Statistical Learning Models for Estimating Retail Tariffs of Electricity in the United States
Open AccessRetail electricity tariffs represent how electric utilities charge their customers for the electricity they consume. They consist of several types of monetary charges: consumption-based, fixed, or based on minimum and maximum demand. This data is necessary for performing economic analysis on clean-energy projects (e.g., rooftop solar energy, home energy storage, etc.) In the United States, this data is scattered across thousands of electric utility websites. To mitigate this problem, in 2012, the US National Renewable Energy Laboratory launched a crowdsourcing website for collecting this data, organizing it, and making it available to the public for free. Despite this fruitful effort, the database includes outdated or no tariffs in thousands of jurisdictions. This makes it challenging to run economic analysis for clean-energy projects in those areas.This praxis explores a variety of statistical learning methods to estimate missing tariffs based on available predictors, such as electricity prices in neighboring jurisdictions, the type and size of electric utility, the average price of electricity in the state, etc. To this end, several statistical inference approaches are considered: some that require ancillary data (deep neural networks, k-nearest neighbors, decision trees, linear regression, support vector machines, areal interpolation) and others that do not require any additional data beyond known tariffs in neighboring jurisdictions (inverse distance weighting, ordinary Kriging, average of the n nearest geographical neighbors, average within a radius). Eleven models are constructed to estimate the different type of charges: fixed charges and energy charges for residential, commercial, and industrial customers while considering both time- dependent and time-independent rate structures. Specifically, this praxis makes three key practical contributions. First, it studies the literature to determine which factors impact the different charges in electricity tariffs. Second, it determines how various inference approaches perform for estimating tariff charges anywhere in the US with a commentary on their advantages and limitations. Third, it proposes a probabilistic formulation to describe the inference error.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.