Artificial intelligence or the key to integrating wind turbines into the network? COMMENTS
By Reda Tahiri, energy and mobility consultant at Wavestone
The wind is blowing hard on French wind turbines. By wanting to “accelerate the ecological transition” and advance in the “massive development of renewable energies”, the French government emphasizes the importance of wind energy. On August 28, 2021, Prime Minister Jean Castex made a statement from the Saint-Nazaire shipyard (Loire-Atlantique), off which the very first French offshore wind farm (non-pilot) will be commissioned at the end of 2022. Quite a symbol. (1)
Wind power a French development
The government will launch seven new calls for tenders to boost the development of offshore (maritime) wind farms in France. The Prime Minister announces that over the next five years, more than 25 billion euros of investment will be made in renewable energies, making it possible to support the creation of more than 25 gigawatts of solar, onshore wind and maritime production capacities , and hydroelectric.
In 2020, more than 19% of the electricity produced in France was of renewable origin. Hydropower came first (13.5% of electricity consumption in 2020), followed by French wind farms (8.8%) and solar (2.8%). Despite the preponderant role of hydropower in electricity production, growth was driven by the development of wind and solar power (94% of the power added in renewable energy in 2020). (2)
Wind power remains controversial and in the context of elections, political decision-makers have had to position themselves on the installation of new capacities. To achieve the objective of 40% of electricity produced from renewable sources (or 32% of energy) by 2030 set by the Ministry of Ecology, the installation of new solar, land-based and maritime wind capacities is urgent.
"More than onshore wind power, it is the development of offshore wind farms which must constitute, after solar, the second priority axis" of the French strategy in renewable energy, says the Prime Minister.
It confirms the government's wish to launch a new wind farm in Normandy, alongside the existing project located off Barfleur. This strategy of expanding existing parks or those under construction is inspired by profitable strategies in the Netherlands and Belgium.
This reduces installation and operating costs, while making lots competitive. (3)
French offshore wind projects, according to the French government (4)
Several factors (and not the least important ones) play in favor of offshore wind farms. On the one hand, their cost should decrease more rapidly than that of onshore wind farms (Wiser et al., 2016 (5)). On the other hand, thanks to technological improvements, they are becoming more efficient: the capacity factors of offshore wind turbines already reach 40 to 50% (International Energy Agency 2019 (6)).
The most competitive energy source per kWh generated per county in the United States in 2019 (7)
Multiple challenges of integrating renewable energy into the existing grid
However, one should keep in mind the many challenges faced by operators of renewable parks.
First of all, the very essence of wind and solar sources is to be fluctuating. Because directly dependent on weather conditions that are uncontrollable. However, the reliability of an energy network is closely linked to its stability (8 Why is the flexibility of electricity networks becoming a major issue?).
Supported by competitive prices, the development of wind power still faces these challenges of predictability and management. The main issues are related to the integration of large wind farms into the electricity grid.
This problem arises more in countries, such as the United States, where the distances between production and consumption areas are significant. Integration between electricity networks is essential to allow, when the wind no longer blows off the Danish coast, for example, to supply Germany with electricity from French nuclear power (9).
Artificial intelligence in wind power
To improve its performance, renewable electricity producers are encouraging cooperation between energy companies and technology companies.
Indeed, in September 2021, Google and ENGIE signed an agreement to supply renewable energy to all Google activities in Germany, Belgium and the Netherlands (10 GAFAM and renewable energies)
The development of Big Data and more particularly the Internet of Things allows energy companies to regulate their production or plan their maintenance activities. As an example, we can cite the agreement between Microsoft and ENGIE for the use of Azure servers dedicated to the Internet of Things and the processing of their data (11).
I miss when my mommy would make boiled peanuts*looks up how to boil peanuts using roasted peanuts* 🥜
— Fallen Angel Thu Jun 17 23:48:26 +0000 2021
Massive use of data by energy companies fuels another cross-cutting technology: artificial intelligence. Indeed, the impacts of AI for the wind industry are multiple:
The maintenance of a wind turbine blade causes the shutdown of a wind turbine and can last several days, and weeks for a wind farm. It is therefore expensive for wind farm operators.
