Optical transport networks (OTNs), while increasingly popular, can be affected in ways that are challenging to restore efficiently. This study investigates the problem of service restoration under OTNs with an optical channel data unit (ODU)-k switching capability (OTN-OSC) environment. An advantage actor-critic-based service restoration (A2CSR) algorithm is presented with the objective of increasing the service restoration rate. In our experimental setup, A2CSR uses the advanced image recognition model MobileNetV2 and an advantage actor-critic reinforcement learning algorithm. Simulation results show that the proposed A2CSR algorithm can achieve better blocking probability and resource utilisation than the benchmark algorithm (first fit (FF)), and the restoration time is within the acceptable range.