Research on cascading failure based on high-order neighbors and residual capacities load redistribution process

Cascading failures caused by overload pose significant threats to critical infrastructure systems, such as server systems, power grids, and network systems. Although previous studies have offered valuable insights into load redistribution strategies to mitigate cascading failures, several critical issues remain underexplored. To address the challenge of inaccurate modeling, this paper integrates both network topology and resource allocation considerations. Using the entropy weight method, the TOPSIS algorithm, and the K-means clustering algorithm, we propose a method for representing the initial load of nodes. Moreover, node load capacity is modeled as a normal distribution to account for the inherent variability in load-bearing capabilities. To resolve the issue of indiscriminate load processing, we introduce a real-time load sorting algorithm that evaluates both node level and load size, prioritizing high-priority loads and reducing system response time. Additionally, we propose a load redistribution algorithm that factors in higher-order neighbors and residual node capacities, thereby optimizing resource utilization and improving system stability. A cascading failure model is also developed to demonstrate the chain reaction of failures caused by overloads. Furthermore, three evaluation metrics – residual load, effective nodes, and waiting time – are defined to comprehensively assess the network performance across multiple dimensions. Extensive experiments conducted on ER networks illustrate the impact of various attack strategies on network performance, validate the effectiveness of the proposed real-time load sorting and load redistribution algorithms, and identify key factors influencing network robustness. This study not only advances the understanding of system stability and robustness but also provides practical recommendations for fault prevention and risk management in complex systems.