Byzantine-Resilient Distributed P2P Energy Trading via Spatial-Temporal Anomaly Detection

TL;DR

A fully distributed P2P energy trading model is developed by accounting for the high-fidelity physical network constraints by leveraging the tensor learning method, and theoretical conditions for guaranteeing optimality and convergence of the distributed P2P energy trading problem with anomaly detection mechanisms are derived.

Abstract

Distributed peer-to-peer (P2P) energy trading mandates an escalating coupling between the physical power network and communication network, necessitating high-frequency sharing of real-time data among prosumers. However, this data-sharing scheme renders the system vulnerable to various malicious behaviors, as Byzantine agents can initiate cyberattacks by injecting sophisticated false data. To better investigate the impacts of malicious Byzantine faults, this paper develops a fully distributed P2P energy trading model by accounting for the high-fidelity physical network constraints. To further detect Byzantine faults and mitigate their impacts on distributed P2P energy trading problem, we propose an online spatial-temporal anomaly detection approach by leveraging the tensor learning method, which is informed by the domain knowledge to enable awesome detection performance. Moreover, to enhance its computational efficiency, we further develop closed-form solutions for the proposed detection approach. Subsequently, we derive theoretical conditions for guaranteeing optimality and convergence of the distributed P2P energy trading problem with anomaly detection mechanisms. Results from numerical simulations validate the effectiveness, optimality, and scalability of the proposed approach.

Figures (12)

• Figure 1: False data injections during distributed optimization.
• Figure 2: Online spatial-temporal anomaly detection via physics-informed tensor learning.
• Figure 3: Scheme of physics-informed Tucker decomposition.
• Figure 4: IEEE 15-bus systems with Byzantine faults.
• Figure 5: Convergence curves without Byzantine faults: (a) primal and dual residuals, (b) evolution of the traded energy.