With the intelligent transformation of Industry 4.0, the limitation of programmable logic controller (PLC) as the core equipment of industrial control has become an obstacle to the development of industrial automation. Cloud PLC is a key technology which supports virtualization deployment of multiple PLCs in a central cloud server and interacts with remote actuators through 5G communication networks. However, it’s an issue about how to make full use of communication and computation resources to bear as many control workloads as possible in this scenario. To this end, this paper proposes a cloud control system for control workloads. Each control workload can be partially processed locally and remotely in parallel. Then, this paper considers the deadline of control workload and the limited resources as constraint boundaries, and proposes a stochastic workload offloading scheme based on the deep reinforcement learning (DRL). The scheme utilizes the twin delayed deep deterministic policy (TD3) scheme to optimize the workload distribution and network resource allocation to guarantee workload execution success rate. The simulation results demonstrate that the proposed algorithm in terms of average system energy consumption outperforms other benchmark ones.