The optical power of the LED array was affected by thermology and electricity. The photo-electric-thermal (PET) parameters of the LED chip were coupled, and the thermal coupling relationship in the LED array was complicated, which makes it difficult to design the structure of the high-power light source. This paper proposes an optical power calculation model based on PET theory. Firstly, according to the working mechanism of the LED chip, the coupling relationship between its electrical power, junction temperature and thermal power was constructed. Secondly, by using the thermal coupling simulation result of the LED array as the training sample of the BP neural network, we obtain a BP neural network with input as layout spacing and thermal power and output as LED junction temperature. Finally, the junction temperature obtained by ANN was used as the temperature condition of the PET equation to calculate the optical power of the LED array. In order to verify the accuracy of the model, the experimental verification was performed. The maximum error was 7.6%. This model can analyze the maximum optical power operating point of the LED array under different layout parameters and heat sink temperatures. The optimization design problem of the LED array layout and heat sink structure was solved.