Forge Bastion IWMS · Energy & Carbon Intelligence

Engine Technical
Design Document

Architecture, pipeline design, model specification, and performance validation across eight AI engines for real-time energy metering, HVAC optimization, carbon emissions accounting, regulatory compliance automation, and decarbonization pathway intelligence. Built in Rust. Every kilowatt measured. Every metric ton accounted.

The fines are not coming. They’re here. $268 per metric ton — and 63% of NYC buildings will exceed their limits by 2030.

8
Intelligence Engines
$268
Per tCO²e Over Limit
40+
US Cities with BPS
18–28%
Energy Cost Reduction
engine_index
Eight engines. One carbon nervous system.
01
Energy Metering
Sub-meter resolution at 15-second intervals
02
HVAC Optimization
Autonomous AI control with 99.6% prediction
03
Carbon Accounting
Hourly Scope 1–3 tracking and penalty forecasting
04
Regulatory Compliance
LL97, BERDO, BEPS — 40+ frameworks auto-generated
05
Decarb Pathways
Net-zero roadmaps with NPV-positive sequencing
06
Demand Response
Peak shaving, battery dispatch, grid intelligence
07
Tenant Engagement
Green lease compliance and behavior-based programs
08
Portfolio Benchmarking
Normalized EUI and acquisition due diligence
executive_summary
An eight-engine architecture for the carbon reckoning that has already begun

Buildings consume 36% of global energy and contribute nearly 40% of CO² emissions worldwide. In New York City, buildings account for nearly 70% of total carbon emissions. For decades, that energy consumption existed as an unmanaged line item — a utility bill paid, filed, and forgotten. That era ended when cities began legislating carbon caps on individual buildings with financial penalties severe enough to reshape the economics of commercial real estate.

NYC’s Local Law 97 — part of the Climate Mobilization Act — began requiring annual emissions reports in May 2025 and assessing penalties of $268 per metric ton of CO²e over the building’s annual limit. While 89% of buildings complied in the first period, limits tighten dramatically in 2030, when 63% of covered buildings are projected to exceed their caps. Boston’s BERDO 2.0 levies $1,000/day fines. Washington DC’s first BEPS compliance cycle concluded in 2026 with maximum exposure reaching $1 million per property. Penalties across major BPS cities will increase an average of 82% between first and second compliance periods. Over 40 US cities will have building performance standards by 2026 — and the number is growing.

Bastion Meridian does not simply report emissions — it actively reduces them through AI-driven HVAC optimization that predicts building thermal state with 99.6% accuracy, then autonomously adjusts individual equipment every 5 minutes. The platform generates audit-ready compliance documentation across all 40+ BPS frameworks from a single data infrastructure, forecasts penalty exposure under multiple scenarios, and models decarbonization pathways that identify the 73% of emissions eliminable at positive net present value. The result: 18–28% energy cost reduction at deployed buildings, with the avoided penalties in Year 1 alone paying for the platform for the next decade.

$268
Per tCO²e Over Limit (NYC LL97)
63%
NYC Buildings Exceeding 2030 Caps
99.6%
AI Thermal State Prediction
40+
US Cities with BPS Regulations
18–28%
Energy Cost Reduction
82%
Penalty Increase Between Periods
ENG 01
Real-Time Energy Metering
Sub-meter granularity at 15-second intervals — attributing energy consumption to individual systems (HVAC, lighting, plug loads, process) because you cannot optimize what you cannot measure, and utility bills measure nothing useful.
15sec
Resolution
Architecture
IoT Sub-Metering + Edge Analytics
CT clamp current sensors on electrical panels; BMS integration via BACnet/Modbus; edge computing for 15-second interval aggregation; system-level attribution via circuit mapping
Performance
System-Level Attribution
Energy consumption broken down by HVAC (heating, cooling, ventilation separately), lighting, plug loads, domestic hot water, and process loads — at 15-second resolution
Features
Anomaly Detection
Identifies systems consuming energy outside expected parameters: AHUs running during unoccupied hours, lighting circuits at full power overnight, equipment cycling abnormally
Impact
5–15 Issues Found Day 1
Immediate visibility typically surfaces 5–15 existing issues: units running 24/7 that should cycle, equipment drawing abnormal power, systems operating during unoccupied periods

Utility bills arrive monthly, report building-level totals, and tell you nothing about where energy is actually consumed. Engine 01 deploys sub-metering at the electrical panel level with CT clamp sensors and integrates with existing BMS data via BACnet and Modbus, creating 15-second-resolution energy profiles attributed to individual building systems. The immediate value is visibility: at deployment, the system typically identifies 5–15 operational anomalies that have been wasting energy invisibly — AHUs running at full capacity through summer break with zero occupants, lighting circuits energized overnight in unoccupied floors, and equipment cycling at frequencies that indicate control system misconfiguration. One university discovered that 30% of its campus energy was consumed by three buildings with 1960s systems running 24/7 — nobody knew because nobody was measuring at the system level.

