Build Real-Time Production Forecasts with TimeGPT-1 and Darts

                      
                     
import os
import pandas as pd
from darts import TimeSeries
from darts.models import RNNModel
from sklearn.model_selection import train_test_split

# Configuration
API_KEY = os.getenv('TIMEGPT_API_KEY')
DATA_PATH = os.getenv('DATA_PATH', 'data/production_data.csv')

# Load and prepare data
try:
    data = pd.read_csv(DATA_PATH)
    series = TimeSeries.from_dataframe(data, time_col='date', value_cols='production')
except Exception as e:
    raise RuntimeError(f"Error loading data: {str(e)}")

# Split data into training and validation sets
train, val = train_test_split(series, train_size=0.8)

# Initialize model
model = RNNModel(input_chunk_length=24, output_chunk_length=12, n_epochs=100)

# Train the model
try:
    model.fit(train)
except Exception as e:
    raise RuntimeError(f"Error during model training: {str(e)}")

# Make predictions
try:
    forecast = model.predict(n=12)
    forecast.plot(label='Forecast')
except Exception as e:
    raise RuntimeError(f"Error during prediction: {str(e)}")

if __name__ == '__main__':
    print(f"Forecast for the next 12 time steps: {forecast}")