By revealing that a key parameter D in Chen-Guo model is related to the temperature T, a new temperature-dependent parameter D(T) trained by the reported experimental data is proposed with Chen-Guo model to improve the prediction accuracy of phase equilibrium conditions of methane hydrate in the range of 145.75 K- 272.2 K. Four equations of state (EOS) such as Redlich-Kwong (RK), Soave-Redlich-Kwong (SRK), Peng-Robinson (PR) and Patel-Teja (PT) are chosen to calculate the gas fugacity respectively and the corresponding prediction accuracy is evaluated by the average absolute deviation percentage (AADP). The process of training D(T) based on the Chen-Guo model employed with SRK EOS is described in detail first, and the prediction results show that the corresponding prediction accuracy is greatly improved, whose AADP decreases from 10.83% to 3.31%. In order to verify that training D(T) is an effective way to improve the prediction accuracy, the same procedure is also used to the Chen-Guo models employed with RK, PR and PT EOS respectively, and the corresponding AADP decreases from similar to 11% to similar to 3%. In addition, the prediction accuracy according to the proposed D(T) is also evaluated by other temperatures different from the data used for training D(T) to verify its feasibility and effectiveness. (C) 2019 Elsevier B.V. All rights reserved.