Research On Remote Monitoring And Inversion Model Of CO₂ Concentration Based On Embedded System

Accurate measurement of carbon sink data onto land surfaced plays an important role in assessing carbon balanced and carbon cycle of terrestrial ecosystem.China’s forests are widely distributed,and the advantages of carbon sinks are obvious.In view of the complex nonlinear relationship between carbon flux and environmental factors in forest ecosystem,the inversion model of CO₂ concentration was studied by using the collected environmental parameters,so as to grasp the changing characteristics of forest CO₂ concentration and provide reference to carbon balance,climate change and emergency mechanism of forest ecosystem.Based on the above background and current research situation,a forest carbon sink monitoring system was developed through investigation and study.The system was used to collect environmental factors related to forest carbon flux in real time,and study the inversion model of CO₂ concentration.The main work are as follows.（1）Based on the micrometeorology and vorticity correlation method,a remote real-time data monitoring system of forest carbon flux based on embedded system was designed independently,combining with the demand of continuous and dynamic sensing information of carbon flux.This system was used to collect and monitor the data of one month onto the experimental site.（2）Based on 20000 groups of field data,the correlation between environmental factors and CO₂ concentration was analyzed.The results showed that the correlation coefficient of temperature-CO₂ concentration was 0.767;The correlation coefficient of humidity-CO₂concentration was 0.698;The correlation coefficient of temperature-humidity was 0.987.（3）Based on the conclusion of work content（2）,the inversion model of CO₂ concentration based on temperature and humidity is put forward.BP neural network and genetic algorithm optimized BP neural network（GA-BP）are used to build the model.The standard deviation（STDEV）of the training set of the model is 15.42,the standard deviation of BP neural network is 13.73,and the standard deviation of the output of GA-BP neural network model is 14.73.The output dispersion of GA-BP network is closer to the real value.（4）In order to verify the stability and accuracy of the model,this paper evaluates the results of the improved neural network inversion model by using decision coefficient（R²）,mean absolute error（MAE）,root mean square error（RMSE）,mean percentage error（MAPE）and standard deviation（STDEV）.The determination coefficient R²)of GA-BP neural network is0.85.The results show that the R² of BP neural network is only 0.77;The average absolute error of GA-BP neural network is 4.91,and that of BP is 6.64;The inversion model of GA-BP neural network has more stable inversion performance and higher inversion precision than that of conventional BP neural network.In this study,a remote real-time carbon flux data monitoring system is designed,which lays a foundation for building a comprehensive platform for quantitative monitoring of regional carbon emissions and carbon sinks and real-time release of carbon revenue and expenditure information;The accurate inversion and inversion of CO₂ concentration are realized,and the changing trend of forest CO₂ concentration is mastered,which is beneficial to provide the basis for carbon trading and providing effective evaluation for carbon emission right.