The study is aimed to develop an ideal method to estimate empirical coefficient of diffuse solar radiation model in the Chinese cities where there is no record of solar radiation data. The ground meteorological and solar radiation data from China Meteorological Data Service Center and the method of regression analysis were used to develop and verify the model. The sunshine duration, global solar radiation and diffuse solar radiation from 2000 to 2017 of 17 stations were selected. 12 individual models based on sunshine duration ratio and clearness index and 3 modified models based on geographical parameters (longitude, latitude, and altitude) were built using the data of the first ten years of each station. The data of the last four years of each station were used to evaluate the accuracy of the models. The models were evaluated using coefficient of determination (R²), mean absolute percentage bias error (MAE), and root mean percentage squared error (RMSE). Results show that the form of the model had a great influence on accuracy. The accuracy of the polynomial model was higher than that of the logarithmic and exponential forms, and the accuracy of the primary function was lower than that of the quadratic and cubic functions. Besides, the model using clearness index and sunshine duration ratio as its input parameters was the most accurate among individual models, and the model based on clearness index was the most accurate among modified models. For unobserved stations, the diffuse solar radiation can be estimated by the proposed model based on latitude, longitude, altitude, and clearness index.