Studies on the Variation of Chloroform and Its Prediction Model in Water Distribution Network

According to the monitoring and analysis of chloroform in the water distribution network in a northern city of China, the variations of chloroform in the water distribution network and the major influence factors were studied. Using principle component analysis method, the prediction model which including 9 water quality indexes was established to predict the concentration of chloroform and the average prediction accuracy was more than 80%. Four prediction models, including neural network model and support vector machine model etc. were established using 4 conventional monitoring indexes such as temperature and residual chlorine. The average prediction accuracies of various models were in the range of 83.42% ~ 88.40%, which could absolutely fulfill practical requirements, namely, the adjustment measures could be taken in time basing on the prediction results to decrease the chloroform concentration in the water distribution network.