The future of energy: forecasting the weather?

NextEra:utilizing machine learning to increase reliability and integrate renewables into the grid

By 2050, the United States’ electricity demand will be 40% greater than today’s.  This growth will rely heavily on the production of wind and solar power, which is expected to make up over 64%[i] of the electrical grid supply.   To meet this demand, machine learning will be needed to optimally integrate power sources into the grid, predicting demand and source capacity, and to limit downtime through predictive maintenance.

NextEra Energy Resources, is the world’s largest supplier of wind and solar energy and has the largest battery storage capability[ii] compared to other providers. Their subsidiary, Florida Power and Light (FPL), currently serves 10 million people at a cost 30% lower to consumers than the national average[iii].  Part of this is through machine learning employed by their predictive analytics team that leverages real time equipment data to predict when reliability issues may occur.  The machine learning algorithms detect anomalies in their critical equipment, such as combustion turbines, and schedules preventative maintenance in advance of potential failures, increasing production reliability[iv]. Since 2006, they have installed over 5 million smart meters resulting in the meters being used by 99% of current customers[v]. The data allowed FPL to proactively pinpoint which transformers need to be replaced, a 25% cost savings in comparison to unplanned replacements[vi].  In addition, it gave customers the ability to view their consumption in real time and adjust their demand to optimize their usage when rates were cheaper[vii].  In 2016, NextEra Energy Services partnered with Autogrid to offer the ControlComm platform. The platform uses machine learning algorithms to forecast demand and allows customers to opt into allowing automatic adjustment of their consumption rates to level demand[viii].   Adjusting demand and raising reliability are not the whole picture when it comes to optimizing the grid.  Demand will never equal full electrical production capacity and investments will need to be made to increase power supply capability.

One of the challenges facing renewables is the inconsistency of their ability to supply power to the grid, due to changes in weather such as wind or cloud coverage and the lack of storage capability when weather conditions produce a surplus to demand.  Currently the difference between renewable energy supply and consumption requirements is made up by conventional sources such as natural gas.  Being able to predict patterns in weather to know when to turn on or off these reserve plants will allow NextEra to close the gap between reserve power production and demand.  NextEra should consider expanding their energy demand predictive analytics to include data from “satellites, climate models, and weather stations[ix]”to forecast energy supply from solar and wind power.  In addition, these predictions can be utilized to understand what areas may benefit from battery storage and utilize machine learning to limit cycling of the batteries to extend their useful lifetime[x].

Going forward, the question is what will have the biggest impact on meeting energy demand through renewable energy: efforts focused on utilizing predictive analytics to modify customers’ usage behaviors or efforts focused on improving energy storage options to store excess renewable energy?

[i] U.S. Energy Information Administration, “Annual Energy Outlook 2018 with projections to 2050,”, accessed November 2018

[ii] NextEra Energy, Inc, “Renewable Energy,”, accessed November 2018

[iii]Florida Power and Light, “Company Pofile”, accessed Novemeber 2018

[iv] NextEra Energy, Inc, Energy Now “Predicting the future of generation”, January 31, 2017

[v] NextEra Energy, Inc,  “2018 EEI ESG Sustainability Reporting Template vF.pdf”, accessed November 2018

[vi] LaMargo Sweezer-Fischer, “Big Data: Using Smart Grid to Improve Operations and Reliability”  July 27-31 2014, IEEE Power & Energy Society General Meeting; Presentation, IEEE-PES website, accessed November 2018

[vii] [vii] NextEra Energy, Inc, “Energy Efficiency”, accessed November 2018

[viii] “NextEra Energy Services Teams Up with AutoGrid to offer New Demand Response Programs in PJM” press release, June 21, 2016, on Business Wire website, , accessed November 2018.

[ix]  Frost & Sullivan, Global Energy & Environment Research Team “Impact of Artificial Intelligence (AI) on Energy and Utilities, 2018” ME1F-14, September 2018

[x] Yongli Wang, Yujing Huang* , Yudong Wang, Ming Zeng, Fang Li, Yunlu Wang, Yuangyuan Zhang, “Energy management of smart micro-grid with response loads and distributed generation considering demand response,” Journal of Cleaner Production, nos. 197, (2008): 1069-1083, Elsevier via Science Direct, accessed November 2018


Open Innovation at Lego – The Back Beat in “Everything is Awesome”


Moley the Robotic Chef: The Future of Cooking or An Expensive Toy?

