A Deep Learning-Driven Bidirectional Power Dispatch Optimization Framework for Smart Grids Using IoT Sensing Data

Abstract

With the development of smart grid and demand response technology, the optimal dispatch of power system has become the key to improve the efficiency and stability of power grid operation. This paper proposes a supply-demand collaborative optimization model based on Internet of Things (IoT) technology and deep learning algorithm, which combines generator scheduling and demand response strategy to achieve dynamic balance and load management of power grid. We propose a supply and demand collaborative optimization model based on Internet of Things (IoT) technology and deep learning algorithms, and build a smart grid bidirectional power dispatch optimization framework with the help of long short-term memory network (LSTM) and multi-layer perceptron (MLP). The experiment uses a data set covering 5,000 users and simulating the operation of the entire [specific fictional city name] power grid. The data set contains information such as power demand, power generation data, and environmental factors in the past 3 years. The results show that compared with the traditional dispatch method, the load forecast error is reduced by 8%, the system operating cost is reduced by 15% during peak hours (06:00-12:00), and by 12% during non-peak hours, while carbon dioxide emissions are reduced by 8%. By real-time monitoring and adjustment of power demand and generator response time, user demand and generator output are flexibly adjusted according to changes in electricity prices to achieve optimal dispatch of the power system. This study provides a new perspective and practical framework for smart grid supply and demand optimization, and has high theoretical value and application potential.