Study Of Smart Monitoring And Prediction Method Of Rural Water Environment Based On Internet Of Things

The quality of rural water environment is directly related to agricultural irrigation safety and rural drinking water safety.The research on smart water environment monitoring methods should be urgently carried out to closely cooperate with the national strategy that "lucid waters and lush mountains are invaluable assets" and the beautiful countryside constructionBased on the technologies of Internet of Things,the paper studies the smart monitoring and prediction methods of water environment in rural areas with the purpose to sense the water environment data in rural areas as comprehensively and timely as possible,therefore we can give out an accurate and comprehensive assessment of the water environment,as well as a reasonable and effective prediction of the changing trend of water quality.The given assessment and prediction have valuable theoretical significance and wide application prospect.Orientated to the needs of water environment monitoring in rural areas,based on the design of the overall structure of smart water,and proceeding from bottom-up in the order of smart water architecture,the paper studies the following aspects of the smart water environment monitoring,namely the architecture of water environment monitoring network,modeling methods and performance analysis methods of Wireless Sensor Network(WSN)(sensing layer),water environment monitoring system and method based on private network communication(transport layer),water environment monitoring WSN data fusion algorithm(processing layer)and deep neural network water quality prediction model(processing layer)The main work and research results of this paper are as follows(1)The paper studies the architecture design of the water environment monitoring network,based on the Internet of Things.The design scheme of the overall smart water architecture is proposed,and potential challenges of the smart water environment monitoring network are analyzed.Furthermore,a distributed and extensible smart network architecture for water environment monitoring is designed so that it can meet the requirements of low latency,high bandwidth and high mobility(2)The paper studies the modeling and performance analysis of water environment monitoring WSN.Based on the analysis of the deployment strategy,coverage requirements and network model of WSN for water environment monitoring in rural areas,a method is proposed to describe WSN,the water environment monitoring system,and the system performance metrics are extracted.The formal modeling method called Performance Evaluation Process Algebra(PEPA)is used to model the clustered water environment monitoring WSN.System parameters are set based on real examples of water quality monitoring and sensor node parameters,and the working process of clustered WSN is simulated using the simulation method.The performance metrics of the constructed network are extracted and analyzed,and the effect of different design schemes on the response time of the system is discussed.On this basis,PEPA and fluid approximation method are used to model and analyze the network system with high-speed node movement and dynamic topology.The results show that this method can present the large system of water environment monitoring system as the interaction between subsystems,making the internal structure of each subsystem clear and understood,and is able to realize the simulation or approximation of the system performance,which can help to optimize the system design.(3)The paper studies the design of the water environment monitoring system and methods based on the Time-Division Long Term Evolution(TD-LTE)private network.A private networking scheme based on the TD-LTE base station is proposed for water environment monitoring,and an intelligent water environment monitoring method based on data fusion and machine learning is proposed.The proposed system and method provide a sharing,smart and integrated water environment monitoring system,which meets the demands of water environment protection workers in rural areas.Moreover,it also enables the water environment to be monitored in real time in case of emergency.(4)The paper studies the data fusion in the clustered water environment monitoring WSN.A data fusion model of cluster head in water environment monitoring WSN based on least-squares is designed,and a recursive least-squares forward data fusion algorithm is proposed to solve the problem of a large amount of computation in the data fusion model of cluster head in the data absence of some sensor nodes.The analysis results show that the proposed algorithm reduces the amount of computation in the course of data fusion and saves cluster head nodes' storage space and energy.(5)The paper studies the medium-and long-term water quality prediction based on Long Short-Term Memory(LSTM)deep neural network.On the basis of pre-processing the missing data,a single-parameter water quality prediction model based on the LSTM deep neural network is constructed.Through the application of the stacked LSTM neural network architecture,the neural network unit parameters and samples are set,and the learning process methods are determined.The model is then used to predict six parameters of drinking water quality.The results show that the model has fast convergence speed,high sample approximation accuracy and strong generalization ability.By comparing this model with other two regular time series models namely Autoregressive Integrated Moving Average(ARIMA)and Support Vector Regression(SVR),it is proved that in predicting step m=10,20,30,60,90,180,this model's dissolved oxygen prediction accuracy is better than that of ARIMA and SVR,especially the medium-and long-term prediction accuracy of dissolved oxygen,which is significantly better than that of ARIMA and SVR.With the increase of prediction step size,the advantages of this model are gradually increased.Based on the correlation analysis of drinking water quality parameters,a multi-parameter medium-and long-term prediction model based on LSTM deep neural network is designed,and the model prediction accuracy of different iterations of the neural network is compared Under the same neural network structure and conditions for parameter setting,the prediction effect of the single-parameter prediction model and the multi-parameter prediction model on dissolved oxygen is compared.According to the results,the multi-parameter water quality prediction model is more accurate in predicting the water quality of small sample data(182 groups)under the premise that the training set and test set are divided reasonably,which reflects the effect of water quality prediction by using the correlation between multiple water quality parameters.It indicates that when the number of samples is sufficient(953 groups),the two water quality prediction models have similar prediction effects.