Optimal Power Flow Using PSO Algorithms Based on Artificial Neural Networks

The need to improve the performance of electrical power systems involves solving the Optimal Power Flow (OPF) problem by determining the optimal controllable setpoints, which is the focus of this paper. Specifically, this study addresses the OPF problem by proposing a novel methodology that combines the Particle Swarm Optimization (PSO) algorithm and Artificial Neural Network (ANN) to decrease costs and transmission losses. The primary objective is to minimize generation fuel costs and transmission line losses, taking into account the system constraints. Previous work on PSO optimization has a significant limitation; the selection of cognitive parameters such as acceleration coefficients and inertia weight is often arbitrary, with little to no investigation into optimal OPF solutions, making it difficult to minimize key objective functions, including cost and loss. This study eliminated these limitations by utilizing modified PSO with an iterative approach which is then integrated with ANN training. The ANN improves the parameter exploration accuracy by PSO by determining the optimal parameters for the optimization problem. Most of the earlier works presented results where they could not determine optimal values of the objective function, this study successfully merges PSO and ANN to present optimal solutions. The proposed technique is applied to the IEEE 30-bus and IEEE 14-bus standard systems. In the IEEE 30-bus system, the model achieved a generation fuel cost of 797.25 ${\$}$ /hour and transmission losses of 2.38 MW. In the IEEE 14-bus system, the respective figures were a fuel cost of 829.34 ${\$}$ / hour and losses of 2.79 MW. The results of experiments demonstrate the superiority of this approach compared to other techniques in the recent literature, especially in the minimization of the objective function.