Article Preview
TopIntroduction
The load flow (LF) problem, in current power system outlook, has become quite a serious issue. Researchers are in continuous quest of improving the power flow analysis as it is highly essential for desired power system scheduling and operation (Duman, Güvenç, Sönmez, & Yörükeren, 2012). The LF can be considered as an elementary tool that helps in identifying the secure states within the system for economical operation. The importance of LF lies in the fact that it forms the backbone in the assessment of the optimum state of operation of an electric network which can be achieved by the optimization of a fitness function representing the load flow objective maintaining the viable constraints. Some out of the numerous approaches are Newton Raphson (NR), Gauss Seidel (GS), and Fast Decoupled Loadflow method (FDLF), etc. (Scott, 1974), that are frequently used to solve nonlinear equations. Due to robustness and fast convergence characteristic, NR is the most popular amongst all the aforesaid techniques. Hence the load flow analysis can be described as an eminently non-extensive and multi-conditional optimization issue (Abou, Abido, & Spea, 2010).
Most of the classical approaches implied for solving the mentioned problem suffer from a few limitations. Some of the limitations are local optima convergence, unsuitable for integer and binary problems, differentiability along with continuity, and in providing solution to systems with large R/X ratios (Stott & Alsac, 1974; Frank, Steponavice & Rebennack, 2012; AlRashidi & El-Hawary, 2009; Bijwe & Kelapure, 2003; Amerongan,1989). Based on classical avenues, load flow algorithms for critical or ill-conditioned power systems have been further developed (Iwamoto & Tamura, 1981; Ajjarapu & Christy, 1992). The problem involved in the conventional approaches is that their performance is highly susceptible to the R/X ratio and their performance gets deteriorated with an increase of R/X ratio, system loadability, complexity and nonlinearity.
Thus, to overcome these problems and have a better solution of load flow, numerous meta-heuristic optimization techniques have been addressed. A few of the techniques available in research literature are genetic algorithm (GA) by Wong et al., particle swarm optimization (PSO), differential evolution (DE), real coded genetic algorithm (RCGA), hybrid differential evolution algorithm (HDE) etc (Wong, Li, & Law, 1997; Karami & Mohammadi, 2008; Fukuyama & Yoshida, 2001; Acharjee & Goswami, 2010; Udatha & Reddy, 2014; Lampinen, 2002; Rout, Swain & Barisal, 2011; Sakr,, EL-Sehiemy & Azmy, 2017). Grey wolf optimization (GWO) introduced by Mirjalili et al. is a powerful meta-heuristic evolutionary optimization technique, based on the social hierarchy and hunting behavior of grey wolves. The technique has been has been able to provide successful result for several power system optimization issues (Bhattacharya & Chattopadhyay, 2010; Herbadji, Slimani & Bouktir, 2013). The biogeography based optimization (BBO) technique (Simon, 2008) is a biogeography inspired evolutionary approach and has proved to be a successful technique in solving the economic load dispatch (ELD) (Bhattacharya, A.,Chattopadhyay, 2010) and optimal power flow (OPF) problems (Herbadji, Slimani & Bouktir, 2013). BBO optimizes a function by improving candidate solutions stochastically and iteratively with respect to some specified measure of quality, or fitness function (Simon, 2008). Based on the fowling habits of humpback whales, whale optimization algorithm (WOA) is a new nature-inspired meta-heuristic optimization algorithm that has been successfully employed to solve the OPF (Bhesdadiya, Parmar, Trivedi, Jangir, Bhoye & Jangir, 2016) and ELD (Touma, 2016) problems.
Load flow analysis carried out by most of the proposed algorithms addressed in the literature so far, have shown improved results. However, most of the above-mentioned population-based techniques suffer from a much heavier computational burden in large-scale problems. Also, when structure intricacy rises with the increment in number of new variables, most of the aforesaid methods are unable to probe the search space adequately.