In view of the increasingly serious problem of corrosion and perforation of the oil pipe caused by water source well in the SZ36-1 oilfield, the corrosion law of the pipe is studied under field conditions and the corrosion is predicted. The Orthogonal experiment of five factors and four levels is designed to analyze the effects of temperature, pressure, flow rate, CO ₂ concentration and salinity on the corrosion rate.The main controlling factors in the corrosion environment of the water source well of SZ36-1 oilfield are temperature and CO ₂ concentration. The corrosion prediction model is established and compared by multiple linear regression analysis and BP neural network. The comparative analysis results show that the corrosion prediction model based on multiple linear regression method has higher prediction accuracy and is more suitable for corrosion prediction of oilfield water source well.