CityLearn Challenge 2022 Launched

PV+Battery storage are the cornerstone of virtual power plants to provide grid flexibility. But what is the optimal charging/discharging policy to reduce costs and emissions from buildings ? And can you transfer knowledge/policies from one building to others?

The CityLearn Challenge 2022 is now live as an official NeurIPS 2022 competition. The task this year is to control a set of electrical batteries in 17 single family homes (with PV) to reduce electricity costs and grid emissions from the operation of the homes.

The gym environment for this challenge is based on a real net zero energy neighborhood case study from California, the data are provided by EPRI. Any method can be applied, RL or MPC. As the competition progresses, the agents are evaluated on more unknown buildings.

You can sign up your team and get started on…

(warmup ends August 15, competition lasts until October 30) Prizes this year include USD 15,000 in prize money, travel grants to ACM BuildSys/e-Energy, Pres of the results at NeurIPS22.

Big thanks to IEL’s awesome PhD student Kingsley and Dipam for AICrowd Research for preparing and releasing this competition!

Zoltan Nagy
Zoltan Nagy
Assistant Professor

My research interests include reinforcement learning for buildings and smart cities.