In this paper, a novel improved Stochastic Fractal Search optimization algorithm (ISFSOA) is proposed for finding effective solutions of a complex optimal reactive power dispatch (ORPD) problem with consideration of all constraints in transmission power network. Three different objectives consisting of total power loss (TPL), total voltage deviation (TVD) and voltage stabilization enhancement index are independently optimized by running the proposed ISFSOA and standard Stochastic Fractal Search optimization algorithm (SFSOA). The potential search of the proposed ISFSOA can be highly improved since diffusion process of SFSOA is modified. Compared to SFSOA, the proposed method can explore large search zones and exploit local search zones effectively based on the comparison of solution quality. One standard IEEE 30-bus system with three study cases is employed for testing the proposed method and compared to other so far applied methods. For each study case, the proposed method together with SFSOA are run fifty run and three main results consisting of the best, mean and standard deviation fitness function are compared. The indication is that the proposed method can find more promising solutions for the three cases and its search ability is always more stable than those of SFSOA. The comparison with other methods also give the same evaluation that the proposed method can be superior to almost all compared methods. As a result, it can conclude that the proposed modification is really appropriate for SFSOA in dealing with ORPD problem and the method can be used for other engineering optimization problems.