AI innovation could help consumers understand water use with minimal disruption
28 January 2020
An innovative application of the latest Artificial Intelligence (AI) technology has the potential to bring unprecedented insight into residential and industrial water usage.
Water is an increasingly valuable commodity and its scarcity is increasing in many regions of the world. The OECD projects that world water demand will increase by 55% between 2000 and 2050. It is not surprising then, that countries are beginning to pay more attention to their water security. With Singapore's total water demand set to almost double by 2060, the city state has become a globally renowned hub for technology partnerships as it seeks to guarantee its own water security.
To address this issue, suppliers are seeking to access more detailed usage data. More data can help detect leaks, bottlenecks and inefficiencies faster and at a higher resolution. More data can unlock new insights, helping them to plan for future demand. More data can help consumers understand their usage at a more granular level, helping them to become aware of unnecessary usage and “drips”.
The last point is possibly the most interesting, and definitely the most challenging. Would consumers be willing to pay for, install and maintain a sensor on each of their water-using appliances to better understand their usage? Unlikely. The hassle, additional cost and maintenance requirement is likely to make this solution unfeasible, despite the benefits. So, how can technology enable appliance level usage data for consumers at the lowest possible cost and with minimal disruption to them?
Cambridge Consultants have developed a working prototype called AquaML. It uses a combination of a single pressure and flowrate sensor with a sophisticated machine learning algorithm to disaggregate water usage in the home. AquaML simulates a domestic water network with four outlets: two taps, a solenoid valve and a shower. When an outlet is opened and closed, a display shows the user which outlet was opened and how much water was dispensed from it, in real time.
This solution works on the principle that each appliance in a water network has its own unique pressure response “signature” when opened and closed. This signature depends on its location in the network and its design. The location of the outlet affects the resonance of the pressure signal; there are often different sizes and lengths of pipes between the outlet and the sensor. Its design affects how it is opened and closed, which can introduce some unique features. For example, the signature of the solenoid opening/closing is abrupt when compared to the signature of the screw tap on a shower.
Such a system could already provide the desired insights at significantly lower costs than installing multiple sensors. But it could be made even simpler and cheaper by employing the local, real time data processing of “AI at the edge”. This uses relatively simple, low-cost and low-power devices to perform complex AI tasks without needing to send data to the cloud. Cambridge Consultants expect to see AI at the edge unleashing a new wave of applications, from smart cities through to predictive maintenance on industrial machinery.
Contact Details and Archive...