[Significance/Objective] The optimization of the service composition manufacturing mode and the search for an evaluation method to support the service composition are hot research issues in the field of cloud manufacturing. With the deepening of manufacturing technology, cloud manufacturing has developed from unit digital manufacturing to networked and intelligent manufacturing. As a comprehensive technology that integrates computer, model and cloud manufacturing service combination theory, modeling and simulation technology plays an important role in the development of cloud manufacturing. In view of the current situation of cloud manufacturing service combination evaluation and simulation research at domestic and abroad, this paper introduces the research background, characteristics and NP-hard problem of cloud manufacturing service composition process. Meanwhile, the theory and evaluation method of cloud manufacturing service composition are discussed, and the research and application of modeling and simulation technology in service composition are summarized, which help to promote the further development of cloud manufacturing service composition. [Methods] Firstly, the cloud service composition evaluation system and evaluation methods are refined to address the problems of cloud manufacturing service composition evaluation theory. Then, reasonable views and suggestions are given from different perspectives such as indexes, algorithms and models. Finally, the existing techniques and applications of simulation models and more reasonable simulation modeling methods are summarized. [Results/Conclusions] Three main evaluation systems exist for cloud manufacturing service composition: Quality of Service (QoS) evaluation, service reputation evaluation and other service evaluation. Based on these three evaluation systems, domestic and abroad scholars have constructed cloud manufacturing service composition evaluation models from different perspectives in traditional and complex environments, and these models have achieved good application results in service composition optimization, IoT service optimization, natural language processing and other fields. It is found that the multi-agent simulation modeling method is more suitable for describing the distribution, dynamics and diversity of cloud manufacturing service composition.