Machine learning can be used to improve energy consumption in cities

The City of Philadelphia has set a goal of halving greenhouse gas emissions from the built environment by 2030. Carbon reduction programs targeting the commercial and residential construction sectors are a top priority, as buildings and commercial facilities are the region’s most significant contributor to greenhouse gas (GHG) emissions

Researchers at Drexel University’s College of Engineering are attempting to predict how energy use will change as neighborhoods change. They’re using a machine learning model they’ve developed to help them.

In 2017, the city stated a goal of becoming carbon neutral by 2050, with reduced greenhouse gas emissions from energy use of buildings accounting for about three-quarters of Philadelphia’s carbon footprint at the time. The problem for Philadelphia, one of the oldest cities in the nation, is that there is no one-size-fits-all approach to energy use due to the sheer variety of building styles.

But to achieve this, it is important to incorporate energy use forecasts into planning decisions that will shape future development, not only for new construction but also for existing buildings.

Simi Hoque, Ph.D., a professor in the College of Engineering who led the research on using machine learning for granular modeling of energy use recently published in the journal Energy & Buildings, said: “For Philadelphia in particular, neighborhoods vary so much from place to place in the prevalence of certain housing characteristics and zoning types that it is important to tailor energy programs for each neighborhood rather than trying to implement blanket carbon reduction policies across the entire the city or county.”

Hoque’s team believes that properly implemented existing machine learning programs can provide insight into how zoning decisions may affect future greenhouse gas emissions from buildings.

She said, “Right now, there is a huge volume of energy usage data, but often it is simply too inconsistent and messy to be used reasonably. For example, a data set corresponding to certain housing characteristics may have usable energy estimates. However, another dataset corresponding to socioeconomic characteristics is missing too many values ​​to be usable. Machine learning is well equipped to handle this challenge because it can iteratively learn and improve through the training process to reduce bias and variance despite these data limitations.

Researchers have developed a technique for extracting insights from fragmented data by combining two machine learning programs. One that can extract models from massive amounts of data and use them to make predictions about future energy, and a second that can pinpoint details in the model that most likely had the most significant impact on changing projections.

They used extensive data on commercial and residential energy use for Philadelphia from the US Energy Information 2015 Residential and Commercial Buildings Energy Use Survey, as well as demographic and socioeconomic information of the city from the US Census American Communities Survey Bureau, to train a deep learning program called Extreme Gradient Boosting (XGBoost).

The program gained enough insights from the data to establish correlations between a long list of factors, including building density, population in a specific area, building size, number of occupants, number of days in which heating or cooling was used and the energy consumption for each dwelling or facility.

While deep learning models like XGBoost are excellent for producing accurate predictions, the complexity of their operations may make it difficult to understand how they work.

The researchers used an analysis of Shapley’s additive explanations, a method used in game theory to distribute credit among factors that contributed to an outcome, to decode the estimates and recommendations of the so-called “black box” program.

They were able to determine how much a change in building density or square footage, for example, affected the projection made by the program.

Hoque said, “Machine learning models like XGBoost learn to browse datasets to perform a specific task such as generating a reliable prediction of a system, but they don’t pretend to understand or represent the field relationships that underlie a phenomenon, and while Shapley the analysis can’t tell us which features have the biggest impact on energy usage, it can explain which features had the biggest impact on the model’s energy usage prediction, which is still pretty useful information.

The team then tested the model by providing data from a hypothetical scenario provided by the Delaware Valley Regional Planning Commission, which predicted Philadelphia’s continued economic development throughout the 20th century.

The scenario was for a 17% increase in population with a corresponding increase in households and a variety of job and income opportunities by region around the city.

The model predicted how future residential and commercial development would impact energy-driven greenhouse gas emissions from buildings in 11 distinct sections of the city for each scenario, and which variables played a key role in creating the projections.

In the 2045 scenario, six of the 11 areas would reduce their energy consumption, especially in low-income areas. Mixed-income regions, such as the northernmost half of the city, including Oak Lane, would likely increase their energy use.

The presence of connected (low energy) versus single family (high energy) single family homes played an important role in Shapley’s analysis, with high monthly electricity costs, sub-acre lot sizes, and a fewer rooms to build all helping to reduce energy use projections.

Overall, the Home Energy Prediction model found that characteristics associated with lower building intensity are associated with lower energy consumption predictions in the model.

They wrote, “These findings provide reason to reexamine the effects of upzoning policies, commonly found as affordable housing solutions in Philadelphia and other US cities, and subsequent changes in energy use for these areas.”

The most important details in this text are that the machine learning model predicted little change in energy use under 2045 conditions, and that Shapley’s analysis identified building square footage and employee count as the most important predictors of energy use for most types of commercial buildings.

Hoque said, “I see a lot of potential in using machine learning models like XGBoost to predict increases or decreases in energy use due to new construction projects or policy changes. For example, the construction of a new rail line in a neighborhood can change the demographics and employment of a neighborhood. Their methods would be ideal for incorporating such information in the context of an energy forecasting model.”

The team understands that more testing is needed and that the program will improve as more data is sent. They propose that research continues by focusing on parts of the city with known high energy use and performing a Shapely Analysis to identify some of the elements that may be contributing to it.

The researcher said, “We hope this will provide a resource for future researchers and policymakers so they don’t have to look at the entire city of Philadelphia, but can focus on neighborhoods and variables that we’ve flagged as areas of potential importance. Ideally, future studies would use more interpretable methods to test whether these features correspond to higher or lower energy estimates in a given area.”

According to the study, commercial buildings in the upper quantiles of square footage and number of employees should be prime targets for energy reduction programs, with an approximate threshold of 10,000 square feet of total building area considered a priority due to the disproportionate influence of the model on energy prediction.

The researchers caution against assuming a direct relationship between variables and changes in energy consumption in the model. However, they believe it is still very useful for its ability to provide planners with both a high-level and granular look at the interaction between zoning decisions and development and their effect on energy use.

The study suggests that commercial buildings in the top quantiles of square footage and number of employees should be prime targets for energy reduction programs, with an approximate threshold of 10,000 square feet of total building area being prioritized. due to their disproportionate influence on the energy prediction of the model.

The researchers caution against assuming a direct link between energy consumption variables and changes in the model, but suggest that it is still quite useful due to its ability to provide planners with both a high-level and granular look at the interplay between decisions. of zoning and development and their effect on energy consumption.

Magazine reference:

  1. Shideh Shams Amiri, S., Mueller, et al. Investigate the application of a commercial and residential energy consumption prediction model for urban planning scenarios with Machine Learning and Shapley Additive explanation methods. Energy and Construction. DOI: 10.1016/j.enbuild.2023.112965

#Machine #learning #improve #energy #consumption #cities

Leave a Comment