Fiorillo, Diana (2021) Smart metering data for urban water demand modelling and Water Distribution Network management. [Tesi di dottorato]


Download (3MB) | Preview
[error in script] [error in script]
Item Type: Tesi di dottorato
Lingua: English
Title: Smart metering data for urban water demand modelling and Water Distribution Network management
Date: 12 July 2021
Number of Pages: 148
Institution: Università degli Studi di Napoli Federico II
Department: Ingegneria Civile, Edile e Ambientale
Dottorato: Ingegneria dei sistemi civili
Ciclo di dottorato: 33
Coordinatore del Corso di dottorato:
Giugni, MaurizioUNSPECIFIED
De Paola, FrancescoUNSPECIFIED
Date: 12 July 2021
Number of Pages: 148
Uncontrolled Keywords: smart meters; water demand; water distribution network; district metered area; water demand model; water distribution network management
Settori scientifico-disciplinari del MIUR: Area 08 - Ingegneria civile e Architettura > ICAR/01 - Idraulica
Area 08 - Ingegneria civile e Architettura > ICAR/02 - Costruzioni idrauliche e marittime e idrologia
Date Deposited: 19 Jul 2021 19:54
Last Modified: 07 Jun 2023 10:42


The mitigation of the impacts of climate changes and anthropic pressures on water security calls for effective strategies for sustainable water distribution networks (WDNs) management. Innovative metering technologies, such as smart metering, able to provide high resolution water demand data, represent a useful tool still not fully exploited. The general objective of this thesis is to develop new water demand models to improve WDNs management, as well as optimize the use of smart meters. To this purpose, smart metering data from the District Metered Area (DMA) of Soccavo (Naples, Italy) were used. A novel bottom-up methodology for the generation of demand time series of WDN users was developed. The methodology applies a copula-based re-sort to demand time series generated through a Beta or Gamma probability distribution. The methodology, applied to the literature case study of Milford (Ohio) and the Soccavo DMA, was able to reproduce the main statistics of measured demand time series and preserve their spatial and temporal cross-correlations. Thus, the methodology can effectively assist water utilities in design and management of WDNs by providing accurate estimates of water demand for numerical simulations of WDNs behaviour. Then, a comparative study of the performance of the proposed bottom-up methodology and a top-down one was carried out. The top-down methodology consists of a non-parametric disaggregation model based on the K-nearest neighbours approach. The comparison was performed by considering two case studies, both referred to the Soccavo DMA, with a different number of users. The bottom-up methodology performed better in reproducing cross-correlations between single users, and between single nodes. Whereas the top-down methodology was more effective in reproducing skewness and rank cross-correlations for spatially aggregated time series. For both methodologies, high levels of aggregation in nodes were found to be beneficial to preserve rank cross-correlations. Successively, the focus was set on the reconstruction of the total temporal demand pattern of a DMA. Such aspect is important for water utilities for a variety of operational tasks, e.g. for the detection of anomalous events, such as unauthorized consumption and leakages. To address this issue, two procedures are proposed that can be applied when smart meters, originally installed at all locations, have to be replaced and when no smart meter is present, respectively. The first procedure uses the stepwise regression for the selection of the smart meters to be replaced for accurately reconstructing the total demand pattern. The second procedure consists of applying different criteria, based on easily available data (e.g. consumption on the annual bill and user typology), to identify the set of representative users to be provided with smart meters. Then, a novel linear model based on users billed annual consumption is used to estimate the total DMA demand. The procedures were applied to the Soccavo DMA. In both cases the accuracy of the total demand pattern reconstruction was good already for low number of selected users. Identifying the users with the highest consumption, while distinguishing between the different categories of users, was the most effective strategy for the accurate reconstruction of the total demand. Overall, both procedures allow water utilities to optimize smart metering systems and reduce their costs. Considering the increasing importance of water demand forecasting in light of the future climate changes, the prediction accuracy of models based on weather variables and a common machine learning technic, i.e. the Random Forests, was also investigated. The analysis was carried out for the Soccavo DMA by disaggregating water consumption based on the social characteristics of the users. The models were able to forecast the aggregated daily water demand, though their performances changed depending on the social characteristics of the users. The obtained results are useful for assessing future variations in water demand due to climate variability, thus reducing risks of supply and operational failures in WDNs. In conclusion, the methodologies and the results presented in this thesis are expected to aid water utilities in developing more sustainable and cost-effective WDN management strategies.


Downloads per month over past year

Actions (login required)

View Item View Item