Aiming at the problems of limited real-time prediction of main gas concentration in low-concentration gas direct current(DC) oxidation systems and insufficient adaptability under sensors such as methane concentration, an adaptive prediction method for main gas concentration in low-concentration gas DC oxidation system was proposed. Firstly, based on grey relational analysis(GRA), characteristic indicators with high correlation with main gas concentration were selected to establish a normalized dataset to enhance the adaptability of the model. Secondly, the main gas concentration prediction model was constructed to predict the main gas concentration. Further, the adaptability and prediction real-time of the model were compared and analyzed from the mean square error and sample running time. Finally, the model was integrated into the main gas concentration prediction platform of DC oxidation system developed based on artificial intelligence and software. The experimental results show that the mean square error of the main gas concentration prediction model is 0.061 1, and the running time is 4 023 s. The prediction effect of the main gas concentration is better than that of long short-term memory (LSTM) and other models. The prediction effect of the main gas concentration prediction platform of the DC oxidation system is close to the monitoring value, which provides technical support for the safe operation of the system.