AI, machine learning to help save energy

AL AIN: A new project which uses artificial intelligence and machine learning approaches to predict the energy consumptions of varying applications is currently being developed at the United Arab Emirates University.

Its aim is to use the cloud computing resources in an intelligent manner, to optimise energy consumption, performance and the quality of services in data centres.

“The cloud computing platform serves as a backbone for smart cities,” said Dr Leila Ismail, Associate Professor at the College of Information Technology. “It is highly solicited by smart cities applications from big data analytics and IoT to AI applications and interactive web applications. And because it is solicited by smart cities applications, the energy consumption of its resources becomes an issue in increasing electricity bills, which is reflected on the users as a cost, and adds environmental issues.”

However, the performance of the applications is not compromised. In the project, entitled An AI-based System for Smart Cities Cloud Computing, researchers use AI and machine learning to predict the energy consumption and the performability of smart cities applications. They integrate them within an automatic framework to affect applications to the resources in a way to minimise the energy consumption without compromising performance. “In our project, we observe and learn from the behaviour of different applications running on different architectures and their corresponding performance and energy consumption,” she said. “We then use these data to develop a brain for the intelligent cloud that I call the green operating system for smart cities. With our system, we can reduce clouds electricity bills, be environment friendly while not compromising the quality of services and an application’s service level agreement.”

With smart cities, the increasing number of applications and the variety of application types, a manual administrative intervention is no more effective. The new system will learn from energy and performance data generated in the cloud and learn by itself without administrative intervention. “The idea is to use cloud computing with AI to increase the power of the cloud for performance and energy savings, and decrease administrative intervention,” she added. “The project needs more experimental verification but there’s been lots of progress.”