Extraction of peanut oil with SFE-CO₂ was studied by combining methods of uniform design and BP neural network. Taking peanut after being semi-baked and crushed as raw materials, four factors, including extraction pressure, temperature, time, and CO₂ flow rate, were tested at ten levels of each factor. Using the experimental data of uniform design as training samples, a neural network prediction model for SFE-CO₂ extraction of peanut oil was established. The extraction process was predicted, and the relationship between the experimental factors and the oil yield rate was analyzed. Therefore, the better technological condiitions were determined. A 4-9-1 neural network model was established. The prediction value of the oil yield was close to the experimental value, and the relative error (absolute value) was less than 2%. The neural network model could predict the trend of peanut oil yield under the influence of various parameters. Under the conditions of the extraction pressure 30MPa, temperature 40.5℃, time 125min, and the CO₂ flow rate 187L/(h·kg), the expected value of peanut oil yield was 47.5%. The method provides a reliable theoretical basis for the prediction and control of SFE-CO₂ extraction of peanut oil.