data_generator.py

import pandas as pd
import numpy as np
import os

def generate_synthetic_data(num_meters=50, days=7):
    """
    Generates synthetic 15-minute interval smart meter data.
    """
    print(f"Generating synthetic data for {num_meters} meters over {days} days...")
    
    # 15-minute intervals
    date_rng = pd.date_range(start='2024-01-01', end=f'2024-01-{1+days}', freq='15min', inclusive='left')
    
    data = []
    
    for i in range(num_meters):
        meter_id = f"MTR_{1000 + i}"
        
        # Base consumption pattern (daily seasonality)
        time_of_day = date_rng.hour + date_rng.minute / 60.0
        # Peak around 18:00 to 22:00, low at night
        base_consumption = 0.5 + 1.5 * np.sin(np.pi * (time_of_day - 6) / 12)
        base_consumption = np.clip(base_consumption, 0.1, 5.0)
        
        # Add random noise
        noise = np.random.normal(0, 0.2, len(date_rng))
        consumption = base_consumption + noise
        consumption = np.clip(consumption, 0, None) # Ensure non-negative
        
        df = pd.DataFrame({
            'Timestamp': date_rng,
            'Meter_ID': meter_id,
            'Consumption_kWh': consumption,
            'Is_Anomaly': 0,
            # Generate a mock Grid Stress Score for the map later (0-100)
            'Grid_Stress': np.random.uniform(20, 80, len(date_rng)) 
        })
        
        # Introduce anomalies (e.g. sudden drop simulating tampering or outage)
        if np.random.rand() < 0.2: # 20% of meters have anomalies
            anomaly_start_idx = np.random.randint(len(df) // 2, len(df) - 20)
            anomaly_duration = np.random.randint(4, 16) # 1 to 4 hours
            df.loc[anomaly_start_idx:anomaly_start_idx+anomaly_duration, 'Consumption_kWh'] *= 0.1 # 90% drop
            df.loc[anomaly_start_idx:anomaly_start_idx+anomaly_duration, 'Is_Anomaly'] = 1
            df.loc[anomaly_start_idx:anomaly_start_idx+anomaly_duration, 'Grid_Stress'] = np.random.uniform(80, 100, anomaly_duration + 1)
        
        data.append(df)
        
    final_df = pd.concat(data, ignore_index=True)
    
    # Also create a mock metadata file for the map (lat, lon for each hash)
    lat_center = 12.9716 # Bangalore lat
    lon_center = 77.5946 # Bangalore lon
    
    metadata = []
    for i in range(num_meters):
        meter_id = f"MTR_{1000 + i}"
        # Randomize locations around Bangalore
        lat = lat_center + np.random.normal(0, 0.05)
        lon = lon_center + np.random.normal(0, 0.05)
        metadata.append({'Meter_ID': meter_id, 'Latitude': lat, 'Longitude': lon})
        
    meta_df = pd.DataFrame(metadata)
    
    # Save to data directory
    os.makedirs('data', exist_ok=True)
    final_df.to_csv('data/raw_interval_data.csv', index=False)
    meta_df.to_csv('data/raw_meter_metadata.csv', index=False)
    print("Raw data generated and saved to data/ directory.")
    
if __name__ == "__main__":
    generate_synthetic_data()