TLDR: This research uses Explainable AI (XAI) methods like decision trees and Pearson’s Correlation Coefficient to analyze electricity usage in New Jersey, correlating infrastructure and socio-demographic data with energy features. It identifies housing tenure and racial demographics as key predictors, showing that renters and racially diverse groups often face higher energy burdens. The study introduces a novel energy equity web portal and an energy burden calculator that provides tailored advice based on user zip codes, aiming to inform energy policy and promote sustainable solutions.
Understanding who has fair access to affordable energy is a critical challenge in today’s world, especially for marginalized communities. This concept, known as energy equity, is deeply influenced by various factors, including income, race, housing type, and the age and quality of infrastructure. Often, communities that are already underserved face significant hurdles like outdated energy systems and inadequate housing, which can severely limit their ability to participate in energy efficiency programs. These issues contribute to what is known as ‘energy burden’ – where a disproportionate amount of income is spent on energy costs – and hinder the transition to more sustainable energy systems.
A recent study titled “Energy Equity, Infrastructure and Demographic Analysis with XAI Methods” by Sarahana Shrestha, Aparna S. Varde, and Pankaj Lal from Montclair State University, delves into these complex issues. The researchers utilized Explainable AI (XAI) methods, such as decision trees and Pearson’s Correlation Coefficient (PCC), to analyze electricity usage patterns. Their goal was to identify the key factors influencing energy consumption and equity, particularly in the Northeastern United States, with a specific focus on New Jersey. The data for this research was primarily sourced from the New Jersey Clean Energy Program and the U.S. Census Bureau, covering the years 2008-2022.
Unveiling Key Predictors of Energy Burden
The study’s findings highlight housing tenure (whether someone owns or rents their home) and racial demographics as crucial predictors of electricity consumption. The analysis revealed that renters and racially diverse groups frequently experience higher energy burdens. For instance, decision tree models showed that renter-occupied housing was a dominant factor, confirming existing disparities between renters and homeowners in terms of energy infrastructure. Furthermore, the importance of Asian-Americans (15.75%) and owned housing (13.12%) as features underscored the impact of homeownership on energy equity.
Pearson’s Correlation Coefficient (PCC) analysis further corroborated these observations, showing strong correlations between race and homeownership. White populations exhibited a high prevalence of homeownership, while Hispanic or Latino and Black populations were more likely to be renters, pointing to systemic disparities in housing infrastructure. The study also found that counties with more low-income households tended to have higher rates of renters. When examining race and income, the White population showed a strong positive correlation across various income categories (low, moderate, and high income), suggesting a broad distribution across income levels. In contrast, Hispanic or Latino populations often showed a more moderate correlation with low income, indicating a concentration in lower-income areas and income disparities within this group across different counties.
Also Read:
- Bridging Advanced AI and Power Grids: The Role of Federated Foundation Models
- Improving Building Efficiency with Causal AI: Introducing GRID
A Novel Solution: The Energy Equity Web Portal and Calculator
To translate these insights into actionable solutions, the researchers developed and demonstrated a novel energy web portal featuring an energy burden calculator. This tool is designed to be user-friendly: a user simply inputs their zip code, and the calculator determines their energy burden using a specific formula that considers annual electricity and heating consumption, their respective rates, and the median household income of the area. If the calculated energy burden is higher than the state average, the portal provides clear, explainable action items and tailored advice to help users reduce their energy costs and promote energy equity. If the burden is below the state average, it informs users of their favorable situation.
This innovative work is among the first to explicitly address energy equity using XAI methods, coupled with a practical web portal and energy burden calculator. It aims to tackle significant challenges in energy policy adaptation and pave the way for next-generation systems that foster greater energy equity. Future work includes making the web portal more interactive, conducting more granular analyses (e.g., by energy source, season, or finer demographic attributes), and enhancing the calculator with expected energy usage projections. For more details on this research, you can refer to the full paper available here.
