Research On Sensor Calibration Method Based On Internet Of Things

In recent years,air pollution is becoming increasingly serious,and improving air quality is the people's urgent expectations.Therefore,the construction of air quality monitoring system has become the focus of environmental protection bureaus and environmental protection enterprises.Firstly,establishing an air quality monitoring system needs to arrange the gas sensors at the location where monitoring is required,and then all gas sensors form a sensor network.Due to the gas-sensing characteristics of the gas sensor,the gas sensor is subject to cross interference when monitoring mixed gas pollutants,resulting in inaccurate monitoring.In this paper,the problem of the electro-chemistry gas sensor around the industrial park generated cross interference when monitoring the inorganic gas pollutants is researched from the aspect of calibration method.At present,the mainstream calibration method is building the sensor calibration model by using the neural networks to train the upload data from the Internet of Things formed of the gas sensor clusters.So that,in this paper,the sensor calibration model based on error back propagation(BP)neural network has been optimized in algorithm and improved.The main work of this paper includes:Firstly,this paper analyzes the BP algorithm used in the conventional BP neural network model.It is pointed out that the BP algorithm is easy to fall into the local optimum when training the calibration model because of its theoretical defect.And an optimization algorithm combining Particle Swarm Optimization(PSO)with BP algorithm is proposed.The optimization algorithm optimizes the initial weight of the network and makes full use of the advantage of global optimization of PSO algorithm and combines with the local optimum of BP algorithm,so that the optimization algorithm can avoid the local minimum in the training process and accelerate the rate of convergence of calibration model training.Then,the working mode of the calibration model and the real monitoring data of the gas sensor are analyzed.From the perspective of information utilization and the actual application environment of the calibration model,it is pointed out that the conventional calibration model based BP neural network can not make full use of the gas sensor monitoring concentration information,so that the calibration accuracy of the calibration model is limited.And the integration of long-short-term memory(LSTM)neural network and BP neural network is proposed to improve the calibration model.The improved model firstly eliminates the influence of unknown gas on the gas sensor through the LSTM network by using the continuous change of the gas concentration in time,and then cascade the BP neural network to realize the improvement of the model and improve the performance of calibration model.Finally,from the viewpoint of data verification,the simulation results show that the optimization algorithm proposed in this paper is superior to BP algorithm in training process of the calibration model,and the improved calibration model has a higher accuracy than the conventional calibration model in the practical application environment.