Indoor thermal environment is a critical factor for animal health and production in confined livestock facilities. In order to improve indoor thermal environment control and save energy, a novel dynamic thermal exchange model was developed using the energy balance equation (EBE) and 87 days of data collected in three different seasons in a pig building to simulate the heat transfer and energy consumption in the building. To evaluate the performances of the EBE model, a comparison was made using adaptive neuro fuzzy inferring system (ANFIS) for indoor air temperature prediction. Also, the EBE model was evaluated comparing its outputs of indoor temperature with the dataset of six days, under three different ventilation modes (Min-vent, Low-vent, and High-vent) that represent for the cold, warm and hot weather, obtained through a monitoring period in pig buildings during the production. The results showed that, under three different ventilations modes, the maximum errors between the EBE model simulated and measured data were 1.5 °C compared with 2.6 °C of the ANFIS model; and the averaged coefficients of determination R2 were 0.945 and 0.743, respectively, for the EBE and ANFIS models. Compared with the present ventilation operation, there was 358.301 kW h power saved with the EBE model in a pig room during the whole research period of 87 days. Therefore, this research has several practical applications: the model can be used in developing strategies of indoor thermal environmental control, it can also increase the knowledge about energy consumption in the livestock house. •A dynamic thermal model for indoor air temperature prediction and energy saving was developed.•Data measured for 87 days in three different seasons from a swine room were used.•Model parameters can be determined with multiple non-linear regression method.•The model contributes to heat exchange and animal building environment control strategies.