Multicomponent online gas measurement in the production of olefin is an important approach for effective control and improvement of the overall efficiency of the production process. In this study, we took the online measurement of CO and CO2 for an olefin cracking furnace coal cleaning process as the application example. A Tunable Diode Laser Absorption Spectroscopy (TDLAS) based analyzing platform was developed to facilitate multicomponent measurement. To simulate the reaction process, we designed 0-5% range CO and CO2 tests. Based on the first set of random concentration mixing tests with 19 collected spectra, single component partial least square fitting algorithm models (PLS1) and a multicomponent partial least square fitting algorithm model (PLS2) were developed and evaluated, along with a multivariate classical least square fitting algorithm model (CLS). In subsequent interference and full range step tests, the maximum errors for PLS1, PLS2, and CLS were less than ±0.05%, less than ±0.10%, and less than ±0.20% for CLS. These results demonstrate that the combination of TDLAS and the PLS1 algorithm performed the best during the multicomponent online measurement in the petrochemical process.