Aproximação funcional dos limites de estabilidade/segurança de tensão por redes neurais artificiais com seleção de atributos em problemas de fluxo de potência ótimo

Abstract

The object of this research is the proposal of a model for Voltage-Stability/Security on strained Optimal Power Flow (VSSC-OPF) problems applied to mid-term energy procurement auction systems in electricity markets. In the existing literature, many of the operational security representations in Optimal Power Flow (OPF) problems are too conservative, first or second-order approximations, with high computational costs and/or even ineffective. In this context, a practical VSSC-OPF problem whose voltage stability/security representation is designated by a functional approximation characterized by Artificial Neural Networks (ANNs) with feature selection is proposed in this research project, making it possible the use of Nonlinear Programming (NLP) resolution methodologies. This research project is methodologically founded on determining voltage stability/security limits associated with power system loading margins accounting for saddle-node and limit induced bifurcations and technical-operational limits on transmission lines; in the inclusion of feature selection for the functional approximation of such limits by ANN; the incorporation of this functional approximation into the VSSC-OPF problem and the resolution of the resulting NLP problem. To accomplish this, the main concepts on power system operational voltage stability/security, mid-term energy procurement auction systems in electricity markets and resolution of NLP problems will be outlined by means of a background review on these subjects. The ANN training and validation processes will be realized on Jupyter Notebook, and the proposed model will be implemented in the AMPL modeling language for mathematical programs and solved by LOQO solver. The functional approximation of power system voltage stability/security limits and the proposed VSSC-OPF model will be validated by numerical tests with test systems commonly deployed for this task in the correlate literature.