Abstract: Permeability is an important parameter for coalbed methane development.The strong heterogeneity of coal reservoirs makes it difficult for reflecting the regional permeability law by a small number of well test results.It is also hard to compare the experimental permeability data obtained under different instruments, methods, sample sizes, pressures, temperatures, and media conditions.This paper compares and analyzes the advantages and disadvantages, applicable conditions, and permeability of different well testing methods(drill stem testing, water tank testing, slug well testing, injection/fall-off testing)and different experimental testing methods(steady-state method, non-steady-state method).Based on the evolution history of permeability test standards, the applicability of permeability simulation models under uniaxial strain conditions, constant volume conditions, and triaxial stress conditions has been systematically reviewed.Also, the reliability of predicting permeability based on single main controlling factors such as fissures, coal body structure, in-situ stress and tectonic curvature and artificial intelligence multi-factor prediction(back propagation neural network, grey relational analysis, support vector regression machine, and multi-level fuzzy comprehensive evaluation)was analyzed.The study showsthe problems existed in the testing, simulation and prediction of coal reservoir permeability and puts forward the prospect of further research.The results show that the field injection test is the main method for obtaining the permeability of in-situ coal reservoirs.The laboratory unsteady-state permeability test is suitable for low permeability coal reservoirs in China.The P-M model and the S-D model under uniaxial strain conditions are suitable for the simulation of constant vertical permeability.For simulation under constant volume conditions, the Ma model can measure most coal input parameters, and the simulation results are more reliable.The Connell model and Zhou model under triaxial stress conditions are more conducive to the verification of simulated permeability by laboratory measurements.The permeability obtained based on the historical matching of coalbed methanewell production data is closer to the in-situ coal reservoir, and the prediction effect of the coal reservoir permeability based on a single main control factor is better in the area without coalbed methane test.The penetration rate predicted by artificial intelligence technology that integrates multiple factors is reliable.