The global energy demand supplies mainly from fossil fuels, which is neither sustainable nor environmentally friendly and aims to global warming. Therefore, both more investments in renewable energy sources such as bioenergy are required, as well as new technologies such as carbon capture and storage (CCS) to handle the emissions from existing combined heat and power (CHP) plants. In this degree project, the focus is to determine the moisture content, ash content, heating value, and elemental compositions of solid biofuel mixtures in real-time by utilizing the optical technique of near-infrared (NIR) spectroscopy. A total number of 150 samples of solid biofuel mixtures were prepared and illuminated by NIR light. All spectra of the samples were recorded in a wavenumber range of 12000 cm-1 – 400 cm-1 in a dish on a turn table which was in a moving mode with a speed of 0.5 m/s. Each sample was scanned three times to avoid, or at least minimize the deviation of the spectra and the samples were mixed between each scan to get more reliable representative spectra data. Partial least square regression models were created to analyze the spectra data. A data split was done randomly, 100 for calibration and 50 for validation. Then the data was pre-processed with different methods including multiplicative scatter correction (MSC), standard normal variate (SNV), Savitzky-Golay 1st derivative (SG 1st), Savitzky-Golay 2nd derivative (SG 2nd), and orthogonal signal correction (OSC) to reduce noise and scatter effect. The results of NIR spectra treated by OSC method obtained  , RMSE and SE of 0.900, 2.241 and 2.204, respectively for prediction of moisture content, 0.424, 0.913 and 0.922 for prediction of ash content, 0.640, 0.370 and 0.368 for prediction of heating value, respectively. The obtained prediction of  , RMSE and SE were 0.687, 0.066 and 0.058 for nitrogen, 0.636, 0.361 and 0.364 for carbon, 0.483, 0.269 and 0.270 for hydrogen, respectively. As the results shows, these models to predict the ash content and hydrogen content has a lower accuracy than what is expected in process modeling while the prediction of moisture content has the highest accuracy.