Executive Summary

Physics AI™ System Architecture

Table of Contents

Chapter 1: The Interconnection Bottleneck

1.1 Scale of the Crisis

The US interconnection queue represents the single largest bottleneck to clean energy deployment. As of 2025, more than 2,600 GW of generation and storage capacity sits in interconnection queues across the country — roughly twice the entire installed US generation fleet. Of that, approximately 1,200 GW is solar, representing hundreds of billions of dollars in potential investment waiting for grid access. The Physics AI™ engine transforms this chaos into quantified, site-specific risk intelligence.

The numbers are stark: average wait times have ballooned to 4–5 years in most regions, with some ISOs reporting queues exceeding 7 years. More critically, over 70% of projects that enter the queue never reach commercial operation — they withdraw due to escalating interconnection costs, study delays, or network upgrade requirements that destroy project economics. Only approximately 15% of queued projects ultimately go operational.

1.2 Impact on Solar Developers

Interconnection queue congestion creates three cascading risks for solar developers:

• Timeline Risk — Queue delays push Commercial Operation Dates (CODs) by 2–5 years, potentially voiding PPAs, tax credit safe harbors, and investor commitments
• Cost Uncertainty — Interconnection studies are non-refundable ($50K–$200K each), and network upgrade costs assigned during cluster studies can range from $0 to $500+/kW — often unknown until years into the process
• Stranded Investment — Pre-development costs (land options, environmental studies, permitting) are sunk costs if a project withdraws from the queue due to unfavorable study results

1.3 FERC Order 2023 — Shifting to Cluster Studies

In July 2023, FERC issued Order No. 2023, fundamentally reforming the interconnection process. The key changes include:

• Serial to Cluster Studies — Projects are now studied in clusters rather than one-at-a-time, reducing total study time but requiring developers to commit earlier
• First-Ready, First-Served — Replaces the old "first-come" approach with readiness criteria (site control, financial commitments, environmental work)
• Financial Commitments — Escalating deposits throughout the study process to discourage speculative queue entries
• Study Deadlines — Mandatory timelines for ISOs to complete studies, with financial penalties for delays

Chapter 2: ISO Coverage & Data Sources

2.1 Seven ISO/RTO Regions

The Physics AI™ engine ingests interconnection queue data from all seven major ISO/RTO regions in the Continental United States, providing comprehensive coverage for utility-scale solar development:

2.2 Data Ingestion Pipeline

Queue data is refreshed through a two-trigger system:

2.3 Queue Entry Data Fields

After ingestion, each entry is enriched with matched substation coordinates (matched_substation_lat, matched_substation_lon, match_confidence) through the spatial linking process described in Chapter 3.

Chapter 3: Spatial Linking Methodology

Raw interconnection queue data provides Point of Interconnection (POI) names but not coordinates. To enable spatial analysis — calculating which queue entries are near a candidate solar site — the Physics AI™ engine fuzzy-matches POI names to physical substation locations from national infrastructure databases.

3.1 POI-to-Substation Fuzzy Matching

Step 1: Name Normalization

Both POI names from queue data and substation names from infrastructure databases are normalized to improve match quality:

• Convert to lowercase
• Remove suffixes: "substation", "sub", "ss", "switching station"
• Remove voltage references: "230kv", "500kv", "115/230"
• Remove MW references: "150mw", "200 mw"
• Strip extra whitespace and punctuation

Step 2: Proprietary Similarity Matching

The Physics AI™ engine uses proprietary string similarity algorithms to match normalized POI names against substation records. The matching approach analyzes character-level patterns within each name, measuring the structural overlap between the queue entry's POI name and known substation names. This method is robust against:

• Abbreviations — "Vincent Sub" matching to "Vincent Substation 230kV"
• Naming variations — ISOs use different conventions for the same physical substation
• Partial matches — Queue entries often omit voltage or directional identifiers

The matching threshold is calibrated to balance precision (avoiding false matches that would corrupt spatial metrics) with recall (capturing legitimate matches despite naming inconsistencies). Multiple matching strategies are employed across the ingestion and analysis pipelines to maximize coverage.

Step 3: Coordinate Assignment

Matched entries receive the substation's latitude and longitude, along with:

• matched_substation_name — The canonical substation name from the infrastructure database
• matched_substation_lat / matched_substation_lon — Geographic coordinates
• match_confidence — Dice coefficient score (0–1.0)

3.2 Distance-Based Filtering

For a given solar site, the Physics AI™ engine calculates the geodesic distance (true Earth-surface distance accounting for curvature) to every matched substation. Only queue entries at substations within a configurable radius are included in metrics calculation.

