Energy Data Engineering-2-Project1

ARIMA models

Energy Data Visualization and Time Series Analysis

Problem Statement: How CO2 is effected by energy sources. A detailed data visualisation needs to be done to identify the trends. Create a synthetic model for the future year by fitting to the best ARIMA model.

Features:

Data Visualization:

• Visualize the energy data to identify patterns and trends effectively.

We can see from the graph that the energy source is more occuring in 2000-4000MW.

Trend and Seasonal Decomposition:

• Perform additive decomposition since seasonal growth is almost varying around 6000.

Decomposition of Daily Profiles:

• Break down daily energy profiles to understand underlying patterns.

Model Selection:

• Residuals Stationarity: Check stationarity of residuals using the Augmented Dickey-Fuller (adfuller) test.
• Autocorrelation Analysis: Plot the autocorrelation and partial autocorrelation functions for residuals.
• ARIMA Model Comparison: Compare at least 5 different ARIMA models based on parameters p, d, and q.
• Evaluate models using the Ljung-Box test to check the quality of fit.

Model Deployment:

Rolling Forecast:

• Deploy a rolling forecast on the last year of training data.
• Visualize the forecast including uncertainty ranges.
• Evaluate forecast quality using RMSE (Root Mean Square Error).

Synthetic Profile Creation:

• Created a synthetic profile for 2024.
• Deployed a statistical model for the residuals to closely imitate the time series in terms of autocorrelation, moving average, variance, and mean.