The Big Picture
CityLearn v3 lets teams simulate realistic neighborhood energy communities—including membership changes, equipment failures and local cost settlements—while recording both what controllers requested and what was actually applied so decisions can be fairly compared and audited.
ON THIS PAGE
Key Findings
A configurable simulator and evaluation framework preserves scenario rules (members, device limits, service windows and settlement) from configuration through to results. It supports running multiple communities together, offers a structured interface that keeps entity identities and availability, and records requested versus applied actions so you can see what a controller tried to do and what actually happened. Role-Based Agent Pattern benchmarks combining energy, cost and service indicators reveal trade-offs: lower cost or grid import can coincide with worse user-level service unless that service is explicitly enforced. orchestration
Test your agentsValidate against real scenarios
By the Numbers
1Smart/community rule-based controllers reduced annual electricity cost from 28.7 kEUR (business-as-usual) to 22.0–22.3 kEUR (≈22–23% lower).
2Grid import fell from 166.8 MWh (BAU) to 133.3–135.2 MWh under smart/community policies (≈19–20% reduction).
3Business-as-usual hit the requested EV charge target within ±5 percentage points only 6.3% of departures, while basic and smart/community policies reached ~98% and ~91% target proximity respectively, with minimum EV service at 100% for the basic policy and ~93.8% for smart/community.
What This Means
Control engineers and researchers can use the framework to test policies under realistic conditions (changing members, missing data, command failures) and to compare performance fairly. Multi-Agent Energy Grid Optimization Grid operators, municipal planners and aggregators can evaluate how settlement rules and electrical limits affect participant costs and delivered services before deployment.
Key Figures
Fig 1: Figure 1 : How CityLearn v3 represents one or more renewable energy communities.
Fig 5: Figure 5 : Time-series and KPI comparison views in the CityLearn UI.
Ready to evaluate your AI agents?
Learn how ReputAgent helps teams build trustworthy AI through systematic evaluation.
Learn MoreYes, But...
Results depend on the scenario configuration and time resolution: aggregating short traces to coarser steps can hide short power peaks. The framework focuses on realistic control experiments and comparisons rather than replacing highly detailed thermal or device-level models; domain-specific model accuracy still matters. Some real-world mechanisms (market rules, extreme failure modes, or local governance nuances) may require custom scenario extensions to reflect local practice. Consensus-Based Decision Pattern
Deep Dive
CityLearn v3 is a configurable simulation and evaluation environment for renewable energy communities that keeps scenario assumptions, constraints and service rules intact from setup through controller evaluation. It supports multi-community synchronized runs, a structured entity-based interface that keeps identifiers and availability, and a mechanism that stores both the requested controller action and the action actually applied after limits, availability and failures are enforced. Per-building and per-phase active-power limits are applied and timestep energy accounting is preserved so cost, energy and service indicators can be linked back to concrete controller decisions. Agent Service Mesh Pattern
Not sure where to start?Get personalized recommendations
Credibility Assessment:
Authors show low h-index scores (1–3), no recognized institutional affiliations provided, and the work is an arXiv preprint with no citations — limited established credibility.