Develop a Hybrid Optimization Technique (DNN - PSO) to Optimal Power Flow

The optimal power flow problem (OPF) is a critical and essential topic in power systems and electrical engineering, as it plays an important part in optimizing the performance and efficiency of these systems, while ensuring their reliability and stability operation within operational constraints and under various conditions, with minimal costs and losses. This research introduces a novel hybrid approach that integrates meta-heuristic optimization with deep learning to address the complex OPF issue. The primary objective functions are to minimize fuel costs of generators and transmission line losses. The proposed methodology is applied on the IEEE 30-bus system. The optimization model employed is the particle swarm optimization (PSO), while the deep learning component utilizes a deep neural network (DNN). One of the key aspects of this research involves identifying the optimal parameters of the meta-heuristic optimization algorithms, which is achieved by broadening the parameter search space within the range of those parameters. This results in a large number of solutions and input-output data, which are subsequently used to train a DNN. The DNN constitutes the second and core stage of optimization, ultimately leading to the attainment of the optimal objective function. By integrating the exploratory capabilities of PSO with the learning and generalization capabilities of the DNN, this approach aims to provide more efficient and effective solutions compared to recent literature, in addition to speed and accuracy. Notably, the results yielded costs of 798.53 $/h and losses of 2.3729 MW.