Seventy years ago, the iconic AT&T Bell Labs unveiled cells that could transform sunlight into power. What started as a potential replacement for batteries in remote locations has now become a global phenomenon. Today, solar panels cover an area the size of Jamaica and provide approximately 6% of the world's electricity.
Six percent may not sound like a lot, but we’ve gotten to this 6% terrifically rapidly thanks to the exponential growth of solar capacity. Words like “exponential” get thrown around pretty indiscriminately in tech these days, but in this case, it really is an accurate term to use: Every three years, the installed solar capacity on our planet doubles, resulting in a tenfold increase each decade. To put this into perspective, in just ten years, solar power has grown from a tenth of its current size (providing about half a percent of the world’s electricity) to where it is today (providing 6%). If this trend were to continue, a tenfold increase over the coming decade would mean providing 60% of the world’s electricity needs with solar power by 2034. For comparison, that’d be equivalent to multiplying the world's entire fleet of nuclear reactors by eight, in less time than it typically takes to build a single fission reactor.
So, by the mid-2030s, solar cells are likely to become the single biggest source of electrical power on our planet. If these exponential projections continue further, by the 2040s solar could be the largest source of all energy, not just electricity. Even more impressive, current trends suggest that the all-in cost of solar-produced electricity could drop to less than half of today's cheapest options.
Now, you might be wondering: Is this too good to be true? Can solar’s exponential growth really continue like it has over the past decade? The economics suggest “yes”. Here’s why:
As production of solar cells increases, costs decrease.
Lower costs drive up demand, which in turn increases production, further reducing costs.
It's a virtuous cycle that shows no signs of slowing down. This virtuous cycle is integral to solar’s continued ongoing exponential growth and its great promise. And, critically, unlike previous energy transitions — from wood to coal, coal to oil, or oil to gas — solar power faces no significant resource constraints. While those historical energy sources — coal, oil, gas — became more scarce the more popular they became, thereby requiring us to use increasingly expensive methods to extract the fuel as its popularity increased, no such negative-feedback loop is expected to occur with the current solar-energy transition. The main ingredients for solar’s growth are silicon-rich sand, sunny places, and human ingenuity — all of which are abundant. Even the energy required to produce solar cells is becoming increasingly available through solar power itself.
Naturally, challenges do remain; many of which can be at least partially addressed by data science and machine learning. For example, solar power needs to be complemented with storage solutions and other technologies to meet our 24/7 energy demands (check out Episode #461 with Sam Hinton on energy-grid management for more on the data-science opportunities in this space).
Electrifying heavy industry, aviation, and freight also presents obstacles. However, advancements in battery technology and electrolysis-created fuels are gradually addressing these issues. And you can leverage data science to help on these fronts. For example, in terms of battery technology advancements:
AI can accelerate materials discovery by rapidly screening and predicting properties of new materials for better electrodes and electrolytes.
Machine learning models can simulate and optimize battery architectures for specific use cases.
AI can enhance battery management systems for better performance and longevity.
And in terms of synthetic fuels produced using electricity (compounds are known by names such as e-methane as a synthetic replacement for natural gas or e-kerosene for powering airplane turbines):
AI can optimize electrolysis processes for higher efficiency and lower costs of synthetic fuel production.
Machine learning can help identify more effective catalysts for electrolysis.
Data science can predict demand for electrolysis-created fuels, aiding in production planning.
Other opportunities for leveraging data science to support solar energy production include:
Using AI to accelerate the discovery and development of new photovoltaic materials.
Generative AI can predict where solar projects will be successful (listen to Episode #783 with Navdeep Martin for more on that).
AI can optimize solar-panel production processes, reducing defects and improving yield.
And, if you’re looking to make an impact with AI on climate change more generally, three more ways examined in recent episodes of this show include:
Episode #789 with Dr. Jason Yosinski for a number of ways that AI can be used to make wind farms and their associated power grids markedly more efficient.
Episode #773 with Prof. Barrett Thomas for how deep reinforcement learning can calculate the more energy-efficient freight-transport routes and even leverage autonomous vehicles and drones to save energy usage.
Episode #459 with Vince Petaccio for a broad range of ideas for using machine learning to tackle climate change.
Hopefully something amongst all these opportunities is inspiring to you! The implications of this energy transition, particularly the currently exponentially-growing solar revolution, are profound. Cheaper energy will boost productivity across all sectors. It will make essential services like lighting and transportation more accessible to billions of people. We could see breakthroughs in water purification, desalination, and artificial intelligence powered by this abundant energy. And, perhaps most exciting, are the innovations we can't even imagine yet — perhaps even your mind has been unlocked to dream about what new innovations you could dream up and deliver in this new era of energy abundance.
The SuperDataScience podcast is available on all major podcasting platforms, YouTube, and at SuperDataScience.com.