The invention discloses a solar heat collection system photo-thermal efficiency prediction method based on GA-GRNN. The method comprises the following steps that 1), parameters of the solar heat collection system are determined; 2), training test data of a network are collected and distributed; 3), a GRNN structure is constructed; 4), an optimal smoothing factor sigma is determined through GA; 5),the GA-GRNN is trained so as to obtain a trained GA-GRNN; 6), the GA-GRNN is tested, and the test data selected in the step 2 are input into the GA-GRNN which is trained in the step 5 for testing; and 7), the photo-thermal efficiency prediction is carried out by using the GA-GRNN trained in the step 6 to obtain the photo-thermal efficiency prediction result of the current solar heat collection system. According to the photo-thermal efficiency prediction method, the uncertainty of weather and climate factors can be remedied, the difficulty of data collection is greatly reduced, and the photo-thermal power of the solar heat collector can be accurately predicted.