The default analysis radius is 50 km, representing a practical maximum gen-tie line distance for utility-scale solar development. This ensures that queue pressure analysis reflects only the substations a developer could realistically interconnect to — not distant POIs that would require prohibitively long and expensive transmission lines.

3.3 ISO Territory Detection

When a site's ISO is not explicitly provided, Earthflow automatically detects the ISO/RTO region based on the site's state using a comprehensive state-to-ISO mapping table. For states that span multiple ISOs (e.g., Kentucky spans both PJM and MISO), the system uses geographic bounding box intersection to select the correct region.

Confidence Scoring

Chapter 4: Queue Pressure Score (QPS)

The Queue Pressure Score (QPS) is Earthflow's Physics AI™ proprietary composite metric that distills the complexity of interconnection queue dynamics into a single 0–100 score. Higher scores indicate greater congestion risk and longer expected interconnection timelines.

4.1 Scoring Methodology

The Physics AI™ QPS is a proprietary weighted composite that evaluates four critical dimensions of interconnection risk. Each dimension is independently scored, normalized to a common scale, and combined using calibrated weights that reflect their relative predictive power for interconnection outcomes.

4.2 Risk Dimensions

Capacity Congestion (Primary Factor)

The most heavily weighted dimension measures the total generation capacity (MW) of active projects competing at the nearest Point of Interconnection. This is the strongest predictor of substation upgrade costs and study delays — a POI with thousands of megawatts ahead in the queue will require significant network upgrades regardless of other factors. The Physics AI™ engine normalizes this against observed congestion ceilings to produce a comparable score across ISOs with vastly different queue sizes.

Queue Viability

This dimension evaluates the ratio of active projects to total projects (including withdrawn and completed entries) at nearby POIs. A high active percentage indicates genuine competition from serious developers. Conversely, a low active percentage — many withdrawals — may signal a problematic POI where previous developers encountered unfavorable study results or prohibitive upgrade costs. Both scenarios carry distinct risk implications that the Physics AI™ scoring captures.

Processing Timeline

The Physics AI™ engine analyzes how long active projects have been waiting in the queue at nearby POIs, normalized against industry benchmark timelines. Extended wait times indicate slow ISO processing, complex study requirements, or systemic congestion that affects all projects at that POI. This temporal dimension captures risk that static capacity numbers alone cannot reveal.

Technology Competition

The final dimension measures solar-specific competition as a proportion of all active projects within the analysis radius. This matters because FERC Order 2023 cluster studies group similar technology types together — solar projects compete directly with other solar projects for the same transmission capacity and study resources. A POI dominated by solar entries presents higher direct competition risk than one with a diversified generation mix.

4.3 Score Interpretation & Timeline Estimates

Chapter 5: Queue Metrics & KPIs

5.1 Complete Metrics Output

For every site analyzed, the Physics AI™ engine returns 16 structured metrics from the QueueMetrics dataclass:

5.2 Interconnection Cost Modeling

Queue metrics feed directly into Earthflow's Physics AI™ development cost models. Rather than relying on static industry averages, the engine estimates interconnection costs by combining spatial, electrical, and queue-specific factors unique to each site:

The Physics AI™ engine combines these three components into a total interconnection cost estimate that reflects the specific conditions at each site — distance to the nearest substation, queue congestion at that POI, transmission voltage capacity, and ISO-specific cost patterns. When detailed site data is limited, the engine applies calibrated fallback estimates based on project capacity and regional cost benchmarks.

5.3 Transmission Distance & Voltage KPIs

Earthflow also calculates complementary transmission infrastructure KPIs that work alongside queue metrics:

Chapter 6: Platform Integration

6.1 Grid Dashboard Heat Map

Physics AI™ Queue Pressure is available as a dedicated heat map layer in Earthflow's Grid Dashboard, alongside soil, slope, erosion, vegetation, and other analysis layers:

Each grid cell popup displays the full queue metrics breakdown when the Queue Pressure layer is selected: QPS score, MW ahead, active projects, solar projects, average wait time, competing generation types, and nearest POI name with distance.

