Fetch.ai’s real-time autonomous energy management system delivers major cost savings
- Trial shows how Fetch.ai autonomous agents can help Warwick University reduce its energy costs
- Fetch.ai’s machine learning solution suggests daily energy costs would fall by 13–18% on the student campus
- Future real-world use cases will demonstrate the benefits of using the Fetch.ai energy management system to tackle more complex energy portfolios, such as those generating renewable energy
Fetch.ai’s technology has demonstrated it can significantly reduce energy costs at Warwick University. The innovative energy management system was shown to reduce the daily cost by 13–18% and increased the electricity generated on-site by 31%. It also reduced electricity imports from the National Grid by 47%. The trial’s success represents the latest development in a long-term partnership between Warwick Business School and Fetch.ai.
In an effort to find out how the university’s energy needs could be met at the lowest possible cost, Warwick University gave Fetch.ai access to two years’ worth of energy usage and price data generated by 16 sensors across the student campus. This gave Fetch.ai’s machine learning scientist Dr Yujian Ye and his team 200,000 data points to use. The energy management strategy of each asset was then optimized using a multi-agent deep reinforcement learning algorithm.
Advancements in AI and machine learning are enabling key parts of the university’s energy management system to become responsive to evolving circumstances in ways not previously considered possible. This is opening up new opportunities to help the university reach its ambitious target of producing ‘net zero’ emissions by 2030.
Fetch.ai’s autonomous energy management system is able to monitor and actively control the consumption, generation, conversion and storage of energy on the campus in real-time. By optimizing these factors, the management system is able to ensure it meets the university’s energy needs in the most economical way possible.
Warwick University has two energy centres. In Fetch.ai’s multi-agent energy management system, each centre is represented by an autonomous agent. Each centre has a similar, yet distinct, energy portfolio and the need to optimize the energy usage of the individual devices (e.g. the combined heat and power engine) requires each agent to adopt a bespoke approach. However, the autonomous agents must also collaborate. They share information aimed at reducing the total cost of meeting the campus’ energy requirements, as well as information about the specific supply and demand for heat, as shown by the orange arrow below:
Figure 1: The structure of Warwick University’s energy system
Fetch.ai’s energy management system uses a multi-agent deep reinforcement learning model featuring an actor-critic architecture. An actor neural network is used to optimize the energy management decisions for each agent representing an energy centre. Both actor networks receive feedback from a common critic neural network regarding the success of their strategies, based on the total energy costs of the campus. This enables the Fetch.ai agents to constantly update the energy management strategies they are adopting so as to best respond to changes in demand and grid prices, and to reduce the overall cost of supplying energy to the campus. This learning process is outlined in Figure 2:
Figure 2: The actor-critic architecture of the Fetch.ai energy management system
By utilizing reinforcement learning, the two autonomous agents in Fetch.ai’s energy management system learn to use the Combined Heat and Power Engines in Figure 1 more efficiently. As a result, more energy is able to be generated on site, so less energy needs to be acquired from the grid. This reduces the overall cost of meeting the energy demand across the campus.
The success of the trial at Warwick University has shown how reinforcement learning can save businesses thousands of pounds through enhanced energy usage and storage measures. By scaling up its energy management system, Fetch.ai intends to demonstrate how its approach can be applied to more complex energy portfolios that include renewable energy sources and electrical vehicles. Look out for more details regarding AI-enabled optimal energy management systems in the coming months.
At a time when real-time, sustainable energy management is receiving unprecedented attention from companies, consumers and legislators, the potential economic benefits of Fetch.ai’s technology are immense.
Get Involved
If you’re a developer, find out more about Fetch.ai’s Autonomous Economic Agents by checking out the documentation section of our website, which includes some demos.
- Watch videos outlining our technology and follow our tutorials on YouTube
- Join our Developer Slack Channel and explore our growing GitHub repository
- If you’re not a developer, you can stay up to date with our progress by joining us on Telegram or by following us on Twitter