The tiered carbon trading mechanism and optimization scheduling model solving algorithm are pivotal for the community integrated energy system (CIES). CIES plays a crucial role in optimizing scheduling, yet existing literature often does not fully consider these two factors. To address this gap, the adoption of the proximal policy optimization (PPO) algorithm is proposed, which incorporates a ladder-type carbon trading mechanism to solve the low-carbon optimization scheduling problem of CIES. This method constructs a reinforcement learning interactive environment based on a low-carbon optimization scheduling model. The intelligent agent's state, action space, and reward function are defined using device status and operating parameters. An intelligent agent capable of generating the optimal policy is obtained through offline training. Case study analysis results demonstrate that the low-carbon optimization scheduling scheme for CIES achieved through the PPO algorithm, effectively leverages the advantages of the tiered carbon trading mechanism, significantly reducing carbon emissions and improving energy utilization efficiency.