Skip to content

GloBI Architecture - Simple Overview#

This diagram provides a high-level overview of the GloBI (Global Building Intelligence) system workflow.

%%{init: {'theme':'base', 'themeVariables': { 'primaryColor':'#4a9eff','primaryTextColor':'#000','primaryBorderColor':'#2563eb','lineColor':'#64748b','secondaryColor':'#fbbf24','tertiaryColor':'#34d399','noteBkgColor':'#fef3c7','noteTextColor':'#000','noteBorderColor':'#f59e0b'}}}%%
flowchart TD
    %% Input Stage
    A[User Manifest File<br/>YAML Configuration] --> B{CLI Command<br/>submit manifest}

    %% Input Files
    C[GIS Building Data<br/>Shapefile/GeoJSON] --> D
    E[Component Database<br/>SQLite/Prisma] --> D
    F[Semantic Fields<br/>YAML Mappings] --> D
    G[Weather Data<br/>EPW Files] --> D

    %% Preprocessing Stage
    B --> D[1. GIS Preprocessing<br/>Validate & Enrich Data]

    %% Spec Generation
    D --> H[2. Building Spec Generation<br/>Create GloBIBuildingSpec]

    %% Allocation Stage
    H --> I[3. Job Allocation<br/>Calculate Branching<br/>& Submit to Hatchet]

    %% Distributed Execution
    I --> J[4. Distributed Simulation<br/>Workers Run EnergyPlus]

    %% Results Collection
    J --> K[5. Results Aggregation<br/>Store in S3/Cloud]

    %% Output Stage
    K --> L{CLI Command<br/>get experiment}
    L --> M[Local Results<br/>Parquet/CSV Files]
    L --> N[Visualization Dashboard<br/>Interactive HTML]

    %% Styling
    style A fill:#60a5fa,stroke:#2563eb,stroke-width:2px,color:#000
    style C fill:#60a5fa,stroke:#2563eb,stroke-width:2px,color:#000
    style E fill:#60a5fa,stroke:#2563eb,stroke-width:2px,color:#000
    style F fill:#60a5fa,stroke:#2563eb,stroke-width:2px,color:#000
    style G fill:#60a5fa,stroke:#2563eb,stroke-width:2px,color:#000
    style D fill:#fcd34d,stroke:#f59e0b,stroke-width:2px,color:#000
    style H fill:#fcd34d,stroke:#f59e0b,stroke-width:2px,color:#000
    style I fill:#fca5a5,stroke:#dc2626,stroke-width:2px,color:#000
    style J fill:#fca5a5,stroke:#dc2626,stroke-width:2px,color:#000
    style K fill:#4ade80,stroke:#16a34a,stroke-width:2px,color:#000
    style M fill:#4ade80,stroke:#16a34a,stroke-width:2px,color:#000
    style N fill:#4ade80,stroke:#16a34a,stroke-width:2px,color:#000

Workflow Stages#

1. Input Configuration#

  • Manifest File: YAML configuration defining the experiment, file paths, and preprocessing parameters
  • GIS Data: Building footprints with properties (height, floors, typology)
  • Component Database: Building components and materials specifications
  • Semantic Fields: Mappings between building categories and properties
  • Weather Data: EPW climate files for simulation

2. GIS Preprocessing#

Validates and enriches building data:

  • Filters buildings by area, height, and geometry validity
  • Converts polygons to rotated rectangles
  • Identifies neighboring buildings for shading analysis
  • Assigns weather files based on location
  • Injects semantic context (building typology, age, region)

3. Building Spec Generation#

Creates structured specifications for each building:

  • Extracts geometry (footprint dimensions, height, neighbors)
  • Assigns building properties (WWR, basement, attic)
  • Links semantic context and weather data
  • Produces GloBIBuildingSpec objects ready for simulation

4. Job Allocation#

Distributes work across compute infrastructure:

  • Calculates optimal branching factor based on payload size
  • Submits jobs to Hatchet workflow orchestrator
  • Distributes building specs across Docker worker containers

5. Distributed Simulation#

Workers execute energy simulations in parallel:

  • Each worker processes assigned building specs
  • Uses component database to construct EnergyPlus models
  • Runs building energy simulations
  • Extracts monthly energy and peak results
  • Optionally captures hourly timeseries data

6. Results Aggregation & Collection#

Consolidates simulation outputs:

  • Aggregates results from all workers
  • Stores in cloud storage (S3) as Parquet files
  • Versions experiments for reproducibility
  • Makes results available for retrieval

7. Output & Visualization#

Delivers results to users:

  • Downloads results to local filesystem
  • Generates interactive D3-based dashboards
  • Provides both Parquet and CSV formats
  • Enables analysis of building stock energy performance

Key Features#

  • Scalability: Processes thousands of buildings in parallel using distributed computing
  • Automation: Minimal manual intervention from GIS data to simulation results
  • Reproducibility: Version-controlled experiments with semantic versioning
  • Flexibility: Configurable preprocessing, semantic mappings, and output options
  • Regional Analysis: Designed for urban-scale building energy modeling