Estimação de estados multi-área de sistemas de distribuição de energia elétrica com suporte de redes neurais artificiais considerando variações topológicas e erros grosseiros

Abstract

This work presents a methodology for State Estimation (SE) of Electrical Distribution Systems (EDS) based on Artificial Neural Networks (ANN), considering the division of EDS into different areas, and performing the estimation using separate ANN for each area, characterizing Multi-Area State Estimation (MASE). The estimations are performed using few electrical measurements obtained from Phasor Measurement Units or Smart Meters. The ANN are trained considering the measurement patterns and their related true states of the electrical network, becoming able to estimate state for future measurements not observed during the trainning process. The EDS modeling considers topology changes, which are identified by monitoring the network switches status, and also encompass the input variables of the ANN. Furthermore, the methodology also includes methods for detection and correction of gross erros on the electrical measurements and monitored switch status, as to ensure the estimatitor robustness against gross errors on the measurements. Lastly, the estimations performed by each area’s estimators are sent to a central agent responsible for synchronizing the estimated electrical variables and composing the estimated state of the EDS as a whole. Tests were performed considering two networks frequently used on the literature, and also on a real brazilian network computationally simulated. The implementations were performed using the MATLAB software, and the obtained results showcase the methodology effectiveness for state estimation considering load variations, topology changes, and gross errors detected during monitoring.