The low temperature phase change heat storage unit with tubular internal flow gridding was designed. The key factors affecting the heat transfer coefficient of the phase change heat storage unit were determined, and the influence of single factor on the heat transfer coefficient of phase change heat storage unit was analyzed. In the case of heat storage and heat release, the prediction model of heat transfer coefficient of the phase change heat storage unit was established by using the improved multivariate nonlinear regression method, and the fitting error was tested. The results showed that the heat transfer coefficient of the phase change heat storage unit was synergistically influenced by the average temperature of the phase change heat storage material side and the qualitative temperature of the heat transfer working medium side. The average temperature of phase change heat storage material side was the main influencing factor, the qualitative temperature of heat transfer medium side was the secondary influencing factor. And there was significant interaction between the above two. In the case of heat storage or heat release, the change law of heat transfer coefficient of phase change heat storage unit with single factor was basically consistent, the heat transfer coefficient in the heat storage stage was obviously higher than that in the heat release stage. The average relative errors of the prediction model of heat transfer coefficient of phase change heat storage boxes were all less than 5.00%. Based on this study, the intelligent phase change constant temperature system configuration in agricultural greenhouses could be guided and optimized.