Farzerl/AI-Driven-Energy-Management-System-for-Institutional-Network-SIMBA-EMS-

# SIMBA-EMS **AI4I 2026 Track 3 — Development | Source and technical-evidence repository** SIMBA-EMS is a locally hosted, AI-assisted institutional energy-management MVP for Estates and facilities teams. It receives electrical meter readings, forecasts demand and power-quality conditions, applies deterministic engineering safety rules, ranks controllable responses, requires operator approval, and records the resulting decision and impact evidence. ## Current product status - Working browser dashboard and FastAPI backend. - Multi-horizon demand forecasting at 30 minutes, 2 hours, 6 hours and 24 hours. - HGB, LSTM, Transformer and Chronos-2 model routing with chronological validation. - Separate active-power and reactive-power forecasting for power-factor and tariff-risk analysis. - Ranked multi-facility action queue. - Dashboard-only approval, rejection, acknowledgement and audit trail. - Gmail notification support with local credentials and dry-run mode. - Server-side software-in-the-loop replay. - External or physical control disabled by default. - Planning and simulated impact are not presented as realised electricity-bill savings. ## Core workflow ```text smart meter or safe simulation → edge validation and buffering → FastAPI ingestion → operational storage → multi-model forecasting → uncertainty and risk assessment → deterministic safety rules → operator approval → simulation or authorised gateway → verification and audit evidence ``` See `docs/architecture.md` and the architecture diagrams under `docs/diagrams`. ## Repository scope This public-review repository contains: - application and API source code; - dashboard source and assets; - Windows launcher source and build script; - dependency manifests and lock files; - anonymised sample and simulation data; - database/data-contract documentation; - AI method, model-card and validation summaries; - automated tests and security/repository-audit evidence; …