Estimação de estado para microrredes no modo de operação ilhada

Abstract

In recent years, the distribution systems have become more complex due to different factors, such as: insertion of conventional and renewable Distributed Generation (DG), installation of Phasor Measurement Units (PMU), utilization of Smart Meters, etc.. In this way, the microgrid concept has emerged as an implementer to accommodate all these changes in electrical systems. Islanded microgrids are more vulnerable to voltage and frequency deviation that increase their complexity of control e operation. This complexity requires the introduction of new functions in the distribution management systems to aid the operator in the decision make process and in the new reality of the distribution networks. Thus, the State Estimation (SE) function becomes essential since it is responsible for providing reliable information about the current system operation point based on available measurements. This dissertation proposes a methodology for SE in islanded microgrid that is able to estimate the complex voltage at all the bus and the frequency of operation in steady state. Frequency measurements and inertia characteristics of the distributed generators were introduced in the proposed formulation aiming to provide a tool with ability of representing more realistically and accurately the microgrid operation. The developed methodology is based on the distributed slack bus power flow to model the absence of the swing bus and the primary control in the islanded operation of the microgrid. This power flow evaluates the network state and the frequency deviations in steady state that are utilized to generate the true values of the measurements used in SE. A microgrid of 38 bus and 5 distributed generators was used to validate the proposed formulation through statistical indices such as: average relative errors between true and estimated values, bias, consistency, quality and risks of the relative errors violate their specified values. The tests results demonstrated that the proposed methodology is more accurate and robust, with regard to bad data, than the conventional state estimator