Taiwan, being an island, possesses abundant offshore resources, making it highly suitable for the development of offshore wind power. Welding plays a crucial role in the generation of offshore wind power because it enables the construction of steel structures that support offshore wind turbines. However, relevant personnel often fail to calculate the CO2 emissions generated from welding because of the complexity of such calculation. Therefore, in this study, we designed a CO2 detection module based on welding. CO2 detection devices were placed at the air intake site of welding spaces to measure the CO2 concentration during and after welding. The measured data were recorded, analyzed, and used for deep learning training with a deep neural network. Multiple detectors were required because of the substantial number of recorded data. Ensuring the reliable operation of fixed-point CO2 detectors is crucial to measuring emissions before large-scale monitoring efforts. Additionally, we verified that the system consistently produces reliable results in multiple CO2 measurement scenarios.