24h-energy-consumption-prediction Mfaj-cod

Project_1_24h-energy-consumption-prediction/.gitignore

@@ -0,0 +1,3 @@
+# Ignore large data directories
+data/*.csv
+datasets/

Project_1_24h-energy-consumption-prediction/artifacts/fast_track/forecast_T_plus_24.csv

@@ -0,0 +1,25 @@
+x_timestamp,yhat
+2020-12-31 00:00:00+05:30,0.06285337389677273
+2020-12-31 01:00:00+05:30,0.0779414067052093
+2020-12-31 02:00:00+05:30,0.06222471008699352
+2020-12-31 03:00:00+05:30,0.05219554075912427
+2020-12-31 04:00:00+05:30,0.05723717969284075
+2020-12-31 05:00:00+05:30,0.058565367357661206
+2020-12-31 06:00:00+05:30,0.10995892204143862
+2020-12-31 07:00:00+05:30,0.18344699974685214
+2020-12-31 08:00:00+05:30,0.15427799736384254
+2020-12-31 09:00:00+05:30,0.2155807264085385
+2020-12-31 10:00:00+05:30,0.20905629912941043
+2020-12-31 11:00:00+05:30,0.15220458542063398
+2020-12-31 12:00:00+05:30,0.0960510283377849
+2020-12-31 13:00:00+05:30,0.12343713819079415
+2020-12-31 14:00:00+05:30,0.11372958748363775
+2020-12-31 15:00:00+05:30,0.12292408973174732
+2020-12-31 16:00:00+05:30,0.0873224892064543
+2020-12-31 17:00:00+05:30,0.10990953989302502
+2020-12-31 18:00:00+05:30,0.12594811958678342
+2020-12-31 19:00:00+05:30,0.13203974778234912
+2020-12-31 20:00:00+05:30,0.10784505058529184
+2020-12-31 21:00:00+05:30,0.12125726524513988
+2020-12-31 22:00:00+05:30,0.08112030239674972
+2020-12-31 23:00:00+05:30,0.07778420342924022

Project_1_24h-energy-consumption-prediction/artifacts/fast_track/forecast_baseline_T_plus_24.csv

@@ -0,0 +1,25 @@
+x_timestamp,yhat
+2020-12-31 00:00:00+05:30,0.034
+2020-12-31 01:00:00+05:30,0.063
+2020-12-31 02:00:00+05:30,0.08
+2020-12-31 03:00:00+05:30,0.067
+2020-12-31 04:00:00+05:30,0.059000000000000004
+2020-12-31 05:00:00+05:30,0.066
+2020-12-31 06:00:00+05:30,0.27899999999999997
+2020-12-31 07:00:00+05:30,0.263
+2020-12-31 08:00:00+05:30,0.101
+2020-12-31 09:00:00+05:30,0.244
+2020-12-31 10:00:00+05:30,0.099
+2020-12-31 11:00:00+05:30,0.242
+2020-12-31 12:00:00+05:30,0.083
+2020-12-31 13:00:00+05:30,0.076
+2020-12-31 14:00:00+05:30,0.122
+2020-12-31 15:00:00+05:30,0.057999999999999996
+2020-12-31 16:00:00+05:30,0.041999999999999996
+2020-12-31 17:00:00+05:30,0.087
+2020-12-31 18:00:00+05:30,0.112
+2020-12-31 19:00:00+05:30,0.243
+2020-12-31 20:00:00+05:30,0.176
+2020-12-31 21:00:00+05:30,0.08700000000000001
+2020-12-31 22:00:00+05:30,0.106
+2020-12-31 23:00:00+05:30,0.06

Project_1_24h-energy-consumption-prediction/artifacts/fast_track/metrics.csv

@@ -0,0 +1,3 @@
+,MAE,WAPE,sMAPE,RMSE
+HistGradientBoosting,0.047507219928675626,0.3489970242694261,0.3604810029785683,0.06551220231679815
+baseline,0.061583333333333344,0.45240281603917976,0.464261941461343,0.08665496715903441

Project_1_24h-energy-consumption-prediction/data/INFO.md

@@ -0,0 +1 @@
+## Place the dataset here (Ex: CEEW - Smart meter data Bareilly 2020.csv)

Project_1_24h-energy-consumption-prediction/readme.md

@@ -0,0 +1,45 @@
+# Fast-Track 24h Electricity Demand Forecast
+
+This project provides a lightweight, end-to-end forecasting pipeline for predicting the next 24 hours of electricity demand for Bareilly and Mathura, using historical smart meter data and optional weather forecasts.
+
+## Why This Exists
+
+Power utilities need quick, reasonably accurate demand forecasts for operational decisions like generation scheduling, demand response, and grid balancing. Traditional forecasting setups often require complex feature pipelines or proprietary software.
+
+This script aims to bridge that gap by:
+- Automating data cleaning & hourly aggregation from smart meter CSVs.
+- Using seasonal naive and ridge regression methods for fast forecasts.
+- Optionally enriching with weather data via Open-Meteo.
+- Producing plots, metrics, and a PDF report within minutes — no complex MLOps stack needed.
+
+## How It Works
+
+1. Data ingestion:  
+   Loads smart meter readings, cleans timestamps, finds the energy column, and resamples to hourly.
+
+2. Preprocessing:  
+   - Imputes short missing gaps (forward fill + interpolation).  
+   - Caps extreme outliers at 1st/99th percentile.
+
+3. Optional Weather: 
+   Fetches hourly temperature forecasts for Bareilly and merges with demand data.
+
+4. Feature Engineering: 
+   Builds time features (hour, sin/cos, day of week), lagged values, and 24h rolling means.
+
+5. Forecasting: 
+   - Baseline: Seasonal naive = repeat yesterday’s hour values.  
+   - Ridge Regression: Trains on last N days, autoregressively predicts next 24 hours.
+
+6. Evaluation & Reporting:  
+   If ground truth is available for the forecast horizon, computes MAE, WAPE, sMAPE and RMSE.  
+   Generates plots and a 2-page PDF report with metrics and visualizations.
+
+## Usage
+
+# Run pipeline for Bareilly, 7 days history, without weather
+python run_forecast.py --city Bareilly --data_path "data\CEEW - Smart meter data Bareilly 2020.csv" --history_window 7 --with_weather false --make_plots true --save_report true
+
+
+# Run pipeline for Mathura, 7 days history, without weather
+python run_forecast.py --city Mathura --data_path "data\CEEW - Smart meter data Mathura 2020.csv" "data\CEEW - Smart meter data Mathura 2020.csv" --history_window 7 --with_weather false --make_plots true --save_report true

Project_1_24h-energy-consumption-prediction/requirements.txt

@@ -0,0 +1,6 @@
+pandas
+numpy
+scikit-learn
+matplotlib
+joblib
+requests