Prediction Of COD Trends In Sewage Treatment Based On Deep Learning

At present,domestic sewage treatment plants are not highly intelligent,mainly relying on manpower to control and monitor equipment.In order to reduce manpower expenditure and increase the degree of intelligence of sewage treatment plants,this article will take a chemical plant in Qingdao as the practical location of the project.This project first collects the index data generated by important equipment,analyzes the collected data and judges a priori knowledge,and then performs preprocessing and noise enhancement operations on the collected data,mainly to process the noise with uniform distribution.After the treatment,the neural network model is used to realize intelligent sewage discharge.This paper systematically studies the prediction process of COD trend by BP neural network,hidden Markov model and CNN-LSTM hybrid neural network.The specific operations are as follows:First,through prior knowledge,we can see that the data fluctuates little,and the noise enhancement operation is performed on the data,which is mainly the processing of uniformly distributed noise.In this chapter,two experimental groups are set up to determine whether noise enhancement is effective.One group performs the noise enhancement operation after data preprocessing,and the other group only performs simple preprocessing on the data without adding uniformly distributed noise,and uses BP neural network to train and test the two groups of data.Through comprehensive comparison of experimental results,which method is better under the same conditions.It can be seen from the experimental results that the model with noise enhancement is better optimized.Therefore,the BP neural network is trained using noise-enhanced data,and the trained model is used to predict the COD trend.The final experimental results are very different from the actual emission standards,and the accuracy is low.The prediction performance of the BP neural network model is not strong enough to reach Standards for factory sewage discharge.The second is to build a hidden Markov model,use noise-enhanced data for model training,and predict the COD trend of the trained model.The results obtained through experiments are compared with the results predicted by the BP neural network.From the results,the prediction results of the hidden Markov model are superior to the BP neural network model.The experimental results show that the prediction accuracy is not high,and it does not meet the standards of factory sewage discharge.Finally,under the premise that the prediction accuracy of the two models for predicting the COD trend is not high and does not meet the standards of factory wastewater discharge,this paper proposes to use the CNN-LSTM hybrid neural network model to predict the COD trend.By building a CNN-LSTM hybrid neural network model,using noise-enhanced data for model training,the trained model is tested.The comparison of the test results shows that the accuracy of the CNN-LSTM hybrid neural network model is higher than that of the hidden Markov model and also meets the standards of factory prediction.In order to further confirm that the CNN-LSTM hybrid model is superior to the first two,this paper conducts experiments on the premise of the same mean square error of the BP neural network,hidden Markov model and CNN-LSTM hybrid neural network.The experimental results will be obtained in Compare accuracy and training rate.After comprehensive comparison,the final result is the optimal CNN-LSTM neural network model.