The City of Philadelphia has set a goal to halve GHG emissions from the built environment by 2030. Commercial buildings and facilities are the largest contributors to greenhouse gases in the region, so carbon reduction programs targeting the business and residential building sectors are a top priority. Emissions (GHG)
Drexel University engineering researchers are trying to predict how energy usage will change as neighborhoods change. They use machine learning models that they develop to help them.
In 2017, the city declared a goal of becoming carbon neutral by 2050. Reducing greenhouse gas emissions from building energy use accounted for about three-quarters of Philadelphia’s carbon footprint at the time. The problem with Philadelphia, one of the oldest cities in the United States, is that there is so much variation in building styles that there is no one-size-fits-all approach to energy use.
To achieve this goal, however, it is critical to incorporate energy use projections into zoning decisions that will shape future development, not just for new construction, but for buildings already in use.
Dr. Simi Hoque, an engineering professor who led a study recently published in Energy & Buildings that used machine learning to model granular energy use, said: “Especially in Philadelphia, the prevalence of particular housing characteristics and zoning types varies greatly by region, so rather than trying to enact comprehensive policies for carbon reduction city-wide, energy programs can be developed on a region-by-region basis. Customization is key. County.”
Hoque’s team believes that well-implemented existing machine learning programs can provide insight into how zoning decisions will affect future greenhouse gas emissions from buildings.
she said, “There is currently an enormous amount of energy use data, but it is often too inconsistent and messy to be rationally used. , there may be energy estimates available, but another dataset corresponding to socioeconomic features has too many unusable values.Machine learning is able to perform well despite the limitations of these data , is well equipped to handle this challenge because it can iteratively learn and improve through the training process, reducing bias and variance.”
The research team combined two machine learning programs to develop a technique for extracting information from fragmented data. One can extract patterns from vast amounts of data and use them to make predictions about future energy, and the other is the model that most likely had the greatest impact on changes in predictions. details can be identified.
They used Philadelphia’s vast commercial and residential energy use data from the 2015 US Energy Information’s Residential Energy Consumption Survey and Commercial Buildings Energy Consumption Survey, as well as city demographics and data from the US Census Bureau’s American Communities Survey. A deep learning program called Extreme Gradient Boosting (XGBoost) using socio-economic information.
The program learns enough knowledge from the data to find out between a variety of factors, such as building density, population in a given area, building size, number of occupants, number of days heated or cooled, and energy use of each household. created a correlation of Or structure.
Deep learning models like XGBoost are great at producing accurate predictions, but their complexity makes it difficult to understand how to operate them.
Researchers used Shapley’s additive explanatory analysis, a technique used in game theory to distribute credit among factors that contributed to an outcome, to decipher the estimates and recommendations of so-called “black box” programs. Did.
For example, we were able to determine how changes in building density and area affected programmatic projections.
Hawk said, “Machine learning models like XGBoost learn how to process datasets to perform specific tasks, such as generating reliable predictions of systems, but they also understand the field relationships that underlie phenomena. The Shapley analysis does not tell us which features have the greatest impact on energy use, but which features have the greatest impact on the model’s energy use predictions. I can explain how I gave it. This is still very useful information.”
The team then tested the model by providing data from a hypothetical scenario given by the Delaware Valley Regional Planning Commission, which predicted continued economic development in Philadelphia throughout the 20th century.
This scenario projects a 17% increase in population, a corresponding increase in the number of households, and a wide variety of job and income opportunities in peri-urban neighborhoods.
The model shows how future residential and commercial development will affect greenhouse gas emissions from building energy use in 11 different city sections for each scenario, and which variables are important in making projections. Predicted how it played its role.
In the 2045 scenario, 6 out of 11 regions will reduce their energy use, mostly in low-income areas. Mixed-income areas, such as the northernmost half of the city, which includes Oak Lane, may see increased energy use.
Shapley’s analysis found that there was a mix of single-family annexed (low energy use) and single-family homes (energy The presence of dwellings played an important role. Build everything that helps reduce your energy usage projections.
Overall, the residential energy prediction model found that traits associated with lower construction intensity were associated with lower energy consumption predictions in the model.
they wrote, “These results provide a rationale for reexamining the impact of upzoning policies, which commonly exist as affordable housing solutions in Philadelphia and other cities across the country, and subsequent changes in energy use in these areas. will give you.”
The most important detail in this text is that the machine learning model predicted little change in energy use under 2045 conditions, and that the Shapley analysis showed that building area and number of employees were the same for most types of commercial buildings. identified as the most important predictor of energy usage.
Hawk said, “We see great potential in using machine learning models like XGBoost to predict changes in energy use due to new construction projects or policy changes. Neighborhood demographics and employment can then change, and their method is ideal for incorporating that information into the context of energy forecasting models.”
The team understands that further testing is needed and that the program will improve as more data is sent. We propose to continue our research by performing a Shapely analysis to identify some of the factors that may have contributed to the focus.
researchers said, “Hopefully this will provide a resource for future researchers and policy makers so they don’t have to survey the entire city of Philadelphia, and can focus on the areas and variables they flag as potentially important.” Ideally, future studies will use more interpretable methods to test whether these features correspond to higher or lower energy estimates in specific regions. will do.”
According to the study, commercial buildings in the upper quantiles of area and number of employees should be prime targets for energy reduction programs, with the model’s disproportionate impact on energy projections resulting in approximately 10,000 square feet is preferred.
Researchers are careful not to assume a direct relationship between variables and changes in energy use within the model. However, they still find it very useful, as it can provide planners with a high-level and detailed view of the interplay of zoning decisions and development, and their impact on energy use.
The study suggests that commercial buildings in the top quantiles of area and number of employees should be prime targets for energy reduction programs. A threshold of approximately 10,000 square feet of total building area is preferred due to its disproportionate impact on energy projections. of the model.
Researchers are careful to assume direct links between model variables and changes in energy use, but they do not provide planners with a high level of detail about the interplay between zoning decisions and development. suggest that the model is still very useful because of its ability to give an accurate perspective. Impact on energy use.
Journal reference:
- Shideh Shams Amiri, S., Mueller, et al. We investigate the application of commercial and residential energy consumption prediction models for urban planning scenarios using machine learning and the Shapley Additive explanation method. energy and buildings. DOI: 10.1016/j.enbuild.2023.112965
