To the sluice pier concrete，temperature control and crack prevention is a complex and multi- factor problem of system optimization related to temperature control measures and material parameters. This paper tries optimal temperature control measures with the known concrete material thermodynamics parameters，takes minimum values of the relationships between the sluice pier concrete structure’s internal/surface principal tensile stress duration curve and tensile strength growth curve as inputs and the sluice pier superficial heat preservation effect，pouring temperature，pipe cooling temperature and duration time as outputs，establishes the neural network model of the optimal temperature control measures，takes the uniform design principle to have the temperature control parameter combination，adopts the pipe cooling finite element method（FEM）to simulate and analyze the temperature field and creep stress field of the sluice pier concrete structure with cooling pipe，and gets samples to train network to describe the nonlinear relationship between the relationship's minimumvalueof the structure's principal tensile stress duration curve and tensile strength growth curve and different temperature control measures. Inputting the appropriate relationship's minimum value of structure’s the principal tensile stress duration curve and tensile strength growth curve to the trained network，it can automatically select the optimal temperature control measures. The example shows that the neural network model of the optimal temperature control measures is feasible.