6.2 Cirra AI Agent Tool

Earthflow's Cirra AI agent includes a dedicated get_interconnection_queue tool that enables natural language querying of queue status:

Cirra AI returns Physics AI™-enriched structured responses combining queue metrics with contextual interpretation, cost estimates, and mitigation strategies — all informed by the full suite of Earthflow environmental and infrastructure data for the site.

6.3 Data Infrastructure & Reliability

Queue data is stored in a cloud-native, real-time database organized by ISO region, ensuring fast retrieval and consistent data freshness across all 7 territories. The infrastructure provides:

• Per-ISO Data Isolation — Each ISO region maintains its own dataset with independent refresh cycles, entry counts, capacity totals, and match rate statistics
• Full Audit Trail — Every queue entry retains its original fields (project name, capacity, queue date, status) alongside enriched spatial data (matched substation coordinates, confidence scores)
• Automated Quality Metrics — Match rates, entry counts, and refresh timestamps are tracked per ISO, enabling data quality monitoring and alerting
• Scalable Access — Queue metrics are available through multiple channels: the Grid Dashboard (visual), Cirra AI (conversational), and programmatic APIs (integration)

Chapter 7: Earthflow Differentiation

Interconnection queue data is publicly available. What makes Earthflow unique is not access to the data — it is the Physics AI™ integration, spatial intelligence, and actionable context that transforms raw queue records into development decisions.

7.1 Why Earthflow is Unique

1. Unified Environmental + Grid Intelligence

No other platform combines 70+ environmental data sources (soil, slope, erosion, flood, fire, vegetation, precipitation, irradiance) with live interconnection queue analysis in a single view. Developers typically toggle between multiple tools: one for environmental due diligence, another for grid studies, another for cost modeling. Earthflow delivers a single, integrated assessment where a site's Queue Pressure Score sits alongside its erosion risk, flood zone classification, and vegetation compliance status.

2. Spatial Queue Analysis

Most queue data providers offer tabular downloads — spreadsheets of projects in queue sorted by ISO or date. Earthflow goes further by spatially linking every queue entry to physical substation coordinates via fuzzy POI matching. This enables:

• "Which POI should I target?" — Compare queue pressure across multiple substations within reach of a site
• "What's the competition radius?" — Visualize all competing projects within 50km on the same map as the site boundary
• "Where is the nearest uncongested POI?" — Identify alternative interconnection points that reduce development risk

3. Physics AI™ Cost Modeling

Interconnection costs are the second-largest cost driver for utility-scale solar after panels and racking. The Physics AI™ engine feeds queue pressure directly into development cost estimates, where every cost variable is grounded in physical measurements — not lookup tables or industry averages:

• Gen-tie line distance (~$3M/mile) computed via Haversine geodesic calculations from spatial POI analysis
• Substation upgrade estimates derived from queue congestion levels and transmission voltage capacity
• Study fee budgets from ISO-specific cost data
• Terrain-adjusted gen-tie routing (slope, land cover, watercourse crossings from Earthflow's terrain analysis)
• Transmission voltage capacity from infrastructure KPIs

This produces a complete interconnection cost estimate at the site screening stage — before a developer spends months and hundreds of thousands of dollars on formal interconnection studies.

The QPS itself is a physics-informed metric: it uses real queue positions (ordinal), MW capacities (power), temporal decay (wait time in months), and technology-specific competition ratios — all physical or measurable quantities rather than subjective ratings or qualitative assessments.

4. AI-Powered Interpretation

Earthflow's Cirra AI agent doesn't just retrieve queue metrics — it interprets them in context. When a developer asks "What's the interconnection risk for this site?", Cirra considers the QPS alongside:

• The site's environmental constraints (does it need a longer gen-tie to avoid sensitive habitats?)
• Access road requirements (does terrain affect construction logistics?)
• Competing land use (are there other solar projects nearby that indicate proven viability?)
• Financial implications (does the interconnection cost break the project's IRR target?)

5. Grid-Cell Resolution Analysis

Queue pressure is overlaid on the same grid-cell analysis framework used for soil bearing capacity, slope stability, erosion control, and vegetation management. This means developers can see interconnection risk in the same spatial context as construction suitability — on the same map, at the same resolution, in the same analysis session.

6. Automated Monitoring

Monthly automated refresh ensures queue data is always current — not a stale snapshot from a one-time download. As projects enter, advance through, or withdraw from the queue, Earthflow's metrics update accordingly. Developers with sites in active development can monitor their competitive position over time.

