Research On Temperature Compensation Algorithm Of Chemical Oxygen Demand Detection Based On Neural Network

Human life and development are inseparable from water for a moment.Water pollution not only harms human health,but also harms the development of agriculture and fisheries,and even affects the quality of water used in the food industry.Chemical oxygen demand,as a comprehensive index for evaluating the degree of water pollution,has attracted widespread attention due to its simple and rapid measurement method.With the needs of the market,water quality detectors are developing towards miniaturization,and semiconductor devices are used for light sources and detectors.The LED light source and the detector are affected by the ambient temperature,resulting in a deviation between the COD measurement value and the actual value.Therefore,in view of the influence of temperature,this paper proposes a temperature compensation method based on neural network,and on the basis of temperature compensation,it further compensates the influence of atmospheric noise generated by lightning on the detection signal.Ensure that the water quality detector can also work normally under thunderstorms.The main contents of this paper are as follows:Establish a temperature compensation model.Taking the environmental temperature,the light power of the LED light source and the dark current of the detector as the research variables,the BP neural network algorithm and the particle swarm algorithm are used to establish the temperature compensation model of the chemical oxygen demand detection according to the influence of temperature on the COD measurement.In the process of model building,it is necessary to solve the problem that BP neural network is easy to fall into the local optimal solution and the inertia weight setting in the particle swarm algorithm.The performance of genetic algorithm,ant colony algorithm and particle swarm algorithm is compared from the three evaluation angles of algorithm accuracy,robustness and convergence speed.Finally,the particle swarm algorithm with the best performance is selected to optimize the shortcomings of the BP neural network algorithm that is easy to fall into the local optimal solution.Because the choice of inertia weight will affect the compensation effect of the entire model.Therefore,two inertia weight selection schemes are designed on the constructed temperature compensation model.The first solution is that the inertia weight adopts a linear decreasing strategy.The second option is to use a non-linear decreasing strategy for the inertia weight.The determination coefficients of the test sets of the two schemes are 0.9544 and 0.9838,respectively,so the inertia weight of the whole model chooses scheme two.Compensation for temperature and random errors.Three-dimensional linear interpolation and particle swarm optimization BP neural network are used to compensate the chemical oxygen demand.For multivariable factors,the temperature compensation effect of BP neural network optimized by particle swarm is better.The simulation results show that the BP neural network optimized by particle swarm has a certain inhibitory effect on the jump of individual samples.For the same sample to be tested,the maximum relative error of the compensation value of the three-dimensional linear interpolation method is 9.8%,and the maximum relative error of the compensation value of the particle swarm optimization BP neural network is 0.81%.In addition,the use of particle swarm optimized BP neural network can suppress the influence of atmospheric noise generated by lightning on the detection signal,it can restore the target signal,and ensure that the water quality detector can work normally under thunderstorms.This is a capability that the three-dimensional linear interpolation method does not possess.