Climate change and climate transition: A Climate Analytics Project
In case people have not noticed, October 2024 was pretty warm in the Northeast. In New York City, the average monthly temperature was the hottest in the past century.
The climate change challenge touches every aspect of our lives from energy to transportation; from food and agriculture to steel, cement and infrastructure; from industry to our homes.
We can be more efficient in our personal choices.
We can try to transition from traditional ways of doing things to more sustainable ways – my family switched to an EV last year, and we all like it. (Note: we live in NYC and I go to school outside of Boston that creates an apparent challenge in driving an EV back-n-forth, but it turns out it is very doable and efficient).
Smart engineering solutions will help us avert the consequences of increased intensity natural disasters such as Storm Daniel in Greece in 2023 https://en.wikipedia.org/wiki/Storm_Daniel (I had just left Greece a week earlier), or hurricanes Helene https://en.wikipedia.org/wiki/Hurricane_Helene and Milton https://en.wikipedia.org/wiki/Hurricane_Milton in the Americas in 2024.
AI is sure to unlock efficiencies, from how to control power consumption in our buildings to how to optimize transportation solutions, to help invent new solutions to help address climate change. It is also consuming a lot of electricity, and I am trying to read more about that topic.
An example: in my junior year CS / data science class I suggested (but, unfortunately, did not get approval to do) a project to study the temperature variation inside and outside Andover buildings, and examine possible thermostat control policies that would try to reduce power consumption. Here is some detail:
I had access to daily temperature data in Andover, MA, and specifically for each day I had the noon temp (or average temp, depending on the source), min temp (probably around 5am) and max temp (probably around 2pm). This is daily data, and I had reliable access to the past 30 to 50 years (depending on the source).
I could, therefore, analyze the data to examine variability month-to-month and year-over-year with focus on Sept-May school year period.
Approximate hourly temperature profiles: I did not have historical data going back decades for hourly data for each day, but I had data for the past few years and certainly the current year. For each month and each day, I was planning to create an hourly temperature profile curve, and combine them to form a typical hourly profile for each month that I would then apply to create approximated hourly profiles for my datasets. Two missing pieces for my analysis that I needed to figure out were:
How to normalize these curves by their (min, max, avg) temperatures, so that I can then combine them.
How to apply the normalized profile to any specific day to achieve the right (min, max, avg) for that day while following that profile closely.
Power usage to heat the buildings is a function of the difference of the desired temperature inside the buildings, to the outside temperature, and depends on the insulation of the buildings. For example, if the target is to keep the building always at 72 degrees during the September to May months, I could compute the daily difference between the avg temp on that day and the target of 72 degrees, and take the absolute value. And, crudely, approximate the power consumption as the absolute value of that difference, allowing for heating and cooling, or try to focus only on the heating requirement on days where the outside average temperature was below the target temperature of 72 degrees.
I would repeat the daily power calculation using the hourly profiles, and summarize the data by month and examine year-over-year fluctuations.
The last step would be to consider how to minimize the power usage by picking the target temperature. The simplest way would be to repeat the previous analysis by varying the target temperature and study the effect. And use the analysis to identify if 1 degree difference would have a significant effect on power consumption. I would hope to do the same analysis by setting different target daytime and overnight temperatures.
There is a huge issue that I do not know yet how to study, which has to do with how building insulation affects the power consumption calculations, and how to customize the study to newer and older buildings that we have on our campus.
I imagine that we should be able to use next day forecast information to change the daily target temperatures but I did not find easy access to that historical forecast information stored for me to download. But I am sure it must exist and be available. Additionally, I believe that one should be able to model the types of temperature dynamics that different buildings may have that can improve the above analysis. Something to try in the summer.
