Search

The water industry in the UK and worldwide has a pressing need to better manage interruptions to water supply caused by various failure events, such as pipe bursts, equipment failures or water treatment work shutdowns. One way of doing this is by making use of the increasingly available real-time sensor data collected in water distribution systems, as well as by using hydraulic models in real-time. Currently, real-time sensor data and real-time hydraulic modelling are not used much in a water utility's control room, especially when it comes to identifying a suitable strategy to respond to failure events in near real-time. This PhD project aims to develop, test/validate and demonstrate a new response methodology to support decisions made by control room operators when dealing with various failure events in a water distribution system. An integral part of this work is to develop an interactive decision-support tool, which will guide/support operators in identifying an effective response solution in near real-time (i.e. usually required up to 1 hour after the event detection/localisation if unforeseen events in the field are not considered). The tool will be used in this thesis to test and validate the response methodology. The proposed response methodology considers: (i) structured yet flexible approach supporting and guiding the operator throughout the entire response process to water network failure events, whilst allowing the operator to have a final say; (ii) novel interaction with the operator in near real-time via the proposed tool (e.g. allowing operators to propose different 'what-if' scenarios without being hydraulic experts); (iii) provision of automatically generated advices (e.g. near optimal response solutions via a novel heuristic algorithm and assessed end-impacts); and (iv) improved impact assessment. An integral part of the response methodology is the development of a novel method to identify near optimal response to failures in water distribution networks. The response problem is formulated as a two-objective optimisation problem with objectives being the minimisation of failure impacts and related operational costs. The heuristic-based method is developed and used to solve this problem. For the first objective (i.e. impact assessment), a new impact assessment method is developed, using realistic impact indicators that cover different aspects of the event - which are consistently calculated for every proposed response solution (to facilitate easy comparison between different response solutions). The response methodology was tested, validated and demonstrated on a semi-real case study. The implementation of the response methodology via the tool enabled operators to identify a response solution better (i.e. with lower end-impact and cost) than the solution based on the current response practice of utilities. The results obtained from this case study, demonstrate that the response methodology works well and that it has a potential to improve water utilities' current practice. The heuristic optimisation method that is integral part of the response methodology was validated and demonstrated on two semi-real case studies. Based on the results obtained it can be concluded that the heuristic-based method works well (i.e. it is reliable and robust) and is able to identify near optimal response solutions in a computationally fast manner. This, in turn, enables this method to be used in near real-time in real-life situations.

Novel technologies can help make progress towards the goal of universal and equitable access to safe and affordable drinking water for all. Rural communities in sub-Saharan Africa suffer from very low levels of water supply and drastic unsustainability of existing systems. The reasons for this include complex interactions between many factors; it is a 'wicked problem'. In recent years, a range of ICT and Internet-of-Things (IoT) innovations have been developed to address limitations to rural water supply. Specifically, pre-payment 'smart meter' innovations for communal water points allow for improved revenue collection, real-time monitoring, and improved access. This technology is in the early phases of deployment across sub-Saharan Africa. Current evaluations, particularly in grey literature, have had limited: focus on technical aspects, independent academic rigour, or systems thinking. Significantly, data collected from smart meters has not been used for longer-term insights into rural water supply. The aim of this research is to investigate how 'smart meters' can improve rural water supply service in sub-Saharan Africa. An evaluation of a selected smart meter model was carried out on technical robustness, specifically accuracy of flow measurement and flow rate reduction from debris, and on effectiveness in the socio-economic context in rural Tanzania. This revealed good technical and socio-economic performance. Next, smart meter data from Tanzania and The Gambia was combined with collected survey and interview data to reveal new insights into water collection patterns from rural water points. Then, smart meter data was combined with rainfall data, and the influence of rainy seasons and days of rainfall are quantified. Based on this, 'weather dependent pricing' (WDP) has been developed as a novel pricing mechanism based on smart meter pre-payment and remote management, with the objective to incentivise users to maintain using clean groundwater. Potential costs per disability adjusted life year (DALY) averted, economic cost-benefit, sensitivity, and uncertainty of the pricing mechanism have been quantified, and an associated decision support tool has been developed. WDP has the potential to be an 'off the shelf' intervention with good value for money that can strengthen rural water supply systems and contribute towards climate change resilience. Key findings from this PhD research include: showing flow measurement is accurate, and flow rate reduction from debris build up can provide a basis for predictive maintenance; empirically showing how smart meters can improve service levels in rural communities; that smart meter data can provide new insights into time and location of water collection, and seasonality; and that weather dependent pricing could significantly and cheaply improve community health. The research has strong impact potential for current and future smart meter development and deployment (an area growing in relevance), and starts to fill four important research gaps. Suitable settings, broader considerations, policy implications, and potential barriers for smart meters in rural sub-Saharan Africa are expounded, along with how smart meters can strengthen rural water supply systems.

The supply of drinking water in sufficient quantity and required quality is a challenging task for water companies. Tackling this task successfully depends largely on ensuring a continuous high quality level of water treatment at Water Treatment Works (WTW). Therefore, processes at WTWs are highly automated and controlled. A reliable and rapid detection of faulty sensor data and failure events at WTWs processes is of prime importance for its efficient and effective operation. Therefore, the vast majority of WTWs operated in the UK make use of event detection systems that automatically generate alarms after the detection of abnormal behaviour on observed signals to ensure an early detection of WTW’s process failures. Event detection systems usually deployed at WTWs apply thresholds to the monitored signals for the recognition of WTW’s faulty processes. The research work described in this thesis investigates new methods for near real-time event detection at WTWs by the implementation of statistical process control and machine learning techniques applied for an automated near real-time recognition of failure events at WTWs processes. The resulting novel Hybrid CUSUM Event Recognition System (HC-ERS) makes use of new online sensor data validation and pre-processing techniques and utilises two distinct detection methodologies: first for fault detection on individual signals and second for the recognition of faulty processes and events at WTWs. The fault detection methodology automatically detects abnormal behaviour of observed water quality parameters in near real-time using the data of the corresponding sensors that is online validated and pre-processed. The methodology utilises CUSUM control charts to predict the presence of faults by tracking the variation of each signal individually to identify abnormal shifts in its mean. The basic CUSUM methodology was refined by investigating optimised interdependent parameters for each signal individually. The combined predictions of CUSUM fault detection on individual signals serves the basis for application of the second event detection methodology. The second event detection methodology automatically identifies faults at WTW’s processes respectively failure events at WTWs in near real-time, utilising the faults detected by CUSUM fault detection on individual signals beforehand. The method applies Random Forest classifiers to predict the presence of an event at WTW’s processes. All methods have been developed to be generic and generalising well across different drinking water treatment processes at WTWs. HC-ERS has proved to be effective in the detection of failure events at WTWs demonstrated by the application on real data of water quality signals with historical events from a UK’s WTWs. The methodology achieved a peak F1 value of 0.84 and generates 0.3 false alarms per week. These results demonstrate the ability of method to automatically and reliably detect failure events at WTW’s processes in near real-time and also show promise for practical application of the HC-ERS in industry. The combination of both methodologies presents a unique contribution to the field of near real-time event detection at WTW.