Application and Performance Analysis of Deep Learning Models in Power Dispatching Automation

Abstract

Amidst the swift advancement of smart grid technology, traditional power dispatching methods have become inadequate in addressing escalating power needs and intricate system management prerequisites. By incorporating a deep learning model, we have refined these methods, facilitating data-driven dispatching decisions and optimizing power resource allocation and dispatching efficiency. Our experimental outcomes reveal that the Long Short-Term Memory network (LSTM) excels in handling intricate time series data, boasting superior accuracy and convergence rates compared to the Recurrent Neural Network (RNN) and Convolutional Neural Network (CNN). Detailed performance evaluations confirm LSTM’s proficiency in capturing long-term dependencies and processing time series traits inherent in power dispatching data. Furthermore, a 10-fold cross-validation underscores the LSTM model’s stability and generalizability. In essence, this study concludes that in the realm of power dispatching automation, the LSTM deep learning model demonstrates remarkable effectiveness and holds vast potential, anticipating crucial support for the reliable operation and optimal dispatching of the electrical power system (EPS).