Water COD Prediction Based On ANFIS Model

Water is an important component of all organisms,and with the development of society and production,water available on earth is running short.Therefore,it is especially important to explore the quality and safety of water environment.Chemical Oxygen Demand(COD)is used to treat water samples with a certain amount of strong oxidant under certain conditions.It is an indicator of the amount of the reducing substance in water.The greater the chemical oxygen demand,the more serious the pollution of organic matter.In this paper,the research area of six rivers and reservoirs in zhejiang province is used to determine the PH,dissolved oxygen(DO),ammonia nitrogen(nh3-n)and total phosphorus(TP).4 water quality indexes are input parameters,and the COD index is used as the output parameter.In each study area,600 sets of data are collected as experimental data.Two commonly used neural network models are used to predict and analyze,namely Back Propagation BP network model and Radial Basis Function(RBF)network model,and their respective advantages and disadvantages are pointed out.Due to the shortcomings of artificial neural network,a fuzzy neural network model is established by combining fuzzy inference system with artificial neural network.After establishing three kinds of network prediction models,the training data of each monitoring point is input into them for training and learning.After the training of BP network,RBF network and ANFIS prediction model,the monitoring data of each monitoring point were input into three prediction models,and the performance of three forecasting models for predicting COD content in water area was obtained.Finally,by comparing the average relative error,the mean square error and the root mean square error to judge the advantages and disadvantages of the three forecasting models,to determine whether the ANFIS model has good predictability and generalization.The experimental results show that the average relative error of the prediction model of ANFIS is 5% lower than that of traditional artificial neural network model and the predicted data dispersion is lower,which is closer to the actual value.From the generalization performance,the ANFIS prediction model has good generalization,and it can predict the content of COD in water easily,quickly and accurately for the water bodies of different locations.