Publications and applications

CityLearn is used to study building and community energy management, controller design, demand response, comfort and resilience. The studies below are grouped by application. Each citation opens the full entry in References.

Project papers

  • CityLearn v1.0 introduces the control environment [26].

  • Standardizing research describes the multi-agent demand-response benchmark [28].

  • CityLearn v2 presents energy-flexible, resilient, occupant-centric and carbon-aware community management [11].

  • CityLearn v3 presents configurable simulation and service-aware evaluation for realistic renewable-energy-community control studies. Available as an arXiv preprint [5].

For ready-to-copy BibTeX entries, see Cite CityLearn. Record the software version and scenario used alongside the scholarly reference.

Studies using CityLearn

Application

Examples

Benchmarking reinforcement learning algorithms

Dhamankar et al. [4], Khayatian et al. [8], Nweye et al. [13], Pinto et al. [20], Qin et al. [22]

Coordinated energy management

Deltetto [2], Glatt et al. [6], Kathirgamanathan et al. [7], Pinto et al. [19], Pinto et al. [21], Qin et al. [23], Vazquez-Canteli et al. [29], Yang et al. [30]

Incentive based demand-response

Deltetto et al. [3], Tolovski [24]

Independent energy management

Chen et al. [1], Nweye et al. [14]

Meta-learning

Zhang et al. [31]

Model predictive control

Zhan, Sicheng et al. [32]

Transfer learning

Nweye et al. [15]

Voltage regulation

Pigott et al. [18]

Related research includes extensions such as GridLearn. Consult each paper for its environment version and additional models.

Teaching material

The reinforcement-learning tutorial for grid-interactive buildings and communities is linked from Tutorials. Start with Your first simulation for an executable CityLearn v3 introduction.