A method for predicting product-yield of corn stalk pyrolysis was established by means of BP neural network model. The model consisted of three neuron layers: input layer with four nodes which affected the pyrolysis process. It included input power, air flow rate, feeding rate and pressure, output layer with pyrolysis liquid yield and hidden layer. If the training data were representative, the results obtained by neural network model could be well in accordance with the experimental results and its errors would be less than 5%. The results obtained by neural network are more accurate than those obtained by non-linear regression.