Estimação estática de estados em sistemas trifásicos desequilibrados: uma abordagem considerando detecção e identificação de erros grosseiros

Abstract

This work presents a decentralised three phase approach for the Static Stet Estimation in power distribution systems. In this thesis, the state estimation is formulated as an optimisation problem with non linear constraints with the objective of minimising the sum of the quadratic dierences among the PMU (Phasor Measurement Units) measured values and their corresponding estimated ones. The measured phasors are branch currents directly connected to the measurement units and voltages in buses in which PMUs are installed. The phasor quantities are treated in rectangular coordinates, representing the phasor obtained by a PMU. For the non-monitored uses, their active and reactive power equations are handled as inequality constraints being the upper and lower limits determined by previous estimated loads obtained for a previous time interval, ‘t-1’. These values can be expanded between a minimum power demand until a maximum value, allowing the system state to be determined for any instant, tracking the time-varying load curve of the system. In order to to solve the optimisation problem, the Safety Barrier Interior Point Method was modified in such a way that the non monitored loads, for each phase, are the closest to their respective limits but never reaching their barrier values and, if necessary, these limits are relaxed during the iterative process, representing an important contribution of the work. Moreover, the distribution system is divided into subsystems considering a PMU allocation procedure in which each lateral feeder have their system states estimated individually, allowing a decentralised approach in which the computational time can be reduced using parallel processing techniques. Based on the developed state estimation method, a novel methodology is proposed for PMUs bad data detection, identification and correction. The detection procedure is based on the analysis of the objective function value of each subsystem, in which each time interval t being the value compared with a corresponding threshold value which is determined by the Monte Carlo method assuming uncertainties with respect to the loads and aleatory errors inherent to the instrumentation. The bad data identification step is performed based on the determination of the largest quadratic residual together with a new index which evaluates the variations between dierent time instants based on a statistical approach. Finally, a correction step is proposed based on the measured value at time interval ‘t-1’ which is corrected based on the load variations between the time intervals t and ‘t-1’. Computational simulations are carried out using the 33-bus ad 70-bus test systems to validate he proposed method and highlight their contributions