Research On Load Characteristic Analysis And Modeling Of Regional Power Grid

As an important part of the electric power system,the power load plays a very important role in the analysis,simulation,calculation and other aspects in the power system analysis.There are many substations in large-regional power grid,thus it is impossible to do the layout and modeling of the load measuring devices to all substations,which is very unfavorable to the economy and security of the power grid.And in the planning stage,problems encountered in the modeling of unfinished substations have not been well resolved.Based on the load composition of the substations,this paper discusses the distribution of the load measuring devices and the prediction of the load models.The composition of the power load is very important for the measurement-based load modeling.The statistical work on the composition of the power load is often completed by the statistical reporting of the substation staff in the past thus it is inconvenient to do that very often and it is hard to ensure the accuracy of the statistical results.As time goes by,the actual load would change,especially the proportions and compositions of various loads.The load curve is easy to obtain,thus quantitative analysis of the degree of correlation between the index of load characteristics and the main factors enables researchers to master,analyze and predict the trend of load characteristics and load compositions to a certain extent.Speculate the load compositions through the load curves by the MATLAB program through the analysis of previous load compositions.And historical data verified that the method can estimate the load compositions accurately and can be used in measurement-based load modeling.In the field of measurement-based load modeling,a variety of problems are complex and difficult to solve.For the problems in grouping modeling of the power load,the load data can be grouped by the cluster analysis theory in statistics,and the commonness of the same group of data can be found as a label for them.The variable adaption of the integrated load model structure while selecting the TVA uses the adaptive genetic algorithm to identify the model parameters,and a load model that is generally applied to the group is identified by multi-curve fitting.The load model can be identified according to the label when being selected.An optimization method of PMU distribution is proposed,that is to classify by the CURE method in the hierarchical clustering method according to the load compositions in the substations.The substations with higher failure rate recorded in the security system of each group can effectively obtain the representative fault datawhich can be used for load modeling.Based on the current research results,the advantages and disadvantages of the widely used neural network algorithms are compared and analyzed,and a method of predicting the load model is proposed based on the extreme learning machine.This method predicts the load models and extracts the load compositions of the load characteristic data based on the extreme learning machine theory.The method plugged into the extreme learning machine model as training set,the parameters of the extreme learning machines are obtained through training.At the load model prediction stage,the load composition of the target load data is taken as the input set,and the predicted load model is obtained by training through the extreme learning machine model,thus provides a new idea for the prediction of load model in the grid planning stage.