Download PDFOpen PDF in browser

Predicting Turbidity in Water Distribution Trunk Mains Using Nonlinear Autoregressive Exogenous Artificial Neural Networks

10 pagesPublished: September 20, 2018

Abstract

A nonlinear autoregressive exogenous artificial neural network model was developed to predict turbidity response in two different trunk mains with measured flow and turbidity data. Models were initially established to prepare the data and automatically select the appropriate events for model training. Then, an autoregressive exogenous network model was developed and applied to predict turbidity responses based on past events in the time series. A per site continual data driven calculation of turbidity event risk was included as an additional input to capture the effect of temporal distance between the selected events as well as increasing the accuracy of the predictions. The calculated normalised mean square error and mean absolute error showed that the developed model combined with the data preparation and pre- processing models provides good regressions on a future event with a period of 7 to 10 hours for a multi-step ahead prediction. Furthermore, the result of the autoregressive exogenous network was compared with the output of a feed-forward network where the former significantly outperformed the latter (R value of approximately 0.97 compared to 0.66).

Keyphrases: ann, machine learning, narx, trunk mains, turbidity, water distribution systems

In: Goffredo La Loggia, Gabriele Freni, Valeria Puleo and Mauro De Marchis (editors). HIC 2018. 13th International Conference on Hydroinformatics, vol 3, pages 1030-1039.

BibTeX entry
@inproceedings{HIC2018:Predicting_Turbidity_Water_Distribution,
  author    = {Ehsan Kazemi and Stephen Mounce and Stewart Husband and Joby Boxall},
  title     = {Predicting Turbidity in Water Distribution Trunk Mains Using Nonlinear Autoregressive Exogenous Artificial Neural Networks},
  booktitle = {HIC 2018. 13th International Conference on Hydroinformatics},
  editor    = {Goffredo La Loggia and Gabriele Freni and Valeria Puleo and Mauro De Marchis},
  series    = {EPiC Series in Engineering},
  volume    = {3},
  publisher = {EasyChair},
  bibsource = {EasyChair, https://easychair.org},
  issn      = {2516-2330},
  url       = {/publications/paper/7jMJ},
  doi       = {10.29007/9r3b},
  pages     = {1030-1039},
  year      = {2018}}
Download PDFOpen PDF in browser