Cities consume over 75% of the world's energy and generate more than 70% of global carbon emissions. As urban populations grow and climate commitments intensify, the imperative to optimize city-scale energy systems has never been more urgent. Smart meter deployments now generate unprecedented visibility into consumption patterns—but visibility alone does not drive optimization.
The challenge is not data scarcity. Advanced Metering Infrastructure (AMI) across a major metropolitan area can produce hundreds of millions of readings daily. The challenge is transforming this deluge of meter data into coordinated, intelligent action—demand forecasting that anticipates rather than reacts, load balancing that prevents rather than responds to stress, and renewable integration that maximizes rather than accommodates clean energy.
While smart meters provide city-wide visibility, real-world action occurs inside Building Management Systems (BMS) and Power Management Systems (PMS).
BMS platforms control HVAC, lighting, lifts, fire systems, and water pumps across commercial and public buildings. PMS platforms manage electrical distribution, transformers, generators, UPS systems, and feeder-level constraints.
Without connecting smart meter intelligence to BMS and PMS, city-scale optimization remains analytical rather than operational.
This blog explores how ElixirData and NexaStack enable city-scale energy optimization by building the Urban Energy Intelligence Layer—a reasoning infrastructure that transforms smart meter data into the foundation for agentic grid management.
What is the Urban Energy Intelligence Layer?
The Urban Energy Intelligence Layer integrates data from smart meters, BMS, and PMS, transforming it into actionable insights for optimized, coordinated city energy management.
The original business case for smart meters centered on operational efficiency: automated meter reading, remote service connection, and improved billing accuracy. These benefits have been realized. But they represent only a fraction of the value that AMI infrastructure can deliver.
A modern smart meter is not merely a consumption recorder—it is a continuous context source. Every 15-minute interval reading carries information about occupancy patterns, equipment behavior, weather response, and grid conditions. Aggregated across thousands of meters, this data reveals the rhythms of urban energy: morning ramp-ups as cities wake, afternoon peaks as cooling loads climb, evening shoulders as residential consumption rises while commercial loads fall. These turns into drivers of BMS/PMS Actions like informs HVAC pre-cooling, load shedding, equipment scheduling and informs transformer loading, generator dispatch, feeder balancing.
The limitation of traditional approaches is that this rich context remains fragmented. Meter data flows into utility billing systems. Weather data sits in separate feeds. Grid telemetry lives in SCADA historians. Tariff structures exist in rate management systems. No single system holds the complete picture needed for intelligent optimization.
This is precisely the problem ElixirData solves. As the Context OS, ElixirData unifies these disparate data streams into a coherent, reasoning-ready representation—what we call the Urban Energy Graph.
Why are smart meters crucial for energy optimization in cities?
Smart meters provide continuous context, allowing for more accurate, real-time optimization of energy consumption and grid management, beyond just recording data.
The Urban Energy Graph is ElixirData's unified model of city-wide energy context. It is not a data warehouse or a time-series database—it is a structured representation of entities, relationships, and decision-relevant state that agents can reason over.
Context Nodes: What the Graph Captures
Nodes alone are insufficient for reasoning. The power of the Urban Energy Graph lies in the relationships it captures:
This graph structure is what transforms raw data into reasoning-ready context. When an agent needs to evaluate a load-shifting opportunity, it does not query multiple databases and correlate results—it traverses the graph to understand the complete context: current load, predicted demand, grid constraints, tariff implications, and historical response patterns.
With the Urban Energy Graph providing contextual intelligence, NexaStack deploys the agents that transform understanding into action. Here are the primary use cases for city-scale energy optimization:
Platform Responsibilities by Use Case
|
Use Case |
ElixirData Role |
NexaStack Role |
|
Predictive Demand Forecasting |
Aggregates meter patterns, weather correlations, event calendars into forecast context |
Deploys forecasting agents, serves predictions to grid operators and downstream systems |
|
Dynamic Load Balancing |
Identifies zone-level imbalances, calculates redistribution options, evaluates constraint feasibility |
Orchestrates load shifting across zones via utility APIs and control systems |
|
EV Charging Coordination |
Models charging demand, grid capacity, user preferences, and tariff optimization opportunities |
Executes optimized charging schedules, coordinates with charging network operators |
|
Peak Demand Management |
Detects peak approach, identifies flexible loads, calculates curtailment options |
Triggers demand response actions, dispatches notifications to enrolled participants |
|
Renewable Integration |
Tracks solar/wind variability, models storage capacity, predicts generation windows |
Dispatches battery storage, adjusts grid imports, coordinates with generation assets |
Traditional demand forecasting relies on historical load curves adjusted for weather and calendar effects. This approach captures broad patterns but misses the granular dynamics that drive real-time optimization opportunities.
