Abstract Composting can be defined as an exothermic process involving the microbiological decomposition of organic substances taking place in aerobic conditions with the active participation of thermophilic microorganisms and mould. The process generates a lot of heat, which is dissipated into the environment. If it were possible to acquire the lost heat energy, it could be then used for various utility purposes. An important problem is estimating the heat loss emitted as a result of the exothermic transformations during the composting process. The purpose of this paper was neural modelling of the composting process of solid natural fertilisers with special attention paid to heat analysis. The paper highlights the problem of neural prediction of heat processes accompanying the composting of selected natural fertilisers. It focuses on the estimation of lost heat generated (which roughly corresponds to the thermal energy generated) as part of the exothermic reactions taking place during the process. The obtained results show that neural modelling can be effectively used in the process of estimating heat energy emitted and lost in the composting process. The model's analysis of sensitivity to input variables showed that the 6 most important parameters in the process of neural estimation of heat lost are (in the following order): T (temperature inside the bioreactor), SM (mineral substance mass), O 2 (% content of oxygen), V (stream volume), CO 2 (% content of carbon dioxide), and TIME (process duration).