Inspiration

Rapid urban growth and harsh arid climates make the UAE one of the highest per-capita consumers of water and electricity in the world, with air conditioning (cooling loads) alone accounting for up to 70% of a building's electricity use. Inspired by the UAE’s Net-Zero 2050 charter, UAE Centennial 2071, and DEWA’s Demand-Side Management strategies, we wanted to build a localized solution to this challenge.

We realized that while smart meters exist, communities lack the predictive, proactive tools needed to act before peak demand spikes happen. We wanted to build an intelligent, accessible ecosystem that transforms campus and residential utility data into automated, climate-smart action.

What it does

AI-Driven Smart City & Resource Optimization is an intelligent, real-time platform designed for university campuses and residential communities in the UAE to predict, visualize, and optimize energy and water consumption.

Predictive Forecasting: Uses machine learning to anticipate cooling loads and water demand 6 hours in advance, factoring in UAE-specific weather (temperature & humidity) and occupancy patterns.

Smart Peak Alerts: Detects when consumption is on track to hit costly peak surge hours and automatically flags them.

Automated Conservation Insights: Offers actionable recommendations, such as suggesting a (+1^\circ\text{C}) thermostat adjustment to drop peak AC load, or flagging abnormal water flow rates that indicate potential leaks.

Granular IoT Dashboard: Visualizes real-time and predicted data across simulated campus nodes (dorms, labs, and common areas) to keep utility managers informed.

How we built it

We built this prototype from the ground up over the weekend using a fast, modern tech stack:

The Brains (Python Backend): We used Python to develop the ML engine. We generated realistic time-series environmental data and utilized an autoregressive forecasting model (using scikit-learn and pandas) to predict cooling loads (L_t) based on temperature (T_t) and historical trends. We exposed these predictive endpoints using a high-performance FastAPI server.

The Simulator: Built a custom Python script simulating real-time IoT smart meter telemetry (power draw in kW, flow rate in L/min, ambient temperature) across multiple university zones.

The Dashboard (Frontend): Crafted a responsive, lightweight React/Next.js application. We implemented dynamic charting using Chart.js to visualize live vs. predicted consumption, integrated alert banners, and customized optimization recommendations based on DEWA’s demand response models.

Challenges we ran into

Calibrating the IoT Simulator: Generating fake data is easy, but generating realistic UAE resource data—factoring in extreme afternoon heat spikes, weekend vs. weekday university occupancy, and indoor cooling load correlations—was incredibly complex. We spent hours tweaking mathematical variance to ensure our AI had realistic patterns to learn from.

Model Inference at Speed: Setting up a pipeline where the AI can process continuous streaming telemetry data, retrain/predict on the fly, and send that over WebSocket/HTTP APIs without lag was difficult. We had to optimize our pandas processing to prevent the FastAPI server from bottlenecking.

Complex UI State: Synchronizing multiple live-updating charts, progress gauges, and recommendation cards simultaneously in React proved to be a state management headache, requiring we implement throttled polling to keep the dashboard smooth.

Accomplishments that we're proud of

Working End-to-End Pipeline: We succeeded in getting live simulated telemetry data to feed into our ML backend, generate accurate future predictions, and reflect on a beautiful UI in under 1 second.

High Predictive Accuracy: Despite using simulated datasets, our lightweight regression-based forecasting successfully caught and predicted "peak cooling" thresholds with high fidelity.

High-Impact UI/UX: Built a truly intuitive dashboard that doesn't just show numbers but translates them into concrete actions (like predicting an estimated (12%) utility cost reduction for the day).

Localized UAE Fit: Designed the application specifically for the UAE's high cooling load demands, making the project directly viable for local campuses or municipal pilots.

What we learned

The "Impact Over Features" Rule: We originally wanted to include carbon credit tracking and waste management, but we quickly learned to cut features to ensure the AI forecasting and dashboard flow were 100% functional for the demo.

Climatic Context Matters: Developing a utility optimizer for the UAE requires heavily weighting cooling metrics—ordinary "smart home" models don't translate directly to extreme desert environments.

Collaboration Under Pressure: We learned how to effectively divide backend data processing from frontend UI design, using strict API contract definitions so both teams could work in parallel.

What's next for AI-Driven Smart City & Resource Optimization

Hardware Integration: We want to take this out of the simulator and test it with physical IoT edge devices, such as ESP32 controllers and split AC adapters.

Building Automation (Active Control): Shift from "recommending" AC and water changes to safely automating them via integrations with Building Management Systems (BMS).

Edge ML Deployment: Optimize the forecasting models to run directly on campus edge servers to save bandwidth and improve privacy.

Government Integration: Align with the "Zero Government Bureaucracy" initiative by connecting optimization alerts with the digital systems of UAE utilities (like DEWA and ADDC) for grid-level demand-response programs.

Built With

Share this project:

Updates

Submission history