Evaluación de la Vulnerabilidad del Sistema Eléctrico de Potencia en Tiempo Real usando Tecnología de Medición Sincrofasorial

Abstract

Este trabajo presenta una metodología innovadora para evaluar, en tiempo real, la vulnerabilidad post-contingencia del sistema eléctrico de potencia (SEP). Usando como datos de entrada señales registradas en unidades de medición sincrofasorial (PMU), se determinan indicadores que brindan alerta temprana del riesgo de ocurrencia de colapsos. La evaluación de vulnerabilidad post-contingencia se estructura considerando cuatro tipos de inestabilidades (transitoria, oscilatoria, de frecuencia, y de voltaje de corto plazo) y posibles sobrecargas. La propuesta permite evaluar las dos tareas involucradas en el concepto de vulnerabilidad: i) valoración del estado de seguridad del sistema, y ii) análisis de la tendencia del sistema a cambiar sus condiciones a un estado crítico. La metodología de evaluación se aplica al sistema de prueba de New England. Los resultados resaltan una alta eficacia del esquema integral de evaluación para brindar alerta temprana de riesgos de eventos en cascada del SEP, valorando las dos tareas involucradas en el concepto de vulnerabilidad.

Voltage Stability Margin Index Estimation Using a Hybrid Kernel Extreme Learning Machine Approach

“…The scenarios are built from the demand of the base case, taken from the original information of the system. Then, the demand is changed in every node using hourly demand according to every type of user as indicated in Figure 2 (residential, industrial or commercial) [43]. Every PQ node is randomly assigned a type of user, guaranteeing the same number of PQ users for each type of user.…”

“…The SVM finds this hyperplane using support vectors and margins (defined by the support vectors); this is achieved through solving an optimization problem [42]. SVM belongs to a set of algorithms named Kernel-based methods and the structural risk minimization is used as the optimization principle [43]; also, SVM has statistical robustness and ability to overcome over adjustment problems [44].…”

“…As a solution to the optimal parameter setting, different metaheuristic optimization techniques have been proposed, such as Particle Swarm Optimization [56], Genetic Algorithms [31] and Mean-Variance Mapping Optimization (MVMO), where an objective function (OF) is minimized and corresponded to the Mean Squared Error (MSE) [43]. MVMO has successfully been applied for the Generally, SVR with kernel type RBF has three parameters to optimize: C, γ and ε.…”

“…Additionally, it is a process of local search whose search range is hard to establish. As a solution to the above mentioned problems, an alternative presented in [43] was developed, there an optimization heuristic algorithm known as MVMO is used to determine the best parameters of the AI with kernel. The optimum parameters that must be identified for SVR are ε, C and γ according to equation (2); while the optimum parameters to identify for KELM are C and γ.…”

“…In this case, 10 repetitions are used in the k-folk cross-validation to ensure a robust regression. Based on [43], the search space of the AI parameters are defined in an exponential scale in the following way: C ∈ 2 −5 , 2 15 , γ ∈ 2 −15 , 2 5 , ε ∈ 2 −5 , 2 0 [42,46].…”