ElixirData enables a fundamentally different approach. By maintaining the Urban Energy Graph with continuous context updates, forecasting agents can reason about demand at multiple levels simultaneously:
NexaStack's forecasting agents consume these multi-level predictions and serve them through APIs to grid operators, energy traders, and automated control systems. The result is anticipatory grid management—positioning resources before demand materializes rather than scrambling to respond after the fact.
Urban grids are not uniform. Some neighborhoods experience afternoon peaks from commercial cooling; others see evening peaks from residential cooking and entertainment. Some feeders approach capacity limits while adjacent feeders remain underutilized. These imbalances create both reliability risks and optimization opportunities.
ElixirData's Urban Energy Graph captures the spatial distribution of load in real-time, overlaid with the physical network topology. When imbalances emerge, the decision plane evaluates options:
The graph identifies EV chargers, battery storage, and enrolled demand response participants.
The graph knows which distributed resources are available and at what cost.
The graph models transformer limits, feeder ratings, and voltage constraints.
NexaStack's load balancing agents then execute the approved strategy via PMS (feeder limits, transformer protection) and BMS (HVAC setpoint changes, equipment staggering) and it sends control signals to enrolled devices, adjusting charging rates, or dispatching storage—while ElixirData logs the complete decision lineage for operational review and regulatory compliance.
Electric vehicle adoption is transforming urban energy patterns. A single EV charger can draw 7-19 kW—equivalent to adding another house to the grid. Unmanaged charging concentrated in evening hours threatens to create new system peaks precisely when grids are already stressed.
But EVs also represent unprecedented flexibility. Most vehicles sit parked for 20+ hours daily, and most charging sessions can tolerate delays without affecting driver needs. This flexibility is valuable—if it can be orchestrated intelligently.
ElixirData models each charging session with its full context: vehicle arrival time, departure deadline, required energy, charger capacity, local grid constraints, and current tariff period. The decision plane calculates optimal charging profiles that satisfy driver needs while minimizing grid impact and energy cost.
NexaStack's EV coordination agents communicate these profiles to charging stations, adjusting rates dynamically as conditions change. If an unexpected grid event occurs, charging can be curtailed temporarily. If renewable generation exceeds expectations, charging can be accelerated. The result is a fleet of mobile batteries that supports rather than stresses the grid.
How does ElixirData coordinate EV charging with grid needs?
ElixirData models EV charging sessions to optimize their timing and energy use, helping to avoid grid peaks and align charging with renewable energy availability.
City-scale optimization requires coordination across multiple agents operating at different scales. A demand response event, for example, might involve thousands of buildings, tens of thousands of devices, and millions of dollars in grid value. No single agent can manage this complexity—but a coordinated multi-agent system can.
NexaStack orchestrates a hierarchy of agents, each with defined scope and authority:
|
Agent Type |
Scope |
Responsibilities |
|
Grid Agent |
Entire service territory |
System-wide stability, wholesale market interaction, aggregate demand response, manages feeders, substations, generation assets via PMS |
|
Zone Agent |
Geographic district or feeder |
Local load balancing, constraint management, distributed resource coordination |
|
Building Agent |
Individual facility |
HVAC optimization, lighting control, on-site generation and storage via BMS |
|
Device Agent |
Individual asset (EV, battery, equipment) |
Asset-level optimization within constraints set by higher-level agents |
|
Coordination Agent |
Cross-agent |
Conflict resolution, goal alignment, resource allocation across agents |
The critical enabler of this coordination is ElixirData's shared context graph. All agents reason over the same Urban Energy Graph, ensuring decisions are based on consistent, current information. When the Grid Agent calls a demand response event, Zone Agents can immediately see the system-wide context. When a Building Agent adjusts its load, the impact propagates through the graph for Zone and Grid Agents to incorporate.
This shared context prevents the coordination failures that plague traditional demand response: conflicting signals, cascading over-response, and the inability to verify actual load reduction. Every agent sees the same reality, and every action is logged with full lineage.
Grid operators, utility executives, and regulators share a common concern about AI-driven optimization: accountability. When an autonomous system makes decisions that affect millions of customers and critical infrastructure, stakeholders need to understand what happened and why.
ElixirData's decision lineage capability provides this accountability. Every optimization action—every load shift, every demand response dispatch, every charging rate adjustment—is logged with complete context:
The relevant portions of the Urban Energy Graph at decision time are captured, including meter readings, grid state, weather conditions, and tariff status.
