Noviembre 2, 7 28 29 y Medellín Colombia Medellín Colombia VII Simposio Internacional sobre VII 2013 1 Abstract— This paper presents a novel approach based on a genetic algorithm combined with an artificial neural network and a reduced variable neighborhood search to find the optimal location of distributed generation in electric distribution systems. The objective function consists in minimizing active power losses. The main contribution of the paper consists in the combination of metaheuristic techniques along with artificial intelligence to solve a multi-modal non-convex problem. The use of an artificial neural network avoids the calculation of power flows, while the neighborhood search, applied at the end of each iteration, allows the algorithm to explore a wider search space and eventually, escape from local optimal solutions. The proposed approach was tested on a 13 bus distribution system showing the robustness and applicability of the model. Index Terms —Distributed generation, genetic algorithms, artificial neural networks, neighborhood search. I. NOMENCLATURE The following nomenclature is used throughout the paper. A. Indexes r : Branch index. , ik : Bus indexes. j : Distributed generation index. B. Parameters Di P : Active demand in bus i . Di Q : Reactive demand in bus i . min i V : Minimum and maximum voltage magnitude at bus i . This work was supported by the “S ustainability Program 2013-2014” of the University of Antioquia, Medellín, Colombia. P. A. Narvaez is with “Interconexión Eléctica S.A – ISA” (e-mail: panarvaez@isa.com.co). E. Velilla Hernández and J.M López-Lezama are with GIMEL (Efficient Energy Management Research Group) Department of Electrical Engineering, Faculty of Engineering, University of Antioquia, Street 70 No 52-21 Medellín, Colombia. (e-mail: evelilla@udea.edu.co; lezama@udea.edu.co ). max i V : Maximum voltage magnitude at bus i . min Gj P : Minimum active power limit of DG unit j . max Gj P : Maximum active power limit of DG unit j . min Gj Q : Minimum reactive power limit of DG unit j . max Gj Q : Maximum reactive power limit of DG unit j . max ik S : Maximum apparent power flow in line connecting nodes i, k . nr : Total number of branches. nb : Total number of buses. ik g : Real part of the i,k element of the admittance bus matrix ik b : Imaginary part of the i,k element of the admittance bus matrix. max ik S : Maximum apparent power flow in line i,k . max ngd : Maximum number of DG units to be allocated in a single bus. C. Variables i  : Binary variable that indicates whether there is (1) or there is not (0) DG in bus i . Gj P : Active power supplied by DG unit j . Gj Q : Reactive power supplied by DG unit j . i V : Voltage magnitude at node i .  : Voltage angle. ik S : Apparent power flow in line connecting nodes i, k . ngd i : Number of DG units to be allocated in bus i . II. INT RODUCT ION ISTRIBUTION system planners must guarantee the supply of economical and reliable electricity to customers. With recent advances in small-scale generation technologies, the use of distributed generation (DG) can provide an economical and environmentally friendly solution to meet the load growth in distribution systems. In recent years, the presence of DG in distribution systems has become increasingly common. The reasons for this trend include the unbundling of electricity markets, along with Hybrid Genetic Algorithm for the Optimal Location of Distributed Generation in Distribution Systems P.A Narvaez, Member, IEEE, E. Velilla Hernández, Member, IEEE, and J. M. López-Lezama, Member, IEEE D