Alocação robusta de chaves para maximização da confiabilidade em redes de distribuição de energia elétrica

Abstract

The predictive reliability analysis of distribution networks aims to estimate duration and frequency indices related to energy supply interruption based on the following parameters: network topology, protection system response, restoration strategy, and reliability data of the equipment (failure rates and repair times). The reliability data of the equipment are subject to uncertainties due to the following factors: sample variability of historical data and realization of the stochastic processes associated with the number of failures for a finite time period. The uncertainties in the input data of the predictive reliability model are disseminated to the estimated reliability indices. Consequently, the estimated reliability indices also are subject to uncertainties and they cannot be characterized only by expected values. The uncertainties in the estimated reliability indices for distribution networks can be quantified by defining confidence intervals. The perception of the variability in the estimated reliability indices introduces uncertainties in the planning strategies oriented to reliability improvement, for example: switches placement, reconfiguration, protection device installation, etc. If these uncertainties are ignored, the benefits obtained with the network reinforcement alternatives can be cancelled due to indices variations around their expected values. For example, the indices variations after the network reinforcement addition can make them worse than their corresponding values for the network configuration without reinforcement. In this way, the main objectives of this dissertation are: to carry out a comparative study about the approaches used to evaluate confidence intervals for the power distribution reliability indices and to perform a robust switch placement in distribution networks considering uncertainties in the reliability data. The techniques considered in the uncertainty propagation studies were: Fuzzy Sets, Interval Arithmetic, Bootstrap, Cumulants e Monte Carlo Simulation (MCS). The allocation robustness is associated with the fact of that the reliability indices are not degraded by the uncertainties in the reliability data after the switches installation. This robust allocation is generated through the maximization of the SAIDI (System Average Interruption Duration Index) probability be lower or equal than a maximum value. This objective function was maximized through the combination of the following techniques: Particle Swarm Optimization and Cumulants method to model uncertainties. The tests results demonstrated that the switches placement obtained with the proposed model made the SAIDI more robust with regard to uncertainties in failure rates and repair times. Furthermore, it was demonstrated that traditional paradigm of minimizing the SAIDI expected value results in poor quality solutions, since the reliability data uncertainties cancel the gains obtained with the optimal switches placement. Additionally, the uncertainty propagation analysis demonstrated that all methods of uncertainty propagation have acceptable accuracy. Nevertheless, the confidence intervals calculated by the Bootstrap and Cumulants are more near the intervals calculated by the MCS (the benchmarking method) than those evaluated by the Fuzzy Sets e Interval Arithmetic.