Research On Distribution Network Reconfiguration Containing Distributed Generation

As an important part of the power grid,the power distribution system's safe and reliable operation is crucial to production and life.At present,the degree of automation of distribution networks in our country is relatively low,and a large amount of network losses are generated in the distribution network.Therefore,in order to avoid energy waste and to facilitate smart distribution network construction better and faster,it is necessary to optimize the distribution network.The distribution network reconfiguration for a certain period of time needs the load data of the period as a basis,so accurate load prediction is needed.At the same time,Distributed Generation(DG)is an effective complement to the current large-scale centralized power supply.Because it has the advantages of saving investment in power transmission,improving the reliability of power supply,and reducing environmental pollution,it has been increasingly into the power grid,this thesis has carried out a study on the reconstruction of distribution networks containing distributed generation based on short-term load forecasting.The main tasks are as follows:This thesis first improves the currently widely used Least Squares Support Vector Machine(LS-SVM)load forecasting model to form an improved least squares support vector machine(PSOBC-LSSVM)forecasting model based on similarity day and bacterial chemotaxis improved particle swarm algorithm.When training samples are selected,the model is trained by selecting historical date data similar to the prediction date conditions.At the same time,Bacterial Chemotactic Particle Swarm Optimization algorithm is used to guide the automatic search of the key parameters of LS-SVM,making the prediction results more accurate,and the prediction result provides a data support module for the reconstruction below.Secondly,the multi-period reconstruction optimization of power distribution system based on predictive load is studied.This thesis expounds the method of time division based on the dual goals of network loss and system stability,and finally determines the optimal reconstruction time for multi-period reconstruction.In the reconstruction,the dynamic two-group particle swarm(DBPSO)was used as the solving algorithm.All the particles were divided into two groups.Each group used different weight coefficients to perform global and local search respectively.With the increase of the number of iterations,the number of two groups of particle swarms dynamically changes,and the disadvantages of particle swarm optimization(PSO)can be overcome to achieve a good search.The DBPSO algorithm is used to reconstruct the test system in the divided period to obtain the optimal reconstruction scheme on the day of prediction.Finally,the reconstruction of the distribution network with DG based on predictive load is studied.The main characteristics of the distributed power supply are introduced,and the power flow calculation models of several typical distributed power sources are analyzed in detail.Then these mathematical models are substituted into the hierarchical-based forward-back generational load flow calculation method,and the influence of DG-connected grid on system reconstruction is studied.Using the above-mentioned multi-period reconstruction method for distribution networks,the impact of DG grid-connection on distribution network reconfiguration is analyzed for the time-division scheme of forecasting days.