Or rather, “Tackling Climate Change with Machine Learning.” Today I will recap some of my favorite sections of the much-publicized article written by David Rolnick and others way back in 2019. You know, the long, long ago. These folks are at the forefront of AI research to mitigate the effects of climate change, and more people should pay attention to their important work.
The Role of Data Scientists, Entrepreneurs, and Technologists
I think we can all agree that with government dysfunction in recent decades, we can’t rely on them to act fast enough…even with the positive signals from the new administration. We also can’t expect human behavior to change on a mass scale, considering our current consumption patterns, the cultural and societal pressures to consume, and the billions of people in emerging economies around the world who want a better life. It would be absurdly hypocritical and unfair to ask people in India or Brazil to not eat meat or travel or consume like we do.
No, the greatest advances have to come through innovation and market-based solutions with the right incentives in place. New York Times veteran tech journalist Kara Swisher recently interviewed Bill Gates on his new book about climate change solutions, and they agreed that: “the world’s first trillionaire will be a green-tech entrepreneur. That’s trillionaire, with a ‘T’.”
Yes, climate change is one of the greatest challenges facing humanity, but like other problems, we come together to solve them. It’s not going to be easy, and we, as data scientists, may wonder how we can help. Enough preamble. Now I’ll highlight a few high leverage mitigation strategies in areas where AI companies are starting to make exciting progress.
Electricity Systems
Many electricity systems are awash in data, and the industry has begun to envision next-generation systems (smart grids) driven by AI. Electricity systems are responsible for about a quarter of human-caused greenhouse gas emissions each year, and there is a lot of waste and inefficiency in electricity generation and transmission. ML can contribute by accelerating the development of clean energy technologies, improving forecasts of demand and generation, optimizing electricity systems and management, and enhancing system monitoring.
Let’s discuss forecasting in particular: many ML methods have used historical demand and weather data, physical model outputs, images, and even video data to create short- to medium-term forecasts of generation and demand. These methods include time series approaches, supervised machine learning, fuzzy logic, and hybrid physical models. At a more granular level, some work has attempted to understand specific categories of demand, for instance by clustering households or by disaggregating electricity signals using game theory, optimization, regression, and online learning.
Myst AI is a San Francisco-based startup that secured a $6 million Series A financing round last autumn to continue developing an AI platform for electricity time series forecasting. My final project at Metis was on Forecasting Electricity Demand in Seattle, and you can find my code on GitHub and my corresponding blog post here to see how my model outperforms the government’s day-ahead forecast. Also, I recently applied to Myst’s data science internship program, and I want everyone reading this right now to wish me luck. Out loud. Say, “Good luck, Mike.”
Thanks, that very kind. Now read about buildings and cities.
Buildings and Cities
Inefficiencies in building management and energy use add up to another 25% of emissions, but ML has the potential to drastically lower that number. An essential step towards energy efficiency is making sense of the increasing amounts of data produced by meters, sensors, and home energy monitors.
Intelligent control systems in buildings can decrease the carbon footprint both by reducing the energy consumed and by providing means to integrate lower-carbon sources into the electricity mix. Specifically, ML can reduce energy usage by allowing devices and systems to adapt to usage patterns. Buildings can respond to signals from the electricity grid, providing flexibility to the grid operator and lowering costs to the consumer.
There’s also energy demand forecasts for specific buildings, which are useful for power companies and in evaluating building design and operation strategies. ML has the potential to speed up these forecasts across buildings by learning to approximate the physical model to reduce the need for expensive simulation (surrogate models).
In New York City, data scientists are building models of electricity load profiles with reinforcement learning and deep belief networks using data on commercial and residential buildings. They then used approximate reinforcement learning and transfer learning to make predictions about new buildings, enabling the transfer of knowledge from commercial to residential buildings, and from gas- to power-heated buildings.
Within a single building, understanding which appliances drive energy use is crucial for targeting efficiency measures, and can motivate behavioral changes. Promising ML approaches to this problem include hidden Markov models, sparse coding algorithms for structured prediction, harmonic analysis that picks out the “signatures” of individual appliances, and deep neural networks.
A company in New York City called Cortex is out to dramatically improve how buildings operate using analytics. I think they’re packaging their value proposition intelligently by focusing on reducing costs for building managers and improving the experience of their occupants, but in essence they are fundamentally improving the sustainability of commercial buildings. (Shoutout to Will Koehrsen - data scientist at Cortex, excellent writer, and inspirational figure among aspiring data scientists.)
Farms and Forests
Our current economy encourages sequestered carbon release through deforestation and unsustainable agriculture, accounting for another 25% of emissions.
How can ML help with this? Precision agriculture could reduce carbon release from the soil and improve crop yield, which in turn could reduce the need for deforestation. Satellite images make it possible to estimate the amount of carbon sequestered in a given area of land, as well as track GHG emissions from it. ML can help monitor the health of forests and peatlands, predict the risk of fire, and contribute to sustainable forestry. Sophisticated computer vision tools are often at the heart of these efforts.
Here I’ll focus on precision agriculture, which is basically the demand for sophisticated tools that allow farmers to work at scale more productively while being less destructive. ML in precision ag can help in disease detection, weed detection, and soil sensing. ML can also guide crop yield prediction and even build models that help farmers predict crop demand and decide what to plant at the beginning of the season. These solutions often use Unmanned Aerial Vehicles (UAVs) with hyperspectral cameras can be used for all of these tasks.
See Tree is an Israeli precision ag company that just secured a whopping $30M Series B funding round last December. They use distinct data collection techniques to scan and analyze hundreds of millions of trees, providing comprehensive metrics per tree, which helps their customers detect non-producing and underperforming trees that result in major losses in crop yield and fruit deterioration. They also help quantify the losses and enable accurate business decision making and root cause analysis.
Conclusion
We are part of the generation who has the power to do something about climate change. Our decisions over the next 10 years will have lasting impacts for generations for come. And that can be either incredibly daunting or thrilling depending on how you look at it. Today I gave some examples of data scientists using their skills to tackle climate change and make a lot of money doing so, and I hope this blog post leaves you with some hope and excitement about solving these sorts of seemingly intractable problems that aren’t going away anytime soon.
