Research On Water Quality Prediction Methods Based On Multi-indicator Timing Data

Water pollution has become the top pollution category that seriously threatens the public,and effective water resource protection is becoming increasingly important.Water quality prediction is an important way to protect water resources,and timely and accurate water quality prediction can effectively prevent the occurrence of serious water pollution events.Therefore,water quality prediction has very important practical significance.Water quality prediction is influenced by various factors,such as the completeness and effectiveness of water quality datasets,whether the selection of water quality prediction methods fully considers the temporal nature of water quality data,and the multiple correlations between water quality data indicators.However,traditional water quality prediction methods currently fail to effectively consider the impact of these factors on water quality prediction,resulting in low prediction accuracy.Therefore,in order to solve the above problems and provide an accurate and effective water quality prediction method,this article mainly conducts the following research:(1)This paper proposes a water quality data interpolation method based on VAE-WGAN-LP to address the issue of low quality and abnormal data in water quality monitoring datasets.Firstly,in order to address the temporal nature of water quality data,an improved Gated Recurrent Unit(GRUI)is used as the basic network of WGAN-LP,and a feature matrix is introduced to label the location of abnormal data;Then,the Variational Auto Encoders(VAE)are combined with WGAN-LP to learn the data distribution between real water quality data samples,and finally,interpolation and filling are performed.Use this method to interpolate the original water quality monitoring dataset(including abnormal data)to obtain a complete normal dataset(excluding abnormal data).Finally,the method is applied to two actual data sets: Wuliangsuhai data set and Taihu Lake data set to verify that the method can indeed improve the interpolation effect of water quality anomaly data.Experimental results show that this method can improve the effect of water quality interpolation.(2)This paper proposes a water quality time series prediction method based on GAT-Transformer to address the issue of temporal variability and multiple correlations between water quality data indicators.Firstly,a Graph Attention Network(GAT)is introduced into the Transformer to establish a water quality prediction model for GAT-Transformer;Secondly,Transformer is used to capture the temporal characteristics of water quality data for mining temporal information,while graph attention networks are used to capture the multiple correlations between water quality data indicators;Then,summarize the process of GAT-Transformer water quality prediction method.Finally,the method is applied to the Wuliangsuhai dataset for prediction experiments.The experimental results show that this method can effectively mine the multivariate correlation between water quality indicators and improve the prediction accuracy.This paper conducts experiments from two aspects of water quality data interpolation and multivariate correlation of water quality data.The research results can improve the accuracy of water quality prediction,provide important support for water resource protection,and effectively prevent serious water pollution incidents.