Skip to main content
Skip to footer
We're taking you to another TCS website now.


 


Highlights

  • Organizations in the energy value chain are looking for innovative ways to improve operational efficiencies. ‘Smart grids’ are a step in that direction.
  • Presenting VidyutVanika, a reinforcement learning-driven energy broker, a bot that predicts energy output based on weather patterns and adapts to smart grid environments, working in a simulated electricity market.
  • TCS Research and International Institute of Information Technology, Hyderabad (IIITH)- collaborated to compete in a simulated energy trading competition―PowerTAC.
     


Why smart grids?

Sustainability is a non-negotiable goal for industries today. The world is fast moving towards a sustainable future with economies pursuing cleaner energy sources. However, industry partners including utilities face challenges as they make this transition and endeavour to meet green objectives.

Renewable energy cannot be produced on demand. Therefore, the industry needs intuitive solutions to predict supply for optimal performance. Moreover, the evolving energy ecosystem model involves players within and outside the traditional energy value chain, such as prosumers who produce their own power. This has further increased complexity in operations.

Organizations in the energy value chain are looking for innovative ways to improve operational efficiencies—‘smart grids’ are a step in that direction as they integrate various components in the ecosystem such as power generation, storage, and distribution to meet energy demands. 
 


The power of synergy: VidyutVanika

TCS Research scientists teamed up with students from the International Institute of Information Technology, Hyderabad, (IIITH), to develop VidyutVanika, a machine-learning-powered autonomous broker agent.

“‘VidyutVanika' translates to ‘electricity trader’ in Sanskrit, signifying the indigenous roots of the smart agent”, says Dr Sanjay Bhat of TCS Research.

The agent, essentially a bot, relies on reinforcement learning to predict outcomes (energy output) based on weather patterns, adapting to smart grid environments, working in a simulated electricity market. Based on the prediction, the agent tries to buy and sell electricity and generate economic value in the process. The team took a Q-learning-based approach and backed the machine learning algorithms with game theory to develop the energy trading solution.

Given the volatile nature of clean energy generation, especially wind energy, where output can be measured only at the end, the autonomous broker agent can be instituted to solve bidding challenges. Output prediction methods and cost-effective bidding inventions like these, enable the industry scale-up.
 


Drawing praise

The work was recognized at PowerTAC, an international competition, based on a simulated smart grid environment. VidyutVanika emerged victorious in the 2021 edition of the contest.

Launched in 2009, PowerTAC was designed as an AI-driven open-source platform to identify how digital information systems could help improve the use and profitability of sustainable energy.

The environment in PowerTAC mimics a town-based smart grid set-up with its own weather patterns, power ecosystem, pricing preferences, and generation and storage infrastructure. The competition requires participants to submit autonomous broker agents, which compete and bid in energy auctions. The smart agents are required to manage additional expenses, including distribution, capacity, and balancing charges.
 


A winning collaboration

The winning team was diverse in terms of the intellectual capital it brought to the collaboration, as it pooled in scientific minds from industry and academia.

“Collaborations of this nature help move the project forward at a greater pace due to deeper technical expertise, as well as novelty of collaborating with industry from a student perspective,” says Dr Pravin Paruchuri, Associate Professor at IIITH.

Dr Paruchuri, Dr Sujit Gujar, and Sanjay Chandelkar, of the IIITH team, have expertise in multi-agent systems, microeconomics, game theory and machine learning. Their strengths were complemented by the team from TCS Research, which included scientists—Dr Sanjay Bhat, Dr Easwar Subramanian, and Bala Suraj Pedasingu. The TCS Research team holds subject matter expertise in mathematical finance, computer science, artificial intelligence, and autonomous systems. While TCS Research developed the algorithms to bid in energy auctions for procuring electricity, IIITH created the solution that enables the sale of the energy procured.

"At a premier multi-agent conference, once the steering committee posed the question, ‘we have written so many impactful papers as a community, but mostly this impact is academic in nature. Have we really made an impact on society through our research?' This collaboration is one step towards it by appropriately engineering multi-agent and game theory-based solutions to deploy on practical test beds,” said Dr Sujit Gujar, Assistant Professor, IIITH, highlighting the essence of what collaboration can bring to the table.

The partnership between the research teams at IIITH and TCS Research dates back to 2017. At that time, TCS Research was exploring the possibility of diversifying into problems that involved multiple decision-making agents, an area of expertise of the Machine Learning Lab at IIITH. In addition to guest lectures being delivered by TCS Research scientists at the institute over the years, TCS has also supported the setting up of the Kohli Centre for Intelligent Systems at IIITH.

The partnership has also led to publications for conferences such as Association for the Advancement of Artificial Intelligence (AAAI).

The recognition VidyutVanika has received from the research community is a validation of the combined strength that collaboration can bring to any research endeavor.