A drone inspection of the outside, but also the inside, of the blades of a wind turbine makes it possible to visualize and detect, thanks to image recognition, any defect. The processing of this data by combining it with the performance of the blade makes it possible to target preventive maintenance actions.
This development allows the wind farm operator a shorter and safer production shutdown for his teams (i.e., working in a confined space and at height in the blade). It also increases the life of the blades.
Dashboard, proposed by Nanonets, of defects in a wind turbine following an inspection (12)
Drone inspection and image recognition, offered by Funke-Gruppe (13) (14)
An AI model, developed by EPRI (16), has notably made it possible to achieve 80% accuracy in the prediction of breakage, and thus to reduce the costs of maintenance by 20%.
Gearbox or gearbox of a wind turbine
AI and control of electricity production
The electricity market encourages wind farms to provide precise forecasts in order to guarantee the stability of the network. Lower-higher-than-production forecasts cause strain on the network. As a result, grid operators impose fines on wind farms when their forecasts are inaccurate.
Indeed, a survey of several network operators around the world revealed that 94% of specialists surveyed consider that "the integration of a significant amount of wind energy will ultimately depend on the accuracy of the wind forecast” (Jones et al., 2011 (17)). While typical forecast errors for energy delivery are between 1% and 3%, wind energy forecast errors are between 15% and 20%.
Some compensatory solutions have been developed thanks to European cooperation. In particular through the intra-day market which makes it possible to exchange very short-term electricity (up to 30 minutes) from renewable energies between several countries of the European Union (18). In short, an Italian electricity distributor can easily buy Dutch wind energy. Therefore, the more precise the delivery commitment on the spot market, the better the business model of wind farms performs. (19-20 BREXIT: WHAT IMPACTS ON THE EUROPEAN ELECTRICITY MARKET?).
Thus, the role of wind farm operators is to forecast their production as well as possible to facilitate the management of the balance of the network. To date, there are several wind forecasting methods that can be classified into three main categories according to the type of data they use.
The physical models are based on the results generated by the numerical weather prediction (NWP) models. In short, it is a question of reducing the scale of national weather forecasts to the precise positioning of the wind farm. Depending on their resolution, they can be global or regional models.
A physical model, according to the World Meteorological Organization (21)
Statistical models use historical information (typically wind speed and direction, air temperature and humidity) to infer temporal patterns that help calibrate other predictions (Kavasseri and Seetharaman ( 22)). These predictions are based on a compromise between generality (diversity) and performance (convergence).
Statistical strategies are also experiencing a resurgence in popularity since machine learning strategies have been (re-)put in the spotlight (Karinotakis et al., 2017 (23)).
In addition, there are hybrid models that exploit both physical and statistical perspectives, using outputs from physical models as inputs for statistical techniques to generate reliable statistical downscaling.
Benefits of prediction based on machine learning, according to DeepMind (24)
Various techniques specially designed to process large amounts of data have been developed in recent years, and have been applied with dazzling success in a wide range of fields.
In the case of short-term prediction of wind or wind power, an ever-increasing number of contributions use hybrid methodologies (Okumus et al., 2016 (25)). Moreover, as pointed out by Liu et al. in 2015 in their study (26), hybrid forecasting models offer better performance than purely statistical or physical models in predicting wind speed.
This is how DeepMind, a subsidiary of Google, managed to increase the economic value of wind farms by 20% (24). These wind farms, totaling 700 MW of power, therefore share 20% more reliable production forecasts with network operators.
In short, the development of renewable energies, including onshore and offshore wind power, will allow a gradual decarbonization of the global energy mix.
This development will also lead to significant fluctuations in energy production. Especially since the predictability and management of energy production is essential for the functioning of an electrified society. Advances in machine learning and artificial intelligence can partially address this issue.
