Agricultural and food residues are sustainable biofuels and their utilization in combustion, carbonization, and gasification reduces greenhouse gas emissions. In this work, the thermal decomposition of corn cobs, rice husks, vine prunings, and palm kernel shells was studied in thermogravimetric tests to elaborate reliable pyrolysis kinetics, which are basic parameters for fixed beds and grate reactors. The adequacy and predictability of different models (single first order reaction model, n-order, distributed activation energy, and a structural model based on the decomposition of cellulose, hemicellulose, and lignin) were quantified. The single first order reaction model had low computational cost but poor accuracy (mean discrepancy 5-8 %, maximum discrepancy > 30 %), while the accuracy increased when a further parameter (reaction order) or a distribution in the exponential factor was introduced. The best accuracy (mean discrepancy 1-2 %, maximum discrepancy 5 %) was found for the structural model, with common kinetic parameters of the chemical components.