Artificial Intelligence (AI) techniques will amplify the existing data-driven transmission network activities: AI is able to extract value from more data gathered by sensing systems deployed along the network facilities. AI-based applications to transmission networks may cover network planning, network operations, asset monitoring and maintenance.

The International Renewable Energy Agency (IRENA) defines AI as an area of computer science that focuses on creating intelligent machines that follow human behaviour, according to the data collected [1]. This might be considered a coherent starting point for its contribution to the development of data-driven services for the transmission networks. However, this concept can sometimes be misunderstood or considered too broad: some techniques inside statistics and Data Mining (DM) may also be located within the AI area, as schematised in the tree diagram below [2]. The explanation of the diagram elements is given below:

• Machine Learning (ML): ML is a sub-field of AI that evolved from pattern recognition to analyse the data structure and to fit it into models that can be understood and replicated by users [3]. ML is segmented into four categories: Supervised Learning, Unsupervised Learning, Reinforcement Learning (RL) and Deep Learning (DL) and aims to predict or describe the existing relationships within the data set [4]. 
• Reinforcement Learning (RL): RL is a computational approach that learns from the interaction with the environment. This means defining how system agents can take actions in their environment to maximise the cumulative reward. RL is implemented mainly to solve energy dispatch problems and building energy management scheduling [2, 5].  
• Deep Learning (DL): DL belongs to the Artificial Neural Networks (ANN) field. DL techniques can be applied to power systems in different scenarios such as fault detection in transformers [6], or day-ahead (DA) electricity market price forecasting [7].  
• Statistics: statistics is a science dealing with analysis and data modelling. Despite the similarity with ML, Statistics is the field of Mathematics that deals with the understanding and interpretation of data [8]. According to Ref [9], statistics use a population sample to draw population inferences, whereas ML determines generalisable predictive patterns from data. 
• Data Mining (DM): DM aims to extract patterns and knowledge from large amounts of data, not at the mining of data itself. It is the knowledge mining from large data sets [10].   

Application for the electricity transmission system [11]: 

• Power flow optimisation [12]; 
• Power grid loss prediction; 
• Optimisation of DA predictions; 
• Weather forecasts; 
• Renewable energy sources (RES) power generation; 
• Asset management: visual inspection, vegetation management, environmental threats, predictive maintenance; 
• Data quality assessment: real-time detection, auto validation processes, estimation and correction; 
• Large Language Model for technical documentation, assistant for system operation and market activities TSO/DSO cooperation supported by gathering securely energy from DER sources or storage [13];  
• Grid health monitoring: real-time monitoring system of transmission lines [12], transformer and breaker monitoring in a substation [12]; 
• Fault detection based on a deep learning intelligent approach in detection and classification of transmission line faults [14]; 
• Using optimisation code for automated energy trading; and 
• Using DLs in drones for maintaining the purposes of overhead lines (OHL).  

Challenges in reaching the scope:  

• Complexity of data gathering with the increasing number of IT systems storing the data; 
• Ensuring data quality and security; 
• Managing the complexity of AI technologies; 
• Convincing the business of the AI adoption profits and its integration in the IT landscape; 
• Workforce adaptation and skills alignment; and 
• Identifying AI within existing systems as well as integrating AI with existing systems, especially for AI systems used as safety components in the management and operation of digital infrastructure, given the uncertainty around the interpretation of safety components as defined by the European Commission’s AI Act [15].

The increased adoption of AI, cloud computing and other digital technologies position utilities to be more capable of responding and adjusting to increasing market dynamics, stemming from decarbonisation, demand volatility and climate changes.

The AI adoption in TSOs can address the following challenges [16]: 

• Structural transitions of electricity systems to accommodate a dynamically increasing number of diversified and distributed power energy resources; 
• Evolution of electricity markets towards the increased diversification of market actors and services; 
• Facilitation of transition processes through new, digital technologies to analyse and optimise electricity demand patterns on the consumer-side; and 
• Increasing requirements towards system resilience, including, for example, climate change and man-made hazards. 

The identified benefits are the following: 

• More grid flexibility and data-driven decision-making capability for operators; 
• Reduction of the connection time of RES plants to the grid and increased transparency on the authorisation process conducted by institutional bodies; 
• Support for the decisions and work of business users towards the same quality standards but in less time; 
• Performance increase and support of uncertainty treatment; and 
• Faster accