Engineering Team
This project is developed by a group of Software Engineering students from Sir Syed University of Engineering and Technology:
Mudasir Ali
unarmudasir@gmail.com
Abdullah Tahir
axise140@gmail.com
Abubakr Saqib
abubakarsaqib60@gmail.com
Haider Rizwan
haidernv006@gmail.comStrategic Sustainability Goals
Our engineering objectives are mapped to the United Nations SDGs, ensuring that AI innovation directly addresses global energy and consumption challenges.
Affordable & Clean Energy
Why: Unpredictable electricity costs create financial instability for residential
consumers.
How: The Hybrid Physics Engine provides preemptive awareness by forecasting unit
consumption, helping users maintain their energy budget before the cycle ends.
Responsible Consumption
Why: Inefficient load patterns increase national peak demand and environmental
strain.
How: Our BiLSTM Forecaster identifies temporal dependencies in usage, shifting
behavioral patterns toward higher efficiency and reducing waste through intelligent scheduling.
Research Foundation (PRECON)
Our optimization logic is built upon the Pakistan Residential Electricity Consumption (PRECON) dataset, published in 2019 at the ACM International Conference on Future Energy Systems (e-Energy '19). This research was pioneered by the Energy Informatics Group (EIG) at the Lahore University of Management Sciences (LUMS).
PRECON is a first-of-its-kind extensive study designed to enable the development of intelligent smart grids and better demand-side management tools in the residential world. Unlike generic datasets, it provides a highly granular view of Pakistani energy dynamics:
- Smart Meter Telemetry: Data was collected using high-frequency smart meters over a continuous one-year period to capture precise seasonal fluctuations in Pakistan.
- Diverse Demographics: The study monitors 42 individual households belonging to various social and financial backgrounds, ensuring our models are calibrated for all economic classes.
- Multi-Level Metadata: Beyond total units, the dataset provides detailed load profiles for high-consumption devices, allowing our AI to recognize specific "Appliance Signatures."
Neural Architecture Stack
Our system utilizes a Tri-Model Hybrid Pipeline to synthesize energy predictions. By combining statistical machine learning, deep neural networks, and first-principles physics, we eliminate the "black box" problem of standard AI.
Random Forest
A sophisticated regressor calibrated on 13 zero-leakage features. It handles long-term forecasting by analyzing appliance counts and seasonal drift.
Bidirectional LSTM
A specialized RNN that recognizes temporal dependencies. It processes a 48-hour lookback window to forecast a 24-hour demand profile.
KNN Matcher
Maps your profile to the closest match among 42 PRECON house archetypes. This ensures realistic seeding even when historical data is absent.
Mathematical Handshake
Computes a structural baseload using occupancy and area factors. Predictions are a synchronized blend of Physics, AI Inference, and Historical Calibration.
Load Signature Mapping
The engine maps your household inventory against the sub-metered appliance signatures established in the e-Energy '19 (PRECON) dataset, augmented by our physics-based baseline constraints:
These specific parameters allow our Random Forest model to recognize consumption spikes and differentiate between essential baseload and high-wattage appliance usage.
Technical Reference
Nadeem, Ahmad and Arshad, Naveed. "PRECON: Pakistan Residential Electricity Consumption Dataset." Proceedings of the Tenth ACM International Conference on Future Energy Systems (e-Energy '19). Phoenix, AZ, USA. pp. 52-57. 2019. doi: 10.1145/3307772.3328317