Functional Principal Component Analysis as a Versatile Technique to Understand and Predict the Electric Consumption Patterns

TL;DR

This paper demonstrates the use of Functional Principal Component Analysis (FPCA) to analyze and predict electric consumption patterns, aiding smart grid management and energy efficiency efforts.

Contribution

It introduces FPCA as a versatile tool for understanding and forecasting electricity consumption at various spatial scales, validated on real-world data from Milan.

Findings

FPCA effectively decomposes load patterns into meaningful principal functions.

The method identifies physical and behavioral factors influencing consumption.

Long-term predictions based on FPCA are accurate and reliable.

Abstract

Understanding and predicting the electric consumption patterns in the short-, mid- and long-term, at the distribution and transmission level, is a fundamental asset for smart grids infrastructure planning, dynamic network reconfiguration, dynamic energy pricing and savings, and thus energy efficiency. This work introduces the Functional Principal Component Analysis (FPCA) as a versatile method to both investigate and predict, at different level of spatial aggregation, the consumption patterns. The method was applied to a unique and sensitive dataset that includes electric consumption and contractual information of Milan metropolitan area. The decomposition of the load patterns into principal functions was found to be a powerful method to identify the physical and behavioral causes underlying the daily consumptions, given knowledge of exogenous variables such as calendar and meteorological data. The effectiveness of long-term predictions based on principal functions was proved on Milan's metropolitan area data and assessed on a publicly-available dataset.