The mathematical model of boundary temperature regime for accumulation and recovery of heat in a stationary switching regenerative heat recovery unit (SRHRU) is constructed. Modelling of the boundary regime in SRHRU is based on experimental data, and it is required to develop algorithms for calculating thermal constants and the general temperature regime. The boundary temperature regime is modelled by a system of nonlinear ordinary differential equations of the first order, which take into account the results of experimental observations and measurements. To determine the unknown parameters of the system of differential equations, artificial neurons are used: a two-layer perceptron is developed with three neurons on the first layer and with five neurons on the second layer.