194 PRZEGLĄD ELEKTROTECHNICZNY, ISSN 0033-2097, R. 92 NR 8/2016 Edmarcio A. BELATI 1 , Lina P. GARCES 2 , William M. DA ROSA 1 , Igor F. do PRADO 3 and Priscila ROSSONI 1 UFABC (1), UFG (2), UESC (3) doi:10.15199/48.2016.08.53 Distributed Generation Allocation Using the Genetic Algorithm of Chu-Beasley and Sensitivity Abstract. This paper presents a methodology for the allocation of distributed generation (DG) units to minimize active power losses in distribution networks. This methodology is based on the genetic algorithm of Chu-Beasley (GACB) and first-order sensitivity (FOS). To evaluate the different proposals of solution, instead of using Load Flow (LF), a FOS technique was used in order to directly estimate the solution of the LF. The proposed methodology was applied to three distribution systems, containing 34, 70 and 126 buses, respectively. Results obtained for the 34 bus system using GACB and FOS technique were compared with that obtained using GACB but solving the LF via Newton-Raphson (NR) method, showing the computational time gain when the FOS technique is used. For the three systems, the best locations for allocation two DG units are shown and the technical impacts in the network, i.e., active power losses and voltage profiles, are verified. Streszczenie. W artykule opisano metodologię lokalizacji rozproszonych źródeł energii przy kryterium minimalizacji strat mocy czynnej. Metoda bazuje na algorytmie genetycznym Chu-Beasley i czułości pierwszego rzędu. Metodę sprawdzono na przykładzie trzech różnych sieci dystrybucyjnych. Analiza lokalizacji rozproszonych generatorów przy wykorzystaniu algorytmu generycznego Chu-Beasley Keywords: Distributed generation; Sensitivity analysis; Genetic algorithm of Chu-Beasley. Słowa kluczowe: generacja rozproszona, algorytm genmetyczny Chu-Beasley. 1. Introduction The efforts towards the expansion of the participation of alternative energy in the electricity generation are growing in the world. Diversifying the energy matrix by implementing new sources of electric generation is the object and desire of many countries. These new forms of power generation may be connected to distribution and transmission networks. The renewable resources [1], e.g., wind, solar, biomass, among others, are classified as environmentally friendly. These resources can be identified as distributed generation (DG) [2]. The insertion of renewable resources in a power system is increasing, especially in power distribution networks. In general, connecting the power generator near the load brings benefits to the distribution networks. Thus, it is possible to reduce losses and improve voltage profiles in the feeders, allowing postponement of investments in an infrastructure. In specialized literature, there are several papers that discuss this problem from different perspectives. In [3] an algorithm for allocation of DG units is presented. The goal was to reduce power losses in the system and ensure that the voltage profile remains within acceptable levels. This work presents an algorithm based on an Optimal Power Flow (OPF), divided into two phases. In the first phase, the classification of the buses is performed according to loss reduction criterion. The second phase is responsible for allocating and calculating new voltage levels after allocation of the DG units. On the other hand, [4] presents a methodology that uses exhaustive enumeration to calculate the optimal size of the DG units minimizing losses in the distribution system. This proposition is attractive when evaluating systems of small size, but applied to larger systems may lead to greater processing time. In [5], an approach to determine the best location and size of a DG unit using Genetic Algorithm (GA) is presented. In this work, Linear Programming (LP) is used to confirm the optimization results obtained by GA and to investigate the influence in the objective function of the allocation of DG units in different places. This approach was tested in a distribution network in Egypt. In the work proposed in [6], a procedure based on GA and Decision Theory is used. The objective is to establish the best location and the capacity power of a DG unit in systems of medium voltage. This study includes the technical limitations of the system, such as transmission capacity of the feeder, voltage profile and short-circuit currents at various points of the feeder. The objective function seeks to evaluate the best way to the network planning considering wind turbines to meet the load with minimal cost. For allocating DG units in electrical distribution systems, and in order to evaluate the network performance in relation to the active power losses and voltage profile it is necessary to solve the power balance equations with a Load Flow (LF). This evaluation consumes a good portion of the processing time of the algorithms based on metaheuristics to the allocation of DG units. In [7], one of the most efficient methods used for solving the problem of LF for transmission systems, the Newton- Raphson (NR) method, is proposed. Although the NR method is used mostly in transmission systems, it can be applied to distribution systems when sparsity techniques are used, [8]. Finally, [9] presents a first-order sensitivity (FOS) technique applied to distribution networks. This technique has demonstrated good performance in estimating active power losses and will be utilized in this work. Thus, in this work we present a methodology for allocation of DG units based on the genetic algorithm of Chu-Beasley (GACB). The GACB is an improvement in the conventional technique of GA, [10]. The active power loss reduction is used for allocating two DG units in the distribution systems, i.e., the objective function is to minimize the active power losses. To improve the performance of GACB, the FOS technique presented in [9] was incorporated in the simulation process. This technique is used to evaluate the candidate solutions, in a direct way, eliminating the utilization of the LF, making the algorithm faster. Thus, the contribution of this paper is to apply the GACB with the FOS technique in allocation of DG units, providing a quick and efficient method. This paper has the following organization: Section 2 describes the genetic algorithm of Chu-Beasley. Section 3 shows the first-order sensitivity technique. Section 4 presents the results obtained from tests on three systems (34, 70, 126-bus). Finally, some concluding remarks are made in Section 5.