Research On Dynamic Reactive Power Optimization Of Distribution Network Based On VGG Network Load Classification And Improved Bacterial Foraging Algorith

With China’s economy getting better and better and the continuous advancement of urbanization,and the structure of power system distribution network is more and more complicated than before,and the power load is becoming more and more diverse,resulting in more and more pressure on power supply and transmission.In the distribution system,the problems of low voltage quality and high line loss caused by insufficient reactive power are becoming more and more serious,which not only affects the security of the power system,but also has a certain impact on the economy of the power grid.Aiming at the problems described above,this paper has carried out the following work:Firstly,this paper summarizes the principle of distribution network optimization,describes the reason and significance of reactive power optimization in distribution network,introduces several commonly used power flow algorithms,and introduces the establishment of different objective functions for different elements to be optimized in distribution network.Finally,combined with this paper,the distribution station area is selected,Considering the network loss and compensation cost,an objective function considering both security and economy is established.Secondly,this paper selects the convolutional neural network vgg16 network to classify the dynamic load of the distribution station area,adopts the load data of the station area for a whole year,takes the load data of the first 351 days as the training set and the last 14 days as the test set,and successively uses the load curve and sound spectrum for training,and compares the accuracy of the two cases,Finally,the sonogram with higher classification accuracy is selected as the most classification data set among 5 categories of results.Then,the original bacterial foraging algorithm is improved.Firstly,its step size is set as an adaptive step size,which can be changed according to the current bacterial individual.Particle swarm optimization algorithm is introduced in the migration link to eliminate the bacteria of the current optimal solution from the migration link,and a selection link is added in the trend link to prevent the optimization result from crossing the boundary,In the process of replication,the current fitness of individual bacteria is taken as the basis for replication.Finally,the ieee33 distribution network model is optimized,and the simulation results show the limitation of the practical method.Finally,according to the previous classification,four of the five load types need reactive power optimization.The particle swarm optimization algorithm,genetic algorithm and the improved bacterial foraging algorithm are used to optimize these four types respectively.Upon analysis of the simulation results,it can be concluded that the improved bacterial foraging algorithm has a significant effect on increasing the node voltage,reducing the network loss and economic cost,The superiority of the algorithm is proved.