AI Did That?
Once-futuristic Artificial Intelligence now serves as an omnipresent facilitator across industries and markets. In a series of articles, EastBanc Technologies zero in on events influenced by AI, worldwide and local problems solved by AI, industries revolutionized by AI and scientific milestones reached with the help of AI.
We aim to showcase AI as a force for good: Advanced and rapidly improving technology that can solve unsolvable problems, facilitate difficult and time-consuming tasks and further open the doors to trailblazing innovation.
With the effects of climate change hitting the world with a vengeance, one of humanity’s most urgent challenges is that of reducing carbon emissions and transitioning to green energy.
Unfortunately, two of the main sources of renewable energy – wind and sun – are hampered by the obvious: The wind doesn’t always blow, and the sun doesn’t always shine. And when the wind does blow and the sun does shine, they don’t always do so with the same intensity. Indeed, one of weather’s key attributes is its high variability. This implies potential volatility for the energy market, much more so than with fossil fuel power stations.
Fortunately, technology can help. Specifically: artificial intelligence (AI) and machine learning (ML).
How can these technologies help us transition to – and maximize the output and use of – green energy sources? We addressed solar energy in an earlier article. In this piece, we will focus on wind power.
Wind is known to be a very unpredictable energy source. However, with the integration of AI-based tools and ML it is now possible to obtain better forecasting and thus operate and manage wind farms more efficiently, and also avoid grid congestion. With the help of sensors and ML algorithms – among which the most common support vector machines and gradient boosting decision trees – wind farm operators can better predict upcoming power production based on present and past weather data. These short-term power generation forecasts are then fed into operating systems that schedule charging and discharge which helps reduce curtailment. This is crucial because energy sources like fossil fuels that can provide a set amount of electricity at a specific time often have more value to the grid and in the market.
DeepMind, Google’s London lab, applied its Machine Learning algorithms and neural network trained on widely available weather forecasts and historical turbine data to some of Google’s wind farms in the American Midwest. The system could predict wind power output 36 hours prior to actual generation and used these predictions to provide operators with precise recommendations on how to deliver optimal hourly input to the grid at least 24 hours in advance. This model has enabled wind energy’s market value and cost-effectiveness to be boosted up to around 20 per cent.
To better predict the energy output for more profitability on the market is one great step towards ensuring wind reliability. But power generation is not the only thing AI can predict. How about demand? Indeed, forecasting power demand is a very tricky and difficult process that can easily lead to curtailments or blackouts when not properly executed. AI is a perfect tool for detecting intricate patterns based on historic consumption data, and thus predicting demand, both at the individual and cumulative level.
Another compelling use case of AI in wind power generation has to do with one of the first and crucial steps in developing and implementing wind farms: site assessment. One of the leading European energy research institutes, the non-profit ZSW foundation, has been aiming to leverage AI and ML to develop innovative MCP (measure-correlate-predict) methods and has applied them to various processes, including site assessment. Before creating a new wind farm, it is necessary to perform a site assessment, taking into consideration a number of parameters, such as orography and weather conditions. This is extremely challenging, especially in complex terrains. A precise estimate usually requires numerous types of measurements over the course of long periods of time. The foundation’s study has demonstrated that the AI-based MCP algorithms led to a significant reduction of yield data errors, in comparison to traditional MCP methods. These AI methods not only provide solid site assessments and clear indications of how to implement the wind farms in a given location, but also reliable insight on required equipment and design criteria.
This is just a glimpse of a few of the many ways in which AI can help organizations generate and use power from clean energy sources, and thus help the world steer towards a greener direction. Renewable energy resources are all part of nature’s unpredictable behavior. Hence, they are all interconnected. Just as all components of the energy grids are. So, when AI turns out to be the answer to one challenge in one single renewable energy development, it ultimately becomes a solution to the global renewable energy transition challenge as a whole, whether it be for risk mitigation, cost reduction, product design, asset maintenance, grid management, power production, storage, supply, etc.