Cluster Analysis and Model Comparison Using Smart Meter Data

Area of Science:

• Electrical Engineering
• Data Science
• Time Series Analysis

Background:

• Load forecasting is essential for smart grid operations, influencing demand response, asset management, and investment decisions.
• Accurate predictions are key for efficient grid management and resource allocation.

Purpose of the Study:

• To explore the benefits of short-term load forecasting using various statistical and mathematical models.
• To address computational challenges in time-series data analysis for load prediction.
• To present a business case for analyzing customer consumption patterns and predicting behavior.

Main Methods:

• Utilized time-series forecasting techniques, including artificial neural networks, auto-regression, and Auto-Regressive Integrated Moving Average (ARIMA) models.
• Developed a business case to cluster data and identify factors influencing load consumption.
• Evaluated model performance based on prediction accuracy.

Main Results:

• The Auto-Regressive Integrated Moving Average (ARIMA) model with parameters (P, D, Q) set to (1, 1, 1) demonstrated the highest prediction accuracy.
• Analysis of customer behavior based on consumption parameters was performed.

Conclusions:

• Short-term load forecasting is critical for smart grid efficiency and planning.
• The ARIMA (1,1,1) model is a highly accurate method for short-term load prediction.
• Understanding customer consumption patterns enhances forecasting capabilities.