Research On State Estimation Methods For The New Characteristics Of Active Distribution Network

With the increasing permeability of distributed power in distribution network and the improvement of measurement system,distribution network is gradually changing from passive system to active system.This brings the following problems to the real-time monitoring of distribution networks:First,there are many zero injection nodes in the distribution network.The traditional state estimation method usually uses zero virtual measurement to estimate zero injection nodes,and gives them a higher weight value to ensure the estimation accuracy of zero injection nodes.However,this will make the Jacobian matrix and gain matrix conditions too large,resulting in the algorithm extremely sensitive to errors or even unconvergence,greatly increasing the calculation time of the algorithm and reducing the robustness of the algorithm.Second,distributed power supply in distribution network brings a lot of uncertainty for distribution network state estimation due to its intermittent and random output.In the distribution network where real-time measurement is scarce,these uncertainties will reduce the estimation accuracy of the state estimation algorithm for the distributed power node state,thus affecting the state estimation to the whole system real-time situation perception level.Third,the measurement system of active distribution network is not complete and the real-time measurement in the system is few.The smart meter,which is the main source of the measurement data,has the problems of long time delay non-synchronization and bad data because of the incomplete communication facilities.Using these data directly for state estimation will result in the decrease of the state perception level and the decrease of the state estimation accuracy.This Thesis has proposed the followingsolutions to the above three problems in active distribution network:First,for processing of the virtual measurements of zero injection buses,the Thesis proposes to use node injection currents in rectangular coordinate form as state variables.Because the node injection currents of zero injection nodes are strictly limited to 0,simplified state variables can be used to estimate the state,and avoid a series of problems caused by the high weightings of the zero injection buses.Second,based on the above method,this Thesis analyzes the historical output data of distributed generators by k-means clustering,and divides it into different output modes.In the process of state estimation,the results of Pre-state estimation by the system measurement discriminates the output mode,constructs pseudo-measure according to the output mode data and the system measurement data according to certain weight superposition and adds it to the secondary state estimation process to improve the estimation accuracy of distributed power nodes.Third,the smart meter measurement data is interpolated to the same time by using the characteristics of the intelligent meter measurement data,and the system is partitioned according to the remote terminal unit(RTU)configuration position.The off-line smart meter data of different partitions are integrated into a virtual power flow measurement,and a virtual power flow measurement vector is formed according to sequence.Then the multi-dimensional k-means clustering algorithm is used to cluster the vector into different virtual power flow clustering,and the intelligent data under the same cluster is averaged and stored.According to the real-time RTU measurement data of the system to determine which cluster mode the system is currently in.Then we use the stored smart meter data in this mode to approximate instead of the actual read smart meter measurement data for state estimation calculation.The method can reduce the situation awareness adverse effects when smart meter measurement data is delayed,out of sync or including bad data.The above method is validated via MATLAB simulations.