Artificial intelligence (AI) is having a significant impact on the renewable energy market, mainly to make accurate predictions and to orchestrate the different parts of the energy system. An overview of the developments. And a call. In the near future, NedZero will share knowledge about the possibilities of using AI within the wind industry. To gain insight into the experiences and specific needs of our members, we would like to receive information.
AI, versus machine learning and deep learning
Artificial intelligence (AI) refers to the simulation of human intelligence in machines capable of performing tasks that normally require human intelligence. This includes skills such as speech recognition, image recognition, natural language processing and problem solving. An AI system is like a digital brain, designed to learn from experience and improve itself.
Machine learning is a subset of AI that focuses on the ability of computer systems to learn and improve without being explicitly programmed. In other words, in Machine Learning we give computer systems data and let them discover patterns and make predictions based on that data.
Deep learning is another subset of Machine Learning and goes one step further. It uses artificial neural networks, inspired by the way our brains work. These neural networks consist of multiple layers of neurons that process information. Deep Learning is able to discover and learn from complex structures in large amounts of unstructured data, such as speech and image recognition at an advanced level.
The market
According to Reuters , the use of AI represents a value of $13 billion in the energy sector alone and contributes to more efficient use of wind energy, better maintenance and increased predictability.
A data-driven global survey last year among 745 startups & scaleups in wind energy shows that AI emerged as number 1 of the top 10 trends. These are the developments and applications that underlie it.
Google, AI and wind energy
Pioneering in the development of AI within the wind industry has been the use of machine learning to increase the predictability of wind energy production.
Major consumer of wind energy Google, together with DeepMind started a joint investigation in 2018. The aim was to better map the predictability of wind and therefore make it more reliable as an energy source.
Looking for a solution, they began applying machine learning algorithms to 700 megawatts of wind energy capacity in the central United States. The wind farms in question, which are part of Google's global fleet of renewable energy projects, collectively generate as much electricity as a medium-sized city needs.
Using a so-called neural network trained with publicly available weather forecasts and historical turbine data, the DeepMind system could then be configured to predict wind energy output 36 hours ahead of actual generation. This also made it possible to develop algorithms that optimize the energy production of wind turbines through advanced prediction models. Based on these predictions, the model was then able to deliver the optimal delivery commitments per hour - a full day in advance.
Important, because energy sources that can be scheduled (i.e. can deliver a certain amount of electricity at a certain time) are often more valuable to the grid. This contributes to more efficient management of wind farms and increases the value of wind energy on the electricity grid.
So far, machine learning has increased the value of wind energy by about 20 percent compared to the base case without time-bound grid commitments. While wind variability cannot be eliminated, this early success suggested that machine learning can make wind energy more predictable and valuable.
Trading systems
The principle of machine learning has now led to the development of fully-fledged trading systems. Norwegian renewable energy producer Statkraft , uses AI in its energy trading activities. They work with forecasts, some of which are supported by machine learning, and algorithms, the so-called trading bots that communicate with the market in an automated way. The electricity generation assets the company manages are programmed to automatically start and stop producing power according to schedules, designed using AI.
The entire system is highly automated and designed to be scalable and operate in a cost-effective manner across a large fleet of assets. That wouldn't be possible without the use of machine learning and algorithms that connect everything together. (Read more here .)
Maintenance and cost savings
AI technologies such as predictive maintenance help identify potential failures before they occur, leading to a reduction in maintenance costs and downtime. This can result in significant savings for wind farm operators. For example, GE Renewable Energy has indicated that its AI-powered maintenance systems can lead to a 10-15% reduction in operating costs.
Siemens Gamesa uses AI for predictive maintenance and operational optimization. They have equipped their systems with sensors that collect and analyze real-time data to identify potential problems early and plan maintenance before failures occur. Their systems combine knowledge of turbine design and technology, historical data and machine learning to predict component behavior and efficiently plan maintenance. (See: Remote Turbine Diagnostic Services and The Power of Big Data )
Vestas has a strong focus on data analytics and smart technology to improve operations. This includes predictive maintenance and operational optimization, using data from hundreds of turbines to identify patterns and anomalies that could indicate future problems.
There are now a multitude of AI techniques that are widely used for surveillance systems and to improve efficiency and maintenance management.
A comprehensive overview of artificial intelligence and wind energy is given in the article: A Comprehensive Review of Artificial Intelligence and Wind Energy | Archives of Computational Methods in Engineering (springer.com) This includes: Artificial Neural Networks, Fuzzy Logic, Genetic Algorithms, Particle Swarm Optimization, Decision Making Techniques, and Statistical Methods.
These techniques help address various aspects such as economic factors, location of the wind farms, non-destructive testing, environmental conditions, schedules, operator decisions, energy production and remaining life.
Performance optimization
Wind turbines are equipped with thousands of sensors that generate enormous amounts of data. AI systems analyze this data to optimize turbine performance and support operational decisions. This leads to better utilization of resources and increased reliability of energy production.
For example, AI helps by improving the placement of turbines and the management of energy flow to the grid. This can increase energy yield by 5-7%, depending on the location and specific conditions of the wind farm.
IBM's Watson is used, among other things, to analyze weather forecasts and optimize energy production. This uses data from hundreds of turbines to identify patterns and deviations that could indicate future problems. ( IBM Watson AI in Energy )
Challenges
While AI brings many opportunities and benefits, it is not without challenges. Wind energy asset management software company Skyspecs mentions the lack of centralization and structured data in the industry as the biggest challenge in its blog 'Artificial Intelligence and its role in Wind Energy' . Larger organizations often work with data silos that are spread throughout the organization, but there is not a single source of truth. The data is not easily accessible and it can feel like finding a needle in a haystack.
Perhaps we can learn from initiatives for collecting big data within the oil and gas sector. Consider the Open AI Energy Initiative by Shell, Baker Hughes, C3 AI and Microsoft: BHC3 | Baker Hughes . Specialized oil and AI modules can be shared via this open source platform for AI solutions for the oil and gas industry. The idea is that oil and gas companies, service providers, equipment suppliers and independent software providers can all offer AI and physics-based models and related services, such as monitoring, diagnostics, prescriptive actions and services. The data obtained can be used to improve AI models by, for example, predicting risks for process and equipment performance.
Another challenge, according to Skyspecs, is the lack of standardization and consistency in the industry. This is reflected in the way damage is marked and categorized and how severity is assessed. There are many different languages that customers use when talking about these categories and severity levels.
Chances
Overall, there are many opportunities for the use of AI in the wind sector, Skyspecs also sees. As data becomes more structured and consistent, domain experts can work more closely with engineers and we can solve more targeted problems with machine learning.
Cautious predictions about developments point towards multimodal learning, where AI could help. This means looking at all the different data modalities together so that a single model can take something like an image and then provide an output. This also offers opportunities for the wind industry, for example in the way analysts assess damage. This isn't just by looking at an image; there are several parameters that need to be assessed such as radial distances before making a decision. That's where AI could come in handy.
Call: what are your experiences and wishes?
The applications of AI will only increase in the near future. It is therefore very important to be and remain informed of the possibilities. NedZero wants to focus more explicitly on sharing knowledge about AI for our members in the near future.
To better understand where the precise need lies, we have the following questions. We would like to hear your insights via: info@nedzero.nl, stating the subject: 'AI opportunities for the wind industry'. We will anonymize your information upon request.
- What projects related to AI and wind energy are you currently working on?
- Which issues would you like to have solved with the help of AI in the near future?
- What insights and suggestions regarding AI and the wind industry would you like to share with us?