ENG 02
AI HVAC Optimization
Autonomous control that predicts building thermal state with 99.6% accuracy and writes back to individual HVAC equipment every 5 minutes — occupancy-responsive, weather-predictive, and grid-aware.
99.6%
Prediction
25%
HVAC Savings
Architecture
Deep RL + BMS Write-Back
Deep reinforcement learning trained on building-specific thermal dynamics; connects to BMS via BACnet/IP; autonomously writes setpoint adjustments to individual AHUs, VAVs, and chillers every 5 minutes
Performance
99.6% Thermal State Prediction
Predicts future building thermal state using weather forecasts, occupancy patterns, thermal mass modeling, and equipment performance curves; pre-cools before peak demand, pre-heats before occupancy
Features
Occupancy-Responsive Control
Integrates with Bastion Horizon occupancy data to condition only spaces with actual occupants; HVAC scales with real-time demand rather than fixed schedules
Toolchain
Python / PyTorch / BACnet
Deep RL with actor-critic architecture; BACnet/IP for BMS integration; Haystack-standard point naming; virtual tests select optimal algorithms per building’s unique thermal behavior

HVAC systems consume 40–60% of commercial building energy, yet most run on fixed schedules and static setpoints that ignore actual conditions. Engine 02 connects to the building management system via BACnet/IP, learns the building’s unique thermal behavior through an initial learning phase, and then autonomously adjusts individual equipment setpoints every 5 minutes based on current and forecasted weather, utility tariff structures, grid emission factors, and occupant density. The AI predicts the future thermal state of the building with 99.6% accuracy, enabling pre-cooling before afternoon peak demand (when electricity is most expensive and carbon-intensive) and pre-heating before morning occupancy (using cheaper overnight electricity). Equipment runtime reduction extends HVAC asset life by up to 50% and defers retrofit capital expenditure. At deployed buildings, HVAC energy costs typically decrease 20–25% with no comfort degradation — and often with improved thermal comfort scores.

