Research On Three-Phase Power Flow And State Estimation Methods For Asymmetrical Distribution Networks

Distribution power flow calculation is the base of distribution network analysis,and is the key of distribution management system. Meanwhile, state estimation is thebase of power flow calculation for distribution networks, and is the source of the data todistribution system analysis, which has an important research value. With thedevelopment of power system, the asymmetry in three-phase parameters or theimbalance in three-phase load is prominent, and asymmetric distribution system isubiquitous. In our country, medium voltage distribution system generally adopts theisolated neutral operation mode, while the structure of three-phase four-wire is used forlow voltage distribution networks. At present, most of power flow researches are basedon grounding system, when the research for ungrounded system is few and inadequate.At the same time, existing three-phase state estimation methods are only applicable tothree-phase three-wire system. A further research of calculating three-phase power flowfor asymmetric distribution networks in ungrounded system and considering thestructure of three-phase four-wire for area load in three-phase state estimation model isstill needed.Therefore, this paper is a research on three-phase power flow and state estimationmethods for asymmetrical distribution networks. Specific contents are as follows:①A three-phase power flow model is proposed for ungrounded distributionnetworks. In a neutral ungrounded system, zero sequence network is open with acharacteristic of zero sequence current being0. Based on this fact, endpoint models,including PQ, PV and balance forms, as well as load models, including constantimpedance and constant power forms, are totally considered. Newton-Raphson methodis used to solve the presented model. Compared with the existing literature model, themodel is without the approximation of zero sequence voltage being0, then reflectsasymmetric current status more accurately for ungrounded system.②A state estimation model is proposed for three-phase four-wire distributionnetworks. Measurement equations are established in the forms of node injection currentand phase voltage amplitude of three-phase or virtual measurement of neutral node, andzero injection current equations of T endpoints are as the equality constraint equations,where the real and imaginary parts of three-phase and neutral voltages are as statevariables. The model adopts weighted least squares method, where a augmented Lagrangian function is as the objective function. The model provides a solution to solvethree-phase four-wire state estimation problems for distribution networks. Consideringthe influence of minute measurement residual, a threshold criterion is introduced, and adetection threshold value of the residual variance is reasonably selected in calculatingthe exponential weight function, which greatly improves the convergence and the speedof the algorithm.③Effectiveness verification and feature analysis of three-phase power flow andstate estimation methods for asymmetric distribution networks are finished.1) Based onthe model in①, the modified IEEE-13bus distribution system is used to contrast theexisting literature model with the presented model, then to verify the applicability andnecessity of the model proposed in this paper. Meantime, the characteristic of zerosequence voltage and the uncontrollability of three-phase power are verified, and therelationship of neutral voltage and zero sequence voltage, also the discrepancy betweenphase-to-ground power and phase-to-neutral power are analyzed;2) Based on the modelin②, the modified IEEE-13bus distribution system is used to verify the effectivenessof the presented method and the necessity of introducing a threshold criterion andreasonably selecting the detection threshold value of residual variance when computingexponential weight function. Concurrently, based on Xiaoshan typically actual area load,the exponential weight function method is constrasted with fixed weight method andestimation method based on weight function. Simulation results verify that thepresented method is robust in identifying bad data, and can automatically exclude theimpact of bad data.