A Novel Application of Generative Adversarial Networks in Enhancing the Resilience and Efficiency of Integrated Power and Gas Systems

Abstract

In the realm of integrated energy systems (IESs), the complexities and vulnerabilities posed by interdependencies between power and gas infrastructures require robust solutions for resilience and efficiency. This research introduces a pioneering approach employing a generative adversarial network (GAN)–based robust optimization model, specifically designed for the IEEE 118-bus system and the Belgian high-calorific gas network (BHCN). This method synergistically combines deep learning algorithms and robust optimization to adaptively manage and optimize system responses under varying attack scenarios and operational stresses. Our methodology involves the utilization of GANs to simulate realistic attack scenarios that challenge the IES’s resilience. By iteratively training the discriminator to assess the system’s response and the generator to create increasingly complex attack vectors, the model refines its capacity to guide system adaptations, ensuring optimal operational resilience and efficiency. The optimization model was applied to key components of the system, including voltage regulation in the power system and pressure management in the gas network. Through this application, we achieved a reduction in response time by up to 15% and enhanced the voltage stability across the power system, maintaining levels within 1% of desired operational voltages. For the gas network, the model successfully regulated pressures, preventing deviations beyond 5% under nominal conditions, even during peak demand periods. The effectiveness of the GAN-based model in enhancing operational stability and reducing susceptibility to disruptions is underscored by these quantifiable improvements. This innovative approach not only supports the theoretical advancement in energy system optimization but also offers practical insights and tools for system operators seeking to enhance infrastructure resilience and operational efficiency. The integration of AI and operations research in this context presents a transformative potential for future developments in energy system management.