Experiences in Smart Water - the Singapore Story

If you were thinking about countries and cities that were at the forefront of innovation in the water sector, would Singapore immediately come to mind?

Singapore has long been researching and implementing methods to conserve water, to reuse water and to work with citizens and the community on understanding their needs - both now and in the future. In many ways, the challenges that Singapore faces are the challenges of the future - a small city with limited access to natural resources, including water; a high-tech economy that provides its citizens with the comforts and benefits associated with a developed country; and a changing climate that is impacting its ability to deliver those benefits.

So, let’s talk about how Singapore is using the latest in data science and machine learning to help solve its water problems!

Singapore’s Public Utilities Board (PUB) is the national water agency responsible for “supplying good water, reclaiming used water and taming stormwater”. As an agency, it is responsible for 17 surface water reservoirs, 4 water reclamation plants, 3 desalination plants, 6 waterworks, 8000 km of drains, rivers and canals, 5500 km of pipes, 48 km of deep tunnels, 3500 km of public sewers, 90000 manholes and 1.5 million customer accounts. The interesting aspect of the Singapore experience is that as a small city state, the PUB is effectively responsible for the entire watershed - which is very different from many countries where stormwater, wastewater, drinking water and water resources management are governed by different entities. That makes it easier to evaluate the entire system and deploy solutions that are beneficial to the entire system - not just certain parts of it. 

About five years ago, the PUB identified the need to implement “smart water technologies” to meet the needs of increasing water demand, rising costs and future challenges in water management due to climate change. “Smart water” covers many areas - primarily efficient monitoring of plant operations and the hydrologic basin through sensors and digital twins, automation of several menial tasks such as cleaning and monitoring pipes using drones and robots, improving the customer experience through easy access to information electronically, and enabling a 360o view of the entire system so that decisions can be made quickly, efficiently and taking the needs of the whole system into account. 

Several pilot-scale trials were conducted across the system for 1) risk management models and automated sensors for pipeline failure 2) process based, statistical and machine learning models for water resources management and flood prediction 3) algorithms integrated with sensors and analytics to monitor and measure leaks in the system 4) a network of high quality sensors including biosensors such as live tiger barbs to monitor water quality in real-time and respond to events and threats to the water system and 5) high resolution sensors to measure water use by customers at a fine scale and implement water conservation measures as required.

Many of these systems were deployed in addition to or as replacements of the manual methods that were previously in place. For example, leaks and high-risk pipelines were traditionally detected by field crews that surveyed the water network - a process that took approximately a year to complete. With the introduction of automated, high-resolution sensors along the pipe and models that combined physics-based models of corrosion, pressure and flow together with statistical and machine learning techniques to detect anomalies -  the network could be monitored much more frequently, threats detected earlier and system components repaired or renewed earlier to ensure the safety and security of water supplies. 

Another example is the automation of monitoring and replacement of pipe components through drones and robots. UAVs with cameras, GPS, sensors and robotic arms enabled the monitoring of areas that were unsafe for humans to venture into and easy replacement of parts by the robot with humans overseeing. We’ve seen quite a few startups globally venture into the fields of robotics for water and sewer system management - they typically combine image processing algorithms with water quality sensors and statistical models to understand what the robot is seeing in the pipes or tunnels and figure out what to do about it. Nexusbit and GenRobotic Innovations are some of the startups whose products are being used in different cities in Asia - an interesting harbinger of the future!

At the end of the pilot trials, the PUB found that they were able to detect several illegal dumping events of heavy metals that impacted water quality, identify threats to water quality from algal blooms in reservoirs and transform dangerous manual inspections of deep tunnels into safer, automated robot systems.

However, they also identified several challenges in the deployment of these smart water systems - chief among them being the lack of a workforce that is familiar with and expert in the use of these methods. Without professionals with the skills to understand the water system and the data science skills to develop the models, it can get challenging to deploy these systems quickly and at scale. The second challenge is related to the inter-operability of the sensors, robots and software systems. With so many startups and large companies in the space - each with their own proprietary technology-  getting the different types of sensors and automated systems to talk to each other and integrate the data from each other is a significant undertaking! Finally, working with the community and the public to develop trust in and understand the value of these systems, means that deploying smart systems at scale will take time and effort.

As water utilities and agencies around the world learn from each other and attempt to deploy these technologies to make people’s lives better, we need professionals in all of these sectors who have a combination of skills - both the knowledge of the water system and an understanding of how, when and where data science and machine can be applied. So, if you’re a water professional interested in learning about data science - join us for our first, cohort based course on “Getting started with data science and machine learning in the water and wastewater sector”.

What our community are reading

Moonshots, Models, IoT and Machine Learning in Agriculture

Our online community space is now live!

How much water should an email consume? Data centers and water use