Energy management of networked energy hub considering risk assessment and cyber security: A deep reinforcement learning approach

Abstract

This research presents a comprehensive analysis of data-driven energy management within the framework of a networked energy hub (NEH), focusing on three objective functions that account for various uncertainties, alongside risk assessment and cybersecurity considerations. The primary objectives of the initial phase encompass the optimal operation of the NEH, which entails maximizing subscriber engagement through integrated demand response initiatives that respond to wholesale and retail market price signals, thereby altering subscriber consumption behaviors, and enhancing the operational efficiency of energy storage systems (ESS) to mitigate operational expenses. The subsequent phase aims to minimize environmental pollution costs, while the final phase is dedicated to evaluating risk costs and conducting a cybersecurity assessment. The model put forward, which utilizes the K-means clustering methodology alongside a probabilistic framework for power generation sources, ESS units, and loads, is articulated through the application of time sequence matrices, auto-correlation matrices, and cross-correlation matrices. This model is constructed using the deep reinforcement learning algorithm, with the soft actor-critic architecture employed to ascertain optimal control strategies. This investigation has been assessed across three scenarios. The first scenario shows the optimal operation of NEH, the second scenario is the same as the first scenario plus IDR performance, and the third scenario is the same as the second scenario with the optimal participation of ESSs. The outcomes show a 15.14 % reduction in total operation costs in the second scenario and 19.49 % in the third scenario.