With the in-depth development of global digital transformation, industrial IoT has put stringent requirements on the transmission delay, reliability, and wide coverage of network communication. TSN, with its deterministic low-delay transmission characteristics, complements the wide coverage advantage of 5G. The TSN-5G heterogeneous network, formed by the deep integration of the two, has become a cutting-edge solution to meet the real-time communication needs of industrial IoT. However, currently there is a lack of optimization methods for fine-grained use of transmission resources in this solution, resulting in the underutilization of limited resources, which in turn affects the success rate of end-to-end transmission. Therefore, based on the TSN and 5G fusion industrial network architecture, this paper designs a whale optimization algorithm based on dual Q-learning. The algorithm uses DDQN to dynamically optimize the execution action selection of “search” or “encirclement” in the whale optimization algorithm, which can accelerate the optimization process and effectively avoid the whale optimization algorithm falling into local optimum. In terms of performance comparison, the algorithm also shows significant advantages. Compared with the traditional whale optimization algorithm and genetic algorithm, The improvement of scheduling success rate ranges from 6.1% to 74%. Thus, the combination of DDQN and whale optimization algorithm can effectively solve the fine-grained resource allocation problem in TSN-5G networks.