In order to obtain the module temperature of PV array, by analyzing the relationship between module temperatures and influencing factors, the function law between various external environmental factors and module temperatures is determined. Combining multivariate nonlinear function fitting with BP neural network, this paper proposes a more comprehensive method to calculate the temperature of PV modules. Aiming at the partial shading of PV arrays, the method is modified by analyzing the relationship between temperature of modules and the maximum power point. The actual data from a photovoltaic power plant verify that the method has high accuracy and strong practicability. It can reflect the changes in the external environment and has a strong self-adaptivity.