Carbon Lighthouse: Illuminating Energy Savings with Machine Learning

Carbon Lighthouse delivers 20-30% energy savings to their clients using a cloud-based, machine-learning AI.

Who is Carbon Lighthouse?

Carbon Lighthouse (CL) is a clean-energy consulting firm that offers advice on how to reduce their clients’ energy consumption. With a portfolio containing over five-hundred buildings and doubling annual revenue growth for the past seven years, they are clearly doing something beyond the every-day competition. What is their edge? They use machine learning and artificial intelligence to uncover hidden efficiencies beyond the typical low hanging-fruit of LED lighting and HVAC improvements. By using a complex network of sensors, CL monitors and corrects building performance to deliver sustained savings through a guaranteed subscription service. The company claims that these savings amount to an additional twenty to thirty percent efficiency gain above the typical Building Management Systems (BMS) offered by competitors.

 

How Does it Work?

[1]

As described by CEO Brenden Millstein above, when CL is hired, they first deploy 50-300 sensors to measure electrical flows, chiller fluid flows, occupancy rates, pressures and lighting flows. The collected data is then sent to a cloud server, where it is analyzed by the company’s proprietary machine-learning AI, known as CLUES® (Carbon Lighthouse Unified Engineering System). CLUES® uses tens of millions of data points to predict the financial and energy-based benefit of thousands of potential actions to identify the best potential energy-savings measures for the client.

These recommendations are twofold. First, CLUES® identifies possible one-time changes such as lighting retrofits, valve changes, and solar additions. Second, the system monitors and optimizes the building’s controls every five minutes. It is this latter machine-learning driven control that differentiates CL from its competitors, generating 80% of the client’s savings over time. [2]

How Important is the Machine Learning?

Incredibly. Buildings are extremely dynamic; weather, occupancy, and energy demand change throughout the day, necessitating real-time control. Moreover, in the clean-energy business, too many building owners have been burned by unsuccessful energy-efficiency projects. Carbon Lighthouse’s VP of Engineering, Matt Gasner, in an interview with GreenBiz reporting, stated that 20% of energy savings projects are performing at less than 50% of their expected benefit within a few years. [3]

Hence, accuracy matters. To this point, studies have shown that in measuring and verifying energy consumption, machine learning methods are up to 50% more effective than traditional means. [4] With such an increase in accuracy one might think one has to move more slowly, but the reverse is true. In fact, CLUES® has also sped up CL’s implementation time from eighty person-days to just three. [5] It’s clear that the megatrend of Machine Learning is essential to CL’s business.

 

Management Addresses the Issues

To combat the perception that energy-efficiency projects don’t payoff, CL has instituted an energy savings guarantee, where they contractually oblige themselves to deliver a dollar amount of savings per annum. Should CL fail to deliver, they refund the difference. Thus, they have removed the upfront risk for the customer, while only exercising refunds in 3% of cases. [2]

Strategically, management has also set up CL for long-term success. Most recently, the company landed a 27-million-dollar round of seed funding in order to pursue growth of their engineering teams and further develop the CLUES® product. [6] Much of that growth has been focused in Hawaii where regulators have set an aggressive 100% renewable energy target by 2045, suggesting a need to reduce demand by four-thousand gigawatt-hours. [3] Additionally the company appears to be focusing on long-term contracts that boost customer lifetime value, an example of which is Tesla, who has signed a six-year contract to reduce utility bills by $90,800 annually. [2]

Thinking more long-term, the company vows it will stay private: “We aren’t trying to build a ‘flip’ company to get acquired or end with an IPO — we’re committed to solving the climate change conundrum for the next 150 years.” [6]

 

Recommendations to Management

As CL’s AI gets faster, consider shifting from large clients to smaller ones. Specifically, focus more heavily on small to medium-sized manufactures, as studies suggest they have significant energy savings potential and are unlikely to be able to reduce their consumption with internal talent. [7] These are companies with fewer than 500 employees, less than 100-million-dollars in sales, and annual energy costs of 100 thousand to 2.5-million-dollars.

