In this study, we designed a metamaterial-based terahertz multi-band sensor integrated with microfluidic channels. We simulated the reflection spectra of the sensor during the detection of ethanol-water mixtures containing different concentrations of ethanol. The simulation results show that an increase in the ethanol concentration correlates with a decrease in the reflectivity at the resonant dips and a blue shift of the resonant frequency. We analyzed the quantitative relationship between resonant frequency or reflectivity of the sensor with ethanol concentration, which is in turn used for the prediction of the ethanol concentration in an ethanol-water mixture. Three resonance dips are used for predicting the ethanol concentration and the predicted errors are smaller than 1%. The above results demonstrate the utility of the terahertz time-domain spectroscopic technique in the rapid, real-time, and infinitesimal material identification and bio-sensing.