In Brazil, thermoelectric power plants (TPP) are responsible for about 25% of the electric energy generated. Of the energy supplied by the thermoelectric plants, 11% comes from the burning of heavy fuel oils (HFO). The monitoring of these oils is of great industrial interest in order to maintain their quality, avoid damage to the engines, and guarantee the energy supply. In this sense, this work investigated the potential of near-infrared (NIR) spectroscopy combined to partial least-squares (PLS) and artificial neural network (ANN) models for online monitoring of temperature, density, water content, and kinematic viscosity of the HFO used at Energetica Suape II S.A. Chemometric models were calibrated to predict the properties of interest in the temperature range typically found at thermoelectric power plants (from 25 to 120 degrees C). The results show that NIR spectroscopy combined with multivariate calibration is a powerful tool for online monitoring of HFO. ANN models presented a better performance than PLS models, mainly for nonlinear properties like kinematic viscosity; however, all determination coefficients were bigger than 0.95 for all properties when compared to standard methods. The next step of this work will comprise the installation of an industrial NIR at TPP and evaluate the model's performance when applied in real industrial process.