Please use this identifier to cite or link to this item:
Full metadata record
|dc.identifier.citation||Enviro 04 Convention & Exhibition [electronic resource] : proceedings: [CD-ROM, 11 p.]||-|
|dc.description.abstract||Disinfection is an important process in the treatment and supply of potable water, as it protects public health from the threat of waterborne diseases. The use of chlorine for disinfection is widespread due to the cost-effectiveness and simplicity of the chlorination process and the persistence of a free chlorine residual within the water distribution system (WDS). Maintenance of a chlorine residual is a commonly adopted strategy for secondary disinfection of potable water supplies, however effective maintenance of a residual is often often hampered by a lack of a suitable control system for regulating chlorine doses. In order to develop appropriate dosing strategies, it is first necessary to be able to describe the behaviour of chlorine residuals within the WDS. Traditionally, dynamic water quality modelling has involved coupling of water quality models to calibrated hydraulic models. Such models are useful for the design and analysis of a WDS, however they are expensive to construct and include assumptions regarding the complex processes that affect the decay of chlorine within the WDS. In contrast, a control-oriented approach aims to describe the dynamic input-output behaviour between an applied chlorine dose and the residual measured downstream at a strategic point within the WDS. Empirical modelling techniques are extremely useful for developing input-output models in the absence of a calibrated hydraulic model of the WDS, and without a great deal of a priori knowledge of chlorine decay processes. Recently, an empirical WDS modelling methodology based on artificial neural networks (ANNs) has been shown to produce models that accurately describe the non-linear inputoutput behaviour of chlorine residuals. This robust type of model has tremendous potential for implementation in intelligent adaptive control schemes. In this paper, the development of a control-oriented model is described for a single case study. The type of ANN model selected for this study was the general regression neural network (GRNN). The methodology applied to model development aims to ensure that the final model performs well and is therefore considered more rigorous than typical GRNN development methodologies. A comparison is made to a linear auto-regressive movingaverage (ARMA) model that is often proposed for control-oriented modelling. The GRNN was found to perform particularly well in this application (r² ~ 0.98) and this was far better than the ARMA model (r² ~ 0.85). This result indicates that the ANN approach to controloriented modelling has increased potential for use in the development of robust control systems for the regulation of chlorine residuals within a WDS. Such a control system will provide the benefits of process optimisation and cost reduction and improved water quality for customers.||-|
|dc.publisher||Australian Water Association||-|
|dc.title||Control-oriented water quality modelling using artificial neural networks||-|
|dc.contributor.conference||Enviro 04 (2004 : Sydney, Australia)||-|
|dc.identifier.orcid||Maier, H. [0000-0002-0277-6887]||-|
|dc.identifier.orcid||Dandy, G. [0000-0001-5846-7365]||-|
|Appears in Collections:||Aurora harvest 2|
Civil and Environmental Engineering publications
Environment Institute publications
Files in This Item:
There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.