Harmful gas concentration recognition model training method, device and system and monitoring method

Abstract

The invention discloses a harmful gas concentration recognition model training method, device and system and a monitoring method, and belongs to the field of gas monitoring. During model training, firstly, an initial recognition model is constructed, then samples are input into the initial recognition model to obtain the predicted concentration of each harmful gas, then a total output error is calculated according to the predicted concentration and the real concentration, and the initial recognition model is adjusted according to the total output error until the total output error is smaller than or equal to a preset error. In this way, the trained recognition model takes the reading obtained by the concentration detection device as input, and then outputs the predicted concentration of each harmful gas in the mixed harmful gases, which is equivalent to that the recognition model can remove the monitoring error of the concentration detection device caused by the mutual influence of each mixed harmful gas. It is guaranteed that the predicted concentration output by the recognition model is the same as the actual concentration, and real-time and accurate regulation and control can be conducted on the air quality in the environment.

Images