Synthesis of a model for optimising a potable water treatment plant and water usage analysis in the Ugu District
Access to clean and adequate water is a universal and basic human right that feeds into the 6th of the 17 Sustainable Development Goals (SDGs). This goal aims at ensuring availability and sustainable management of water and sanitation for all. Clean water is referred to as potable water, which is safe for human consumption and offer low risk of immediate or long term harm. Raw water undergoes rigorous processing which consists of coagulation, sedimentation, filtration, disinfection and storage, to produce potable water. Each module or stage consumes chemicals and energy resources and thus incurs costs. To achieve the aim of the study, which was to synthesize an optimised potable water treatment network and a water usage analysis model, the Umzinto Water Treatment Plant (UWTP) and its distribution system was used as the study area. This treatment plant is located within Umdoni, a local municipality of the Ugu District Municipality in KwaZulu-Natal Province, South Africa. This study’s objectives were fourfold and the first objective was to identify and quantify key raw water quality parameters affecting treatment at the UWTP. The second objective was to design a genetic algorithm for the potable water treatment process control. The third objective was to evaluate the Umzinto Water Distribution System’s Non-Revenue Water (NRW) while the fourth objective was to develop a model for water usage analysis. For the first objective, data for water quality parameters for the water treatment from July 2006 to June 2013 were statistically analysed. This data were collected from the UWTP’s historical records. To improve the data’s integrity it was pre-processed using cubic hermite interpolation. After the pre-processing trend lines and box plots were used to determine the parameters’ significance compared to the standard values stipulated in the South African National Standard (SANS 241). The trend lines were used to analyse the frequency of observations that were higher than the standard values according to SANS 241. The box plots were used to determine the minimum, median, maximum and mean of the data sets. The mean values for each parameter were compared to the SANS 241 value to determine their significance. The raw water quality parameters were then correlated to the chemical dosages for lime, polymer, potassium permanganate and chlorine. The key parameters selected from the correlation analysis were algal count, manganese (Mn), iron (Fe), Escherichia coli, total coliforms, colour, odour, conductivity, turbidity, suspended solids (SS), pH, temperature, total organic carbon (TOC,) and Hardness. A number of methods can be used to achieve such optimisation, including artificial neural networks, dynamic programming, linear and non-linear programming, and this study utilised a genetic algorithm as an optimisation tool to achieve the second objective of optimising water treatment at the UWTP. For the model development, data from the correlations obtained for objective 1 were used. The model was aimed at reducing the cost of chemical dosage and four chemical dosage prediction models were developed using genetic algorithms and these were then used to produce a combined chemical dosage cost prediction model. The programming interface utilised for these models was Matlab. In developing these models, the data were first pre-processed to remove outliers and fill in the blanks using a Microsoft Excel Add-in that was developed for this particular purpose. The next step involved a curve fitting exercise in Microsoft Excel 2013. Matlab was then used to code the genetic algorithm that combined and optimised the solutions obtained from the curve fittings. The results showed that genetic algorithms can be reliably used to predict the chemical dosages and hence reduce water treatment costs. After treatment, water is pumped into the distribution system for consumption. It is therefore important to ensure that all the pumped out treated water reaches the consumer. The third objective therefore assessed the NRW for the Umzinto Water Distribution System for the period between July 2013 and June 2014. The data used for this objective was provided by the Ugu District Municipality. The method used combined the top-down approach and the component-based approach. This combined approach was modified to enable the calculation of all the components that are required in a standard South African Water Balance. The results showed that the distribution system had a high value of NRW, which was 27.9% of the System Input Volume. The major component of NRW was Real Losses, that is, losses that can be mitigated by improving maintenance. The fourth objective was to develop a model for water usage analysis that would reduce the time taken to evaluate NRW and also improve the analysis of the NRW components using Microsoft Visual Basics 2012 and Microsoft SQL Server 2012 development interfaces. The Visual Basics enabled the development of a graphic user interface that was user-friendly and minimised the time taken to learn the software. The software platform developed was able to import the data required to construct a standard International Water Asssociation (IWA) Water Balance, calculate all the components of NRW, store historical data for the water distribution systems and report on a rolling year basis. A model for water usage analysis was developed and made available for usage by practitioners in Ugu District. The model was developed for the specific study area and further studies would be required in order to validate it in a different setting. The results obtained for the first objective led to the conclusion that, there was very high pollution emanating from communities and activities close to the raw water sources, especially the EJ Smith Dam. The results from the first objective were also used to determine parameters for the models developed in the second objective. From objective two it was concluded that genetic algorithms can be reliably used to predict chemical dosages and hence reduce water treatment costs. The third objective’s results showed that 27.9% of treated water pumped into the distribution system is NRW. Which is a concern because 65% of this are real losses which have maintenance related problems. The fourth objective’s results showed the practicality of designing model that could be used determine all the important components of NRW that would take time to evaluate manually. It would also store historical data for the water distribution system and report on a rolling year basis. Implementation of this software would help minimise the errors associated with manual calculation of NRW and improve the availability of data for research and analysis. From the research findings, it is recommended that the treatment plant should change the way it is dosing chemicals in the balancing tank. The method currently being used is prone to error. The analysis of NRW showed that Real Losses were a major challenge in the Umzinto Distribution System. There is need to develop a maintenance program to cater for leakage. Communities also need to be educated on the importance of reporting leakage in the network.