Research on the Optimization Algorithm of Computing Power Network Resource Allocation

This paper proposes a dynamic resource allocation system based on the integration of Particle Swarm Optimization (PSO) and Soft Actor-Critic (SAC) deep reinforcement learning. By establishing a multi-dimensional network topology model and integrating a real-time load prediction mechanism and protection strategy, efficient resource scheduling is achieved in complex network environments. The system uses a hybrid scoring mechanism to dynamically select optimization algorithms and combines a feedback-driven parameter adjustment strategy to significantly improve global load balancing and Quality of Service (QoS). Through simulations on power communication network topologies and the selection of relevant evaluation indicators, comparisons with existing algorithms show that the hybrid optimization algorithm proposed in this paper outperforms comparative algorithms in terms of reliability, resource utilization, and robustness.