 

Food for Thought

  • Can CL use its rich data sets to inform product development of companies specializing in cleantech and IoT?
  • CL’s stated mission is to end global warming. CEO Brenden Millstein has remarked that they have saved four power plants worth of CO2, so that leaves only fifty-thousand more. How can CL scale their business more effectively?
  • Do you agree that the company should remain private?

 

(790 words)

 

References

[1] Carbon Lighthouse. (2018). How We Work. [online] Available at: https://www.carbonlighthouse.com/how-we-work/ [Accessed 13 Nov. 2018].

[2] St. John, J. (2018). How Carbon Lighthouse Mines Wasted Energy From the Built Environment. [online] Greentechmedia.com. Available at: https://www.greentechmedia.com/articles/read/how-carbon-lighthouse-mines-wasted-energy-from-the-built-environment#gs.=qCwBK8 [Accessed 13 Nov. 2018].

[3] GreenBiz. (2018). Day 1, Sidebar Discussion with Matt Ganser, Carbon Lighthouse. [online] Available at: https://www.greenbiz.com/video/day-1-sidebar-discussion-matt-ganser-carbon-lighthouse [Accessed 13 Nov. 2018].

[4] Gallagher, C., Bruton, K., Leahy, K. and O’Sullivan, D. (2018). The Suitability of Machine Learning to Minimise Uncertainty in the Measurement and Verification of Energy Savings. Energy and Buildings, 158, pp.647-655.

[5] Carbon Lighthouse. (2018). How Carbon Lighthouse Got Its CLUES. [online] Available at: https://www.carbonlighthouse.com/carbon-lighthouse-got-clues/ [Accessed 13 Nov. 2018].

[6] Carbon Lighthouse. (2018). Carbon Lighthouse Closes $27M Growth Round. [online] Available at: https://www.carbonlighthouse.com/carbon-lighthouse-closes-27m-growth-round-proving-profit-driven-carbon-elimination-is-a-thing/ [Accessed 13 Nov. 2018].

[7] DuttaGupta, A. & Egbue, O., PhD. 2017, “Energy Efficiency Using Machine Learning – Targeting Small and Medium- Sized Manufactures”, IIE Annual Conference.Proceedings, pp. 976-981.

Additional Background

[8] Heikell, L. (2014). Meet Carnegie Mellon’s Energy Sleuths. [online] Microsoft Stories. Available at: https://news.microsoft.com/features/meet-carnegie-mellons-energy-sleuths/ [Accessed 13 Nov. 2018].

[9] Hsieh, E. (2017). How Carbon Lighthouse uses machine learning to achieve energy efficiency. [online] GreenBiz. Available at: https://www.greenbiz.com/video/how-carbon-lighthouse-uses-machine-learning-achieve-energy-efficiency [Accessed 13 Nov. 2018].

 

Previous:

The Beauty of Crowdsourcing

Next:

Grupo Aval: Utilizing open innovation to build a new business model

Student comments on Carbon Lighthouse: Illuminating Energy Savings with Machine Learning

  1. I like how Machine Learning enables 1) solving problems in a more tangible way, and 2) have a clear idea of cost savings possible. I’d be curious the size of the total addressable market here – can they only do buildings or can they pivot to more energy intensive systems like manufacturing plant. I’d love to see a dashboard for companies so they can visualize this usage for these applications, and prioritize initiatives.

  2. Great article! And love the concept! I initially assumed this business was infinitely scalable (every company uses energy and wants to reduce its bills etc etc), but when I read that Tesla only was able to save $90k per year, I began to doubt how many companies would find this technology worth the hassle. While it seems relatively unobtrusive in the way it optimizes energy use, I’m curious whether the average company will find it worth the headache to go into business with these guys if the ultimate savings are <$100k

  3. Providing energy data to these companies as a service seems like a great concept. There’s a pretty simple value offering here. I wonder though what kind of changes the end customers have to implement in order to realize the energy savings they propose. Are these companies implementing infrastructural changes or simply changing the time that heating systems go on in the morning and off in the afternoon? To this question, offering a customer guarantee opens them up to a lot of risk, but does signal their commitment to their product.

Leave a comment