Research On Intelligent Water Quality Detection And Early Warning Methods

With the speeding up of industrialization and rapid development of economy,water quality safety problem is more and more prominent in our country.In recent years,a number of water pollution incidents have not only led to disastrous environmental problems,but also caused serious social impact.The restoration of the water environment in the later stage consumes a lot of manpower and material resources.Therefore,rapid detection of water quality,prediction of pollutants in water,real-time prediction and early warning of water quality,rapid identification of water pollution areas through satellite images and other intelligent water quality detection and early warning methods are of great significance to realize automatic abnormal water quality early warning and help people find and curb the impact of water pollution events as soon as possible.In this context,this paper carries out the research of intelligent water quality detection and warning methods for water pollution events.Environment is a complex system affected by many factors,so there are various problems in water quality monitoring tasks:(1)In the process of analyzing the total alkalinity of groundwater quality in eco-industrial parks,the accuracy of feature analysis is low,and the accuracy of total alkalinity detection is not high;(2)In the process of predicting ammonia nitrogen(NH4-N)concentration and chemical oxygen demand(COD)in groundwater,the method based on mean error cannot make full use of statistical information of data,and it is easy to be affected by non-zero mean noise;(3)Due to the special geographical location,short data time and other unknown factors,the real-time monitoring and early warning ability of water quality is insufficient and the precision is low;(4)The segmentation efficiency and accuracy of satellite remote sensing images need to be improved,and the real-time early warning technology of rapid identification of polluted waters needs to be improved.Therefore,in order to solve the above problems,this paper makes an in-depth study,the content is as follows:1.In order to solve the problem that the accuracy of characteristic analysis is low in the process of analyzing the total alkalinity of water quality in eco-industrial parks,an improved genetic algorithm adaptive neural fuzzy reasoning system model was proposed to detect the total alkalinity of groundwater quality in eco-industrial parks.The model uses genetic algorithm to optimize the adaptive neural fuzzy inference system,and trains and iterates the weight value of the network in the adaptive neural fuzzy inference system,so that it has better fitting ability and stronger detection ability.At the same time,it realizes the simulation calculation of the total alkalinity distribution combined with the pollutant transport equation.Through the construction of the statistical information mining model of the total alkalinity of groundwater and the concentration characteristics of related pollutants in the eco-industrial park,the proposed model is used to simulate the experimental data and comparative analysis,proving the effectiveness of the model.2.In the process of real-time measurement and prediction of ammonia nitrogen concentration and chemical oxygen demand in groundwater,the method based on mean error cannot make full use of statistical information of data and is easily affected by non-zero mean noise,a probability density fuzzy neural network model is proposed.Firstly,the modeling error probability density function criterion is established to evaluate the modeling performance of fuzzy neural networks.Secondly,gradient descent method with adaptive learning rate is used to optimize the parameters of fuzzy neural network.In addition,the convergence analysis of the proposed probability density fuzzy neural network is carried out to ensure its effectiveness in practice.Finally,the proposed model is applied to the actual data set to achieve the prediction of ammonia nitrogen concentration and chemical oxygen demand,and the model is compared with other models,which proves that the proposed model has better performance in the prediction accuracy and anti-Gaussian noise ability.3.In the early warning process of abnormal water quality,parameters are affected by climate,human interference and other conditions,which lead to information limitation,short recording period and insufficient change trend.An effective data-driven surface water quality prediction model is proposed to analyze the internal change trend of surface water quality according to historical observation data and provide real-time early warning.This model combines improved genetic algorithm with backpropagation neural network.The former is used to select the optimal initial weight parameters of the neural network to prevent the model from selecting the local optimal results,while the latter is used to adjust the connection structure of the neural network to identify the characteristics of water quality changes,and then integrates with the data-driven model with the sliding window prediction mode to achieve the purpose of real-time warning.Finally,the simulation experiment and comparative analysis of the proposed model are carried out.4.Aiming at the problems that the segmentation efficiency and accuracy of satellite remote sensing images need to be improved,and the real-time warning technology of rapid identification of polluted waters needs to be improved,a new multi-scale residual U-Net model is proposed based on U-Net model.In this model,a multi-scale extrusion excitation module is added to each standard convolutional block,and a residual structure is added to the standard convolutional block.Meanwhile,in order to make the network have a larger receiving field,the direct connection at the bottom of U-Net network is replaced by a mixed cavity convolutional layer.A multi-scale residual U-Net model is proposed to obtain useful features of images at multiple nodes and suppress the interference of independent regions.Finally,the simulation experiment and comparative analysis of the proposed model are carried out.