Deep-Learning-Based Approach for Accelerated Economic Dispatch in Hierarchical Distributed Power Systems With Internet of Things

With the increasing scale of the Internet of things (IoT) in power systems, the conventional centralized economic dispatch (ED) approach fails to satisfy the demands for computing large-scale power systems. Thus, this study introduces the hierarchical distributed optimization method into the power system, aiming to simplify the difficulty of solving large-scale power systems through hierarchical scheduling. To overcome the shortcomings of slow computation speed and poor optimization performance of existing optimization algorithms, this study proposes a hierarchical distributed African vulture acceleration optimization method based on multi-channel self-attention network (MCSAN-HDAVAO). MCSAN-HDAVAO combines deep neural networks with African vulture optimizers to accelerate the computation speed of power systems and improve ED efficiency. The MCSAN-HDAVAO fills the gap in the application of deep learning accelerated optimization algorithm technology to IoT-enabled power systems. In this study, the performance of MCSAN-HDAVAO is evaluated by 9 benchmark functions. In addition, this study shows through simulations in IEEE 118 and 1790 systems that the proposed MCSAN-HDAVAO can reduce the generation cost by obtaining a more optimal scheduling scheme in a shorter period of time.