Applied AI / 2026
AI Energy Forecasting Dashboard
From demand forecasts to explainable planning.
01 / THE PROBLEM
Forecasts you can interrogate.
Operations teams need to inspect forecast drivers and uncertainty as well as expected energy demand. This prototype brings those views into one decision-support interface.
My contribution
Hackathon project contributorI contributed to the energy-demand forecasting project for the PowerMind Hackathon. The team’s solution connects a forecasting backend with a dashboard for exploring demand, uncertainty and what-if scenarios.
02 / TECHNICAL APPROACH
Inside the
implementation.
- 01
The file-based energy dataset is cleaned by timestamp sorting, deduplication, numeric conversion and interpolation. Macro forecasts use daily MWh totals.
- 02
Hourly and daily Prophet models capture trend, seasonal and holiday components, making the forecast structure and intervals easier to interpret.
- 03
A FastAPI service exposes forecasts, seasonality, residuals and anomalies. A Next.js interface supports date selection, what-if scenarios and structured range explanations.
03 / SYSTEM ARCHITECTURE
Follow the signal.
Select a stage to explore
how the system works.
Consumption data
Timestamped electricity consumption in MWh is loaded, cleaned and aggregated from the project dataset.
05 / RESULTS & OUTCOMES
What the work delivers.
The project earned second runner-up at the PowerMind Hackathon. It brings forecasts, seasonal patterns, residuals and anomalies into one interface for exploring energy demand.
Scope & limitations
- Hackathon prototype using a file-based dataset; no connected live telemetry or production authentication.
- Running forecasts requires the project’s dataset and backend; this case study provides an architecture walkthrough.
06 / ENGINEERING DECISIONS
The interesting
constraints.
Future lags are not available
A forecast cannot rely on future lag values that are unavailable at prediction time. The active macro path uses calendar and seasonal structure instead.
A useful forecasting product connects the prediction to its drivers, uncertainty and operational context.
07 / BUILDING BLOCKS
Tools with
a purpose.
- Prophet / Pandas
- Interpretable time-series forecasts
- FastAPI
- Forecast and analysis endpoints
- Next.js / TypeScript
- Dashboard and interaction
Go to the source
NEXT PROJECT / 05