Modified Genetic Algorithm in state space pruning for power system reliability evaluation and its parameter determination

Genetic Algorithm (GA) is emerging as a popular tool in the optimization problems of power systems. In reliability indices calculation and adequacy assessment, methods have been previously developed to use GA as the sampling tool. One of the techniques developed is to use GA as the state space pruning tool in order to truncate the state space before calculating the reliability indices. This means to generate a pruned state space, in which the density of failure states is much higher than the original state space, and then Monte Carlo simulation (MCS) is used as the final tool to assess the state space and have the reliability indices calculated. GA is used as the state space pruning tool to remove as many success states as possible, with overall computational efficiency in the residual space better than only using Monte Carlo simulation over the entire. The GA has selection, crossover, and mutation operations, with associated parameters controlling every step. The decision of the parameters and the stopping criterion does not have obvious rules. This paper presents a modified GA as the state space pruning tool, with higher efficiency and controllable stopping criterion as well as parameter selection. The modified GA has better efficiency than previous methods, and it is easier to have its parameters selected. This method is tested using the IEEE Reliability Test System (RTS 79 and MRTS), and is compared with original GA-MCS method.