Student comments on The future of energy: forecasting the weather?

  1. Thanks for the contribution, Lori. I’m curious about your point that increasing the accuracy of weather predicition will enable power providers such as NextEra to better close the gap between reserve power and demand. My previous understanding was that power production would be dynamically adjusted as grid demand rose and fell (e.g. gas plants being turned on during periods of peak demand). What additional options are opened up if you have greater foreknowledge as to when those spikes will occur? Does it change the power generation source you use (e.g. using a plant that takes longer to spin up rather than gas), or is it some other mechanism that provides a benefit?

  2. Enjoyed the post, Lori! I think the question about consumer behavior depends on price sensitivity and a bit on psychology. A major driver of electricity consumption in Florida must be air conditioning, so it may not be enough to say, “Prices are going up. Turn down your air conditioning” because people are pretty set in their ways. The app may need to integrate with other smart home devices to say, “If you turn down your AC by 2 degrees, you’ll save $x on today’s bill.” Some people still may not be compelled, so they may need to consider a steeper rate difference to see a major effect. I know Opower (now part of Oracle) tried to tackle consumer demand reduction using similar tactics, with mixed success.

    On the weather side, I think that global warming will necessitate machine learning advancements in weather prediction for a number of industries, including renewable energy. If they succeed in building accurate models, they would have a broad market outside of their own core business.

  3. Very interesting Lori! Thanks! My first thought while reading your article was about all the different stakeholders that would benefit from the mission of optimizing energy consumption using renewable energy and machine learning algorithms focused on maintenance and weather predictions. It seems like there could be so many great partnerships within the private and the public sector! Hence, an additional key question I would try to think about in the short and medium term is: which organizations should we aim to partner with?

  4. Lori, this was such a cool read! The implications of weather on the production and accessibility of renewable energy is one that I found fascinating and will continue to educate myself on. To answer the questions you posed at the end of your article, I believe that we have an opportunity to do both. We can reduce energy consumption patterns of households across the United States and globally, while also expanding energy storage capacity. I find it remarkable that despite the increase in the number of appliances, devices, electronics in the average American household, energy consumption patterns per household have stayed relatively flat since the 1970s. This indicates to me that the push following the Oil Crises of the late 1970s which spurred a greater awareness on energy efficiency was in fact successful. The challenge moving forward in my mind is how can we leverage machine learning technologies like the one provided by NextEra to bend the consumption curve downwards, while also changing our energy mix away from carbon-sources and towards renewables— a shift that will need the incorporation of energy storage.

    1. Thanks Lori, I enjoyed reading it! I agree with Toni that we need to work on the both initiatives, as renewable / ESS are still on the early commercialization stage. Especially when it comes to the commercial / industrial sector, the back-up energy in ESS is critical as they cannot operate if the energy supply is blocked. Also I’m curious if the commercial / industrial sectors value the NextEra’s solution – would it be even possible for them to use more during the lower-rate time and less during the higher-rate time?

  5. Thanks for the interesting read, Lori! I’m heavily in the camp of focusing on improving today’s storage options over changing customer behavior. When I think about which specific behavior the consumer needs to change, I am coming up with a host of them (e.g. turn off the AC, run the washing machine in the early morning) and to add fossil fuel to the fire (lol) there is more than one protagonist per household that needs to change their ways. Currently, the price of energy is not even close to high enough for consumers to feel the pain of their poor consumption choices.

  6. This post is really cool, thanks Lori! Nice to finally see something other than the age-old buzzword of “storage” with regards to renewable energy innovation. I definitely agree that this will be critical to optimizing smart grid infrastructure – I’m curious about the conclusions NextEra ss drawing from it and whether it will cause them to shift their future capital allocation in terms of region or power source (wind v solar?). I also wonder how aware the fossil fuel industry is of this competitive threat.

Leave a comment