With the explosive growth of IoT devices and data volumes, traditional centralized machine learning methods can no longer meet the demands of modern application scenarios. Hierarchical Federated Learning (HFL), as an enhanced federated learning framework, builds a multi-level collaborative mechanism among devices at different layers. It ensures data security while maintaining high scalability, offering broad application prospects. However, existing HFL schemes still face numerous challenges in task scheduling and resource allocation. This paper proposes a joint optimization solution to address the task scheduling and resource allocation problems within the HFL framework. Firstly, we model the task scheduling and resource allocation problem in the HFL system as a constrained optimization problem, with the objective of minimizing global training time and energy consumption. Subsequently, we introduce a hybrid heuristic optimization algorithm to solve this problem. This hybrid algorithm leverages the strengths of both PSO and SA, enabling fast solutions while reducing the likelihood of getting trapped in local optima. Experimental results demonstrate that our algorithm can efficiently complete task scheduling and resource allocation in the hierarchical federated learning system with high quality.