ENG 03
Carbon Emissions Accounting
Hourly Scope 1, 2, and 3 emissions tracking with penalty exposure forecasting — because annual emissions reports that arrive in May tell you what you owed in January, not what you can do about it.
Scope 1–3
Hourly Tracking
Architecture
Metering + Grid Emission Factors
On-site combustion (Scope 1) from fuel metering; purchased electricity (Scope 2) using hourly grid emission factors; tenant operations and embodied carbon (Scope 3) via activity data and emission factor libraries
Performance
Real-Time Penalty Exposure
Continuous year-to-date emissions vs. annual cap comparison; projected year-end penalty exposure updated hourly; scenario modeling for intervention impact on final penalty
Features
Rolling Forecast
Projects year-end emissions trajectory based on current consumption patterns and seasonal modeling; identifies the month when the building will cross its annual cap at current rates
Impact
$1.8M Penalties Avoided (Y1)
Manhattan REIT avoided $1.8M in LL97 penalties while simultaneously reducing energy costs $3.4M annually — net positive return from compliance
ENG 04
Regulatory Compliance Automation
Auto-generates compliance filings across 40+ building performance standard frameworks from a single data infrastructure — LL97, BERDO, BEPS, Energize Denver, and every new city that legislates carbon caps.
40+
Frameworks
Architecture
Template Engine + Regulatory DB
Jurisdiction-specific filing templates auto-populated from building energy data; regulatory database tracking compliance deadlines, limit changes, and penalty structures across all 40+ BPS cities
Performance
One Data Source → All Filings
A national REIT with buildings in NYC, Boston, DC, and Denver generates all four compliance filings from a single Meridian data infrastructure — eliminating four separate compliance workflows
Features
Deadline & Limit Tracking
Automated alerts for filing deadlines; proactive notification when limit changes are proposed or enacted; penalty impact modeling for proposed regulatory changes
Impact
GRESB 62 → 84
National mixed-use REIT improved GRESB score from 62 to 84 after unifying compliance data across four BPS-regulated cities onto a single platform — investors noticed
ENG 05
Decarbonization Pathway Modeling
Net-zero roadmaps that sequence interventions by NPV — because 73% of building emissions are eliminable through measures with positive net present value, and the order you implement them determines whether decarbonization creates or destroys value.
73%
NPV-Positive
Architecture
Optimization + Monte Carlo
Intervention sequencing via integer programming; Monte Carlo simulation for uncertainty (energy prices, carbon prices, technology costs); sensitivity analysis on electrification pathways
Performance
NPV-Optimized Sequencing
Identifies which interventions to implement first (LED retrofit before chiller replacement before envelope upgrade) to maximize cumulative NPV across the decarbonization journey
Features
Technology-Agnostic Modeling
Heat pumps, VRF, geothermal, solar, battery storage, building envelope, LED, BMS upgrades — every option modeled against the building’s specific characteristics and utility tariff structure
Impact
Value Creation, Not Cost
Transforms decarbonization from a regulatory cost into a value-creating capital program; most buildings achieve compliance through interventions that pay for themselves within 5–8 years
ENG 06
Demand Response & Grid Intelligence
Peak shaving, battery dispatch optimization, and grid signal response — reducing demand charges (often 30–50% of the utility bill) while earning revenue from grid flexibility programs.
30–50%
Of Bill = Demand
Architecture
Load Prediction + Battery MPC
15-minute load forecasting for demand charge avoidance; model predictive control (MPC) for battery storage dispatch; automated response to utility demand response signals and real-time pricing
Performance
Peak Demand Reduction 15–25%
Pre-cooling before peak periods; battery discharge during demand peaks; load shifting of deferrable equipment to off-peak hours; time-of-use tariff optimization
Features
Grid Carbon Matching
Shifts consumption to hours when the grid is cleanest (high renewable generation); reports actual hourly carbon intensity rather than annual grid averages — critical for Scope 2 accuracy
Impact
Revenue from Flexibility
Buildings with battery storage and flexible loads earn demand response revenue; one commercial tower generates $180K annually from grid flexibility programs
ENG 07
Tenant Energy Engagement
Green lease compliance monitoring, tenant-level energy attribution, and behavior-based reduction programs — because building owners control the base building, but tenants control 40–60% of the energy consumption.
40–60%
Tenant Energy
Architecture
Tenant Sub-Metering + Portal
Tenant-level energy attribution via sub-metering or allocation algorithms; tenant energy dashboard with benchmarking against peers; green lease clause compliance tracking
Performance
8–15% Tenant Energy Reduction
Visibility and benchmarking alone drive 8–15% tenant energy reduction; behavior-based programs with gamification and incentives achieve additional 5–10%
Features
Green Lease Enforcement
Monitors tenant compliance with green lease energy clauses; automated reporting to tenants on their energy performance; supports cost-recovery for landlord-funded improvements
Impact
Shared Responsibility
Under LL97, the building owner pays the penalty regardless of tenant behavior; Meridian creates the data infrastructure to share responsibility through green lease mechanisms
ENG 08
Portfolio Energy Benchmarking
Normalized energy use intensity (EUI) comparison across the portfolio, ENERGY STAR scoring, GRESB integration, and acquisition due diligence — identifying the buildings where investment in efficiency has the highest return.
EUI
Normalized
Architecture
Normalized KPI + Peer Comparison
Weather-normalized, occupancy-adjusted EUI per building; peer benchmarking by building type, vintage, climate zone; ENERGY STAR Portfolio Manager integration; GRESB reporting automation
Performance
Portfolio-Wide Visibility
Identifies bottom-quartile buildings where efficiency investment has the highest ROI; surfaces buildings approaching compliance thresholds before they cross
Features
Acquisition Due Diligence
Carbon risk scoring for acquisition targets: current emissions vs. applicable caps, projected penalty exposure, estimated decarbonization cost, and impact on portfolio-level compliance
Impact
Carbon as Financial Risk
Transforms carbon from an ESG reporting metric into a financial risk variable integrated into acquisition underwriting, asset management, and disposition decisions
carbon_impact
$268
Per tCO²e penalty (NYC LL97)
63%
NYC buildings exceeding 2030 caps
$5.2M
Penalties avoided Year 1 (case study)
62→84
GRESB score improvement (national REIT)