the use of contour coefficient evaluation and T-distribution random adjacent embedding solves the problem that the clustering effect is difficult to quantify.By comparing the clustering effect of the improved algorithm with the original algorithm,it is shown that the improved algorithm can improve the accuracy of clustering.(2)For parameter identification,firstly,the principle and mathematical model of the firefly algorithm are introduced,and the defects of the firefly algorithm are explained,including the possibility of the objective function and the easy generation of local optimal solutions.After the algorithm is improved,the above-mentioned defects are eliminated.The simulation verifies that the improved algorithm has higher accuracy,which lays the foundation for the identification work in Chapter 5.(3)For the measured load modeling work of electrolytic aluminum,this paper uses the PMU data of multiple electrolytic aluminum plant power supply lines and applies the improved Firefly algorithm to various load models including constant current+10%induction motor model,ZIP model and comprehensive load model.The results show that the comprehensive load model is better than the ZIP model and the constant current+10%induction motor model.At the same time,for the load with strong fluctuations,this paper proposes an integral memory model,which is verified by measured data.The comparison of the above-mentioned multiple models proves its effectiveness.