Research On State Estimation Of Active Distribution Network With Distributed Photovoltaic System

With the rapid development of solar photovoltaic technology,more and more solar photovoltaic power supply(PV)access to the distribution network,making the traditional single power distribution system gradually transformed into active distribution network(ADN)with multi-power supply.Compared to the traditional distribution network,coordination and control technology of ADN will become more complex.Coordination and optimization control of ADN is based on intelligent measurement and communication technology.In order to make the operation of power grid efficient,reliable,safe and stable,the summary of the massive data of ADN and the efficient and accurate processing have become the core link,which puts forward new requirements for the state estimation method which is suitable for the development of ADN.Therefore,state estimation method of ADN with distributed PV is studied in this thesis.The specific research includes basic algorithm of state estimation,photovoltaic power generation state estimation model and output prediction method,optimization method of measurement device and distributed state estimation method.Firstly,the basic algorithm of distribution network state estimation is studied combined with the characteristics of ADN.The input and output model of the state estimation is established,which analyzes the mathematical expression of measurement function,the network parameters and the topology information and the state estimator based on the weighted least squares method.The mathematical functions of the measurement method based on the node voltage method and the tributary current method are deduced respectively,and the concrete steps and processes of the state estimation are formed.Secondly,the photovoltaic power generation state estimation model and the solar photovoltaic output prediction method are studied considering the access problem of solar PV.The measurement and non-measurement models of PV power generation nodes and their processing methods are introduced,and based on the analysis of the main factors influencing the photovoltaic output,methods for gray correlation analysis of gray correlation degrees for sorted samples of different time spans and the BP neural network training method based on genetic algorithm optimization are proposed.On this basis,the solar photovoltaic output prediction method is formed and the prediction results show that the proposed method has certain accuracy.Then,considering the observability of state estimation,an optimal allocation method of ADN measurement device is proposed.The concept of power system observability is introduced,and a power system observability analysis method based on the observability of tributary states is proposed.The advantages of the phasor measurement unit(PMU)applied in the distribution network are described,and a method of partitioning a distribution network with PMUs as boundary nodes is proposed.A two-layer optimization model of the measurement device configuration is established.Optimized configuration of PMUs is carried out in the upper layer,taking the computational burden of each subsystem as the goal,and the lower layer is based on the upper layer,in which the traditional measurement device is optimized,taking the number of measure devices and economical efficiency as the goal and the observability of subsystems as constraints.The optimal configuration scheme is selected by the evaluation of the calculation precision of the state estimation and the number of traditional measurement configurations.The correctness and validity of the proposed method are verified by an example.At last,a distributed state estimation method for ADN with PV is proposed in order to improve the accuracy and convergence of state estimation.A partition method is proposed considering the PMU configuration,at which the system is decoupled and the reference node is based.Considering the uncertainty of the prediction of PV output,a collinearity test method based on the normalized Lagrangian multiplier is proposed.The method of distributed state estimation is proposed,in which sub-partition take full advantage of the original state estimation method and consider the impact from coordination side as pseudo-measurement,while the Jacobian-free Newton-GMRES(JFNG)method is used to solve the nonlinear equations in the coordination side.The accuracy and superiority of the method are verified through the simulation test.