Rapid analysis of biomass fuel is of great importance to the energy utilization of agricultural and forestry waste. The models for predicting the moisture, ash, volatile matter, fixed carbon and calorific value of three kinds of agricultural and forestry waste such as pine wood, cedar wood and cotton stalk are established by using a visible and near-infrared spectroscopy. All of these models have a determination coefficient greater than 0.88 after cross validation. When the artificial neural network (ANN) modeling with several latent variables is used, the models have the average determination coefficient of up to 0.95 for moisture. The result shows that the visible and near-infrared spectroscopy combined with chemometrics can be used to replace the traditional analysis methods in industry completely and can provide a new method for the rapid detection of the biomass fuel properties of agricultural and forestry waste.