7. FERC Order 2023 Awareness

Earthflow's timeline estimates and cost projections reflect the post-FERC Order 2023 reality of cluster studies, escalating deposits, and readiness requirements. The QPS interpretation table accounts for the transition from serial to cluster processing, providing realistic timeline estimates for today's regulatory environment.

7.2 Value Across the Development Lifecycle

Earthflow's queue intelligence isn't a one-time report — it delivers value at every stage of the solar development lifecycle:

7.3 Competitive Value Matrix

The following matrix maps Earthflow's Physics AI™ capabilities against what is available through alternative approaches. Green indicates a capability Earthflow delivers; amber indicates partial or manual alternatives exist; red indicates no viable alternative.

* Free queue downloads require significant manual effort to process, spatially link, and interpret — the hidden cost is analyst time.

Chapter 8: Roadmap & Future Capabilities

The Physics AI™ queue intelligence engine is designed as an extensible framework. The following capabilities build directly on the existing QPS methodology, spatial linking infrastructure, and 7-ISO data pipeline.

8.1 Near-Term Enhancements

These extensions leverage existing infrastructure and can be deployed incrementally:

Historical QPS Trend Tracking

Store monthly QPS snapshots for every matched POI, enabling time-series analysis of queue pressure changes. Developers will see whether congestion is worsening or improving at their target substations — e.g., "QPS at Vincent Sub has increased 15 points over the last 6 months, driven by 800 MW of new solar entries."

Queue Position Change Alerts

Real-time notifications when queue events affect a developer's target POIs: new project entries, withdrawals, status changes (e.g., "moved to Facilities Study"), and study completions. Delivered via Earthflow dashboard, email, or Slack integration.

Multi-POI Comparison Tool

Compare 3–5 nearest substations simultaneously with side-by-side QPS, MW ahead, wait time, gen-tie distance, and cost estimates. Presented as an interactive comparison table in the Grid Dashboard, enabling rapid "which POI should I target?" decisions.

ISO-Specific Withdrawal Rate Analysis

Track and model withdrawal rates by ISO, year, and technology type to refine QPS timeline estimates. For example, ERCOT's ~80% withdrawal rate produces different timeline implications than PJM's ~65% rate. Physics AI™ will use these ISO-specific patterns to calibrate QPS interpretation.

8.2 Advanced Analytics

Machine learning and predictive capabilities that extend the Physics AI™ foundation:

Predictive Queue Modeling

ML-based interconnection study completion ETA using historical study durations, cluster sizes, ISO processing capacity, and project characteristics as features. Rather than relying on broad timeline ranges (e.g., "3–5 years"), Physics AI™ will predict site-specific expected completion dates with confidence intervals.

Curtailment Risk Analysis

Post-interconnection generation curtailment probability based on transmission congestion patterns, renewable penetration levels, and nodal pricing data. A site may successfully interconnect but face significant curtailment if the local transmission network is already saturated with renewable generation.

Storage Co-Location Optimization

Analyze queue position advantages of solar+storage configurations versus standalone solar. Battery co-location often receives preferential treatment in cluster studies and can reduce assigned network upgrade costs. Physics AI™ will quantify the cost-benefit of adding storage to improve queue outcomes.

Real-Time Queue Event Feeds

Upgrade from monthly batch refresh to streaming updates from ISO queue portals — new entries, withdrawals, and status changes reflected within hours rather than weeks. Critical for developers in fast-moving markets like ERCOT and CAISO.

8.3 Data Source Expansion

New external data integrations that deepen the Physics AI™ grid intelligence capability:

Available Transfer Capability (ATC) Monitoring

Integrate OASIS ATC data to show available transmission capacity at each POI, adding a supply-side dimension to the demand-side QPS. A POI with high queue pressure but ample ATC may still offer viable interconnection paths.

Hosting Capacity Map Integration

Ingest utility-published hosting capacity data (where available) to identify distribution-level interconnection opportunities that bypass the transmission queue entirely. Particularly relevant for smaller solar projects (<20 MW) that can interconnect at distribution voltage levels.

Utility IRP Integration

Cross-reference ISO queue data with utility Integrated Resource Plans to identify POIs where planned transmission upgrades will reduce future congestion. A POI with high current QPS but a planned 500 kV transmission expansion in the utility's IRP represents a different risk profile than one with no planned upgrades.