4 minute read

Transitions in Green Energy Driven by AI

Paula Madetzke

Powering the world in a sustainable way will be an increasingly important endeavor astime goes on, and green energy will need to be a core component of this transition. However,relying on fluctuating sources such as wind and solar, in addition to our current lack ofubiquitous battery storage of the energy results in some structural difficulty. However, theseissues largely boil down to a matter of predicting the supply and demand of energy. This task ofpattern recognition makes the problem an opportune one for artificial intelligence to solve, andmany strides have already been taken in the field.

In order to predict the supply of wind and solar, one must first predict the weather. Dueto the notorious difficulty of this task with human made models, AI which can create the modelfrom observations of data is well suited to the tasks. Currently, Nnergix and Xcel are bothprojects implemented to aid weather forecasting, They harness weather data that is alreadybeing recorded at weather stations, wind farms, and local satellites to train and feed to the AI.The clearer weather predictions enable managers of power grids to make better decisions aboutwhen natural energy will be abundant, vs when energy will be scarce.

The knowledge of when energy is abundant makes it possible to schedule non-urgentbut energy intensive tasks to occur when energy is the most abundant, and presumablycheaper. AI could be employed to use the predictions of the weather forecasting AI in addition topredictions about power usage optimize the scheduling of tasks such as running a washingmachine or even energy intensive commercial tasks at a larger scale. With the double task offorecasting weather patterns and their effects on energy supply in addition to predictingconsumer demand over time, it is difficult to imagine this kind of energy adaptation beingpossible without AI.

Another important task of AI in green energy has less to do with prediction supply anddemand, and more to do with simply reducing demand with increased power efficiency. Oneexample of this is Google’s DeepMind, which was able to significantly reduce the energy costsof cooling large data servers by a whopping 40%. This was achieved by training the AI with datafrom sensors that were attached to the servers over the course of two years. As countriesdevelop and the demand for industry increases globally, being more efficient with energy usewill become increasingly important. Not only will this increased efficiency increase the supply ofpower in the short term, it will help slow the growth of the power demands as it becomes morefreely available to use.

Another useful application of AI for green energy is the problem of storage. One of themain roadblocks to AI becoming more prominent is the fact that we do not currently have thesorts of massive batteries needed to store the total amount of energy required to get throughperiods of scarce energy in centralized locations like power plants. However, as more electriccars become part of the power grid, there are more opportunities to form “micro-grids.” Thiswould involve a grid that is able to tap into edge devices such as parked electric car batterieswhen power from solar or wind is scarce and transfer the power to the devices that require thepower more. This would reduce the pressure on utility companies to have and maintain suchlarge battery stores. However, it introduces a new level of grid complexity that could onlyrealistically be managed properly by AI.

Before having a closer look, I was not aware how prominent or promising AI was in thefield of green energy. Previously, I had only heard about how the complexity of the variability inweather and the problem of battery storage would be what prevents green energy frombecoming the standard for power. AI not only seems promising for the future of green energy,but it appears to be a significant portion of the solution to the roadblocks that have long plaguedthe field.

Resources

[^3] Online Resource https://www.weforum.org/agenda/2018/05/how-ai-can-help-meet-global-energy-demand