Article Preview
TopIntroduction
For the last decade, a lot of studies in the field of electrical engineering have been dedicated to enhancing the electric power networks in terms of power quality and reliability (Jordehi, 2015; Chattopadhyay et al., 2016). The first study that aimed to improve the distribution network performance was done in 1975 when Merlin and Back introduced the idea of the distribution network reconfiguration for the active power loss reduction (Merlin & Back, 1975). The distribution network reconfiguration can be defined as a process that handles the open /close status of sectionalizing switches and tie-switches in an electrical distribution system, in order to optimize different criteria (active power loss, reliability, voltage profile, etc.).
In the literature, the distribution network reconfiguration is commonly solved for the minimization of active power loss and/or voltage deviations (Nguyen & Truong, 2015; Teimourzadeh & Zare, 2014; Kumar & Jayabarathi, 2012). Several authors have used well known heuristics to solve the reconfiguration problem, namely branch and bound (Morton & Mareels, 2000), branch exchange (Ababei & Kavasseri, 2011), another heuristic (Aman et al., 2014) and meta-heuristic rules (Mirhoseini et al., 2015). However, these optimization techniques could not converge to the global optimal solution because they depended only on the operation sequence of switches or on the network initial status.
The network reconfiguration can be considered as a complex, nonlinear and combinatorial problem that requires robust optimization techniques for its resolution. Recently, various studies have proposed multi-objective optimization algorithms based on evolutionary methods to solve the network reconfiguration problem, e.g. the genetic algorithms (Shamsudin et al., 2014), the binary group search optimization (Teimourzadeh & Zare, 2014); the bacterial foraging optimization problem (Kumar & Jayabarathi, 2012) and the enhanced genetic algorithm (Duan et al., 2015). Most of these techniques transform the multi-objective problem into a mono-objective optimization using weighted summation or converting the objective functions into the same measurement unit. Therefore, the multi-objective reconfiguration problem is solved improperly because the objective functions are not optimized equally.
In this context, evolutionary algorithms based on the Pareto optimality approach are proposed by several reconfiguration studies, including power loss reduction and reliability improvement, e.g. the micro-genetic (Mendoza et al., 2009), the modified cuckoo search algorithm (Kavousi-Fard et al., 2014) and the NSGAII (Vitorino et al., 2015). These evolutionary algorithms have proved their effectiveness, providing a set of optimal Pareto solutions in a single run instead of the traditional heuristic techniques that have to perform a series of separate runs.
The main goal of this article is to put forward a novel method based on evolutionary computation to efficiently solve the network reconfiguration problem, thus minimizing active power loss (PL) and improving reliability indices (System Average Interruption Frequency Index (SAIFI), System Average Interruption Duration Index (SAIDI) and Energy Not Supplied (ENS)). The considered constraints are associated with the nodal voltage drops, the branches overloads and violations of the radial topology of the distribution network.
The proposed method is based on Strength Pareto Evolutionary Algorithm 2 (SPEA2), which is an improved version of SPEA developed by the scholar Zitzler in his report (Zitzler et al., 2001). SPEA2 uses the concept of non-dominance and Pareto optimality to solve the multi-objective optimization problems. This technique is developed with some new features, like the use of an external archive to preserve the best non-dominated solutions through iterations and the application of a truncation operator is applied to keep the diversity and well spread of Pareto optimal solutions.