The decision plane evaluates multiple options; lineage records what was considered and why the selected action was preferred.
Every decision carries a confidence score based on context completeness, model certainty, and historical accuracy.
Post-decision metrics are linked to the decision record, enabling continuous learning and performance verification.
This lineage serves multiple purposes. Operations teams can investigate unexpected outcomes. Compliance officers can demonstrate regulatory adherence. Performance analysts can identify improvement opportunities. And executives can trust that AI-driven optimization is governed, not autonomous in the ungoverned sense.
How does ElixirData ensure transparency in AI-driven decisions?
ElixirData logs every decision, including context, alternatives considered, confidence levels, and outcomes, enabling continuous learning and performance verification.
Smart cities generate millions of data points daily from meters, sensors, and grid systems. ElixirData structures this into the Urban Energy Graph—a living, reasoning-ready model of city-wide energy flows. NexaStack deploys the agents that act on this intelligence, coordinating across meters, buildings, EV chargers, and grid infrastructure.
The result is not just smarter monitoring—it is autonomous, governed optimization at urban scale.
Deploying the Urban Energy Intelligence Layer requires thoughtful integration with existing utility systems and processes. Here are key considerations:
ElixirData connects to AMI head-end systems through standard protocols and APIs. Most modern AMI platforms support data export via web services, message queues, or direct database connections. Weather data integrates through commercial APIs. SCADA integration typically uses OPC-UA or historian APIs.
The key is establishing real-time or near-real-time data flows. Batch processing that was acceptable for billing is insufficient for optimization. ElixirData's ingestion layer is designed for streaming data at scale—millions of meter readings processed continuously.
NexaStack's agents need pathways to effect change. For demand response, this typically means integration with demand response management systems (DRMS) or direct communication with enrolled devices through OpenADR or proprietary protocols. For EV charging, integration with charging network operators' APIs enables schedule optimization.
Importantly, NexaStack integrates with existing BMS controllers, PMS controllers, SCADA, and DRMS systems, ensuring no bypass of safety-critical infrastructure. The utility's SCADA system remains the authoritative control layer for critical grid operations. NexaStack provides optimized setpoints and recommendations; existing systems execute and enforce safety limits.
We recommend a phased approach that builds confidence progressively:
Phase 1 — Context Foundation: Deploy ElixirData with meter data integration, build the Urban Energy Graph, and establish baseline forecasting accuracy.
Phase 2 — Monitoring Agents: Deploy NexaStack agents in monitoring mode, generating recommendations without automated execution.
Phase 3 — Controlled Execution: Enable automated execution for low-risk, high-confidence decisions while maintaining human approval for significant actions.
Phase 4— Expanded Autonomy: Progressively expand agent authority as performance is validated and operator confidence grows.
The smart meter revolution promised to transform how utilities understand and manage energy. That promise remains largely unfulfilled—not because the data is lacking, but because the infrastructure to reason over that data has been missing.
ElixirData and NexaStack provide that infrastructure. The Urban Energy Graph transforms fragmented meter data into unified, reasoning-ready context. Agentic optimization converts that context into intelligent, coordinated action. Decision lineage ensures every action is transparent and accountable.
The result is not incremental improvement in grid operations. It is a fundamental shift from reactive grid management to anticipatory optimization—from dashboards that show what happened to systems that determine what should happen and make it so.
In the next blog, we will move from city-scale to building-scale optimization, exploring how ElixirData and NexaStack transform traditional Building Management Systems into Intelligent Management Systems capable of autonomous, governed energy control.
The Urban Energy Intelligence Layer, powered by ElixirData and NexaStack, transforms fragmented smart meter data into actionable insights for city-scale energy optimization. By unifying data from smart meters, BMS, and PMS into the Urban Energy Graph, cities can move from reactive to anticipatory energy management. This intelligent, coordinated approach enhances energy efficiency, reduces costs, and supports a more sustainable grid. The result is a smarter, more resilient urban energy ecosystem that benefits both city operators and residents alike.
Download our whitepaper on city-scale energy optimization with agentic AI, including detailed architecture diagrams and integration patterns for AMI, SCADA, and DRMS systems.
Contact XenonStack to schedule a demonstration of the Urban Energy Graph and multi-agent coordination capabilities.
← Previous: Blog 1 — From Reactive Energy Systems to Agentic Energy Intelligence
→ Next: Blog 3 — Agentic AI for Intelligent Buildings: Agentic EMS and IMS for Autonomous Energy Control