Distribution networks are always exposed to many blackouts due to their sizes. Reducing downtime, reducing energy not supplied, increasing reliability, and thus, increasing customer satisfaction are important goals of distribution companies. The performance of power systems is affected by many factors such as various faults, network maneuvers, and unwanted equipment outages. Protective switches can reduce subscribers' downtime, energy not supplied, and associated costs by isolating faulty network points. On the other hand, due to the high costs of purchasing, installing, and maintaining protective equipment such as reclosers and sectionalizers, distribution companies seek to minimize these costs. Therefore, determining the minimum number of reclosers and sectionalizer switches required for reliable protection of the network and their optimal installation location is one of the basic needs of distribution network operators. The main purpose of this study is to determine the number and optimal location of recloser and sectionalizer switches in a distribution network using an improved genetic algorithm. The proposed objective function, which has been formulated to improve four reliability indices, consists of four terms, i.e., System Average Interruption Duration Index (SAIDI), Energy Not Supplied (ENS), Average Interruption Frequency Index (MAIFI), and System Average Interruption Frequency Index (SAIFI). Also, the limitations of operation are related to the number and installation place of protective devices. Considering the radiality of the feeders studied in this study, in order to implement the mentioned limitations, first all the branches and nodes under each branch in the feeder graph should be obtained. The nodes and branches under each branch in the network graph are the nodes and branches that will become isolated from the reference bus and lose their power if the branch is removed. Therefore, a practical method based on graph theory is presented to calculate all the nodes and branches under each branch.
The genetic algorithm is a very practical algorithm  in solving complex and non-convex optimization problems, but in the optimization process, some defective genes may enter the next generation and the algorithm may face local optima. Therefore, the study adopts the elitism mechanism, so genes with higher competence enter the next generation before entering the crossover and mutation process. The proposed method is tested on one of the distribution feeders in Bojnourd city called Nader and its results are compared with other evolutionary methods such as the genetic algorithm, shuffled frog leaping, and particle swarm optimization. The results reveal that the proposed method outperforms other evolutionary algorithms