This paper presents a modified water cycle algorithm (WCA) adapted to the problem of finding the location and size of distributed generation (DG). Power losses minimization was used as an objective function to compare the proposed algorithm with particle swarm optimization (PSO), the bat- inspired Algorithm (BA), and harmony search (HS). The test scenarios consisted of locating five to seven generators with a maximum real and reactive power in the 33-node and 69-node radial distribution networks. The experiment was designed to start iterations from the same initial population to identify the algorithms' performance when searching for the best solutions. The results demonstrate that the modified WCA found the minimum power losses after locating and sizing distributed generators for most of the test scenarios. The algorithm converged quickly to the best solution and the solutions for all repetitions tested were close to the best for each case simulated.