Why Renewable Energy Needs AI
They're more eco-friendly and now increasingly attractive to energy investors. But renewables still need a boost. Luke Smith expects AI to step to the challenge.
Renewable energy uptake is growing rapidly, making up almost half of the UK’s energy mix in Q1 2020, with the UK also going two months without coal-fired power this year. Over the past decade renewable energy has increased by 13.7% annually making up 11% of total energy consumption in 2019.
Costs are down, appeal is up
The growth in renewable energy is partially due to rapidly decreasing costs, with the cost of solar power dropping 82% since 2010 and the cost of onshore wind down 39% in that time. Wider usage has also led to cost reductions as the technologies achieve economies of scale.
Costs for renewables are now below coal and gas and are expected to continue to fall. This has made them attractive not only from a purely economic standpoint but equally from the standpoint of investors who are less and less interested in funding new power plants unlikely to be economically viable.
Climate change is driving renewables
The increased uptake of renewables is also partially driven by the growing awareness of the impact of climate change, prompted by events like California’s record fire season with more than 4 million acres burned. The need to mitigate the growth in carbon levels makes renewable energy more attractive and government legislation to prevent further damage to the climate is likely to further drive uptake.
This means that renewables are also increasingly important for the economy. Consider, for example, that the solar industry in the US now employs around 242k people and generates tens of billions of economic value and the wind industry supports 120k jobs.
As the sector continues to make up a larger portion of the economy, a market will emerge seeking solutions to address the industry’s pain points, such as intermittency and the challenge of matching supply and demand. Many of these solutions will rely on machine learning.
Some of the areas I’m interested in are:
- Predictive maintenance
- Energy generation forecasting
- Site identification
Wind turbines move constantly. It’s pretty much the point. Things that move break down and the maintenance requirements of wind turbines are significant. Operations and maintenance is estimated at 20–25% of the total costs over the lifetime of a turbine.
Wind turbines also generate a lot of data, raising the potential to use the data to improve efficiency and drive down maintenance costs. There is a growing market for digital twin software, which builds a digital copy of a physical asset and uses it to predict performance and required maintenance with wind installations likely to be a key market going forward.
I expect increasingly sophisticated machine learning models that understand the future maintenance requirements of an asset and automatically schedule work at the optimum time.
Energy generation forecasting
Unlike the power generated from coal or gas, wind and solar power isn’t constant or controllable. This intermittency will be an increasing problem as their share of the total power mix increases.
In 2019, Google used AI to predict wind output 36 hours in advance, allowing the wind farm to commit to providing energy in advance and increasing the price achieved by 20% versus providing energy without committing in advance.
I anticipate AI will enable energy producers to make more accurate predictions of energy generation based on factors such as local weather, asset performance and position, allowing more efficient matching of supply and demand and increasing the value of the power generated
The final use case I expect to emerge is in identifying the best sites for renewable energy assets. Both wind turbines and solar panels rely on local conditions for power generation, meaning that a complex array of factors such as weather, land prices, grid connectivity and installation cost need to be considered when deciding where to build.
As the rate of installation increases, I expect energy companies will use to specialised software to decide which sites to build on, optimising for a range of factors to find the best locations. Being able to combine multiple data sources and tailor decisions to specific use cases is likely to be a key competency for players in the space.
Working on a startup in this space?
The growth of renewable energy offers up exciting opportunities for companies to improve the world, while also building an attractive business.
If you’re building something in the renewables space, I’d love to hear from you.