References:
- Les Echos – The State pleads for a second offshore wind project off the coast of Normandy – last consultation on 25.10.2021https://www.lesechos.fr/industrie-services/energie-environnement/ the state-pleads-for-a-second-offshore-wind-power-project-au-large-de-la-normandie-1341605
- RTE – The panorama of renewable electricity – 06.10.2021 https:/ /www.rte-france.com/analyses-tendances-et-prospectives/le-panorama-de-lelectricite-renouvelable
- La Tribune – Offshore wind power: what is pushing the government to want two wind farms in place of one – last consultation on 25.10.2021 https://www.latribune.fr/entreprises-finance/transitions-ecologiques/eolien-marin-ce-qui-pousse-le-gouvernement-a-vouloir-deux- parc-au-lieu-d-un-seul-891681.html
- Wind turbines at sea – last consulted on 25.10.2021 https://www.eoliennesenmer.fr/
- Wiser et al. 2016 Expert elicitation survey on future wind energy costs | Nature Energy
- IEA – Global CO2 emissions in 2019 – 11.02.2020 https://www.iea.org/articles/global-co2-emissions-in-2019
- Wall Street Journal – Building the Wind Turbines Was Easy. The Hard Part Was Plugging Them In – last accessed 25.10.2021
- EnergyStream – Power Grid Flexibility – last accessed 25.10.2021 https://www.energystream-wavestone.com/2021/06/ flexibilite-des-reseaux-electriques/
- Power Technology – The Great State of Texas: explaining the power crisis and what happens next – last consulted on 25.10.2021 https://www.power-technology.com /features/the-great-state-of-texas-explaining-the-power-crisis-and-what-happens-next/)
- EnergyStream – GAFAM and renewable energies – last accessed 25.10.2021 https://www.energystream-wavestone.com/2021/02/gafam-et-energies-renouvelables/)
- Microsoft – ENGIE brings AI to renewable energy platform, helps decarbonize energy production with Azure – latest consulted on 25.10.2021 https://customers.microsoft.com/en-gb/story/808631-engie-energy-azure-iot
- Nanonets – Fault Detection in Wind Turbines – last consulted on 25.10. 2021https://nanonets.com/drone/wind-turbines/)
- Funke Gruppe – How drones and AI are revolutionizing the inspection of wind turbines – last accessed 25.10.2021
- Bloomber – How Wind Turbine Blades Are Inspected By Drones – last accessed 25.10.2021
- T&D World – EPRI Develops AI Model to Reduce Wind Turbine Operations Costs – last accessed 25.10.2021https://www .tdworld.com/renewables/article/21174662/epri-develops-ai-model-to-reduce-wind-turbine-operations-costs-utilities-see-significant-benefits
- T&D World – EPRI Develops AI Model to Reduce Wind Turbine Operations Costs – last accessed 25.10.2021 https://www.tdworld.com/renewables/article/21174662/epri-develops-ai-model-to-reduce-wind-turbine-operations -costs-utilities-see-significant-benefits
- OSTI.GOV – Strategies and Decision Support Systems for Integrating Variable Energy Resources in Control Centers for Reliable Grid Operations (Technical Report) – last accessed 25.10.2021< /li>
- Directorate General for Energy (European Comission) The future electricity intraday market design – 22.03.2019https://op.europa.eu/en/publication-detail/-/publication/f85fbc70-4f81-11e9-a8ed -01aa75ed71a1/language-en
- EPEX spot – Basics of the power market – last accessed 25.10.2021 https://www.epexspot.com/en/basicspowermarket)
- EnergyStream – Brexit what impacts on the European electricity market – last consulted on 25.10.2021 https://www.energystream-wavestone.com/2019/05/brexit-quels-impacts-sur-le-marche-de-lelectricite -europeen/
- Rucong Yu, Jian Li and Pengqun Jia – Three characteristics of numerical models determine the development of operational weather forecasting – 2019https://public.wmo.int/en/resources/bulletin/three -characteristics-of-the-mod%C3%A8the-numerics-of%C3%A9end-the-development-of-the
- https://www .sciencedirect.com/science/article/pii/S0960148108003327
- Kariniotakis et al. Wind power forecasting using advanced neural network models | IEEE Journals & Magazine – 12.1996
- DeepMind – Machine learning can boost the value of wind energy – last accessed 25.10.2021
- https://linkinghub.elsevier.com/retrieve/pii/S0196890416305428
- https://linkinghub.elsevier.com/retrieve/pii/S0960148115300276