Turk J Elec Eng & Comp Sci, Vol.19, No.6, 2011, c  T ¨ UB ˙ ITAK doi:10.3906/elk-1004-2 Using learning automata for multi-objective generation dispatch considering cost, voltage stability and power losses Arif KARAKAS ¸ 1,∗ , Celal KOCATEPE 1 , Fangxing LI 2 1 Department of Electrical Engineering, Yıldız Technical University, ˙ Istanbul, 34349, TURKEY e-mails: akarakas@yildiz.edu.tr, kocatepe@yildiz.edu.tr 2 Department of Electrical Engineering and Computer Science, University of Tennessee, Knoxville, TN 37996, USA e-mail: fli6@utk.edu Received: 01.04.2010 Abstract The economical and secure operation of power systems has significant importance. Due to technical limitations, the best economicaloperation point is not always the desired operatingpoint for system stability or power losses. In this study, first, the most economical operating point is obtained by solving the non- linear, network-constrained economic dispatch problem using a genetic algorithm. Then, the system voltage stability is analyzed to compare the different possible operating points using V-Q sensitivity analysis. The power losses, obtained for various operating points, are considered the third objective function. Finally, these 3 aspects of cost, voltage stability, and power losses are combined, using the learning automata technique, to achieve a multi-objective optimization solution. The methodology was implemented in MATLAB 7.8 and applied to the IEEE 30-bus test system. The same technique of learning automata may be applied in the future to similar problems that need multi-objective consideration. Key Words: Economic dispatch, voltage stability, genetic algorithm, learning automata 1. Introduction The common formulation of economic dispatch (ED) is to find the optimal generation cost, subject to a number of equality and inequality constraints. For instance, ED needs to consider the generator’s minimum and maximum capacity limits. Also, the transmission network constraints and losses must be considered in order to perform the economic dispatch accurately because power plants are generally not located near load centers. The consideration of these constraints, as well as the quadratic cost function of generators, typically makes ED a non-linear programming problem. * Corresponding author: Department of Electrical Engineering, Yıldız Technical University, ˙ Istanbul, 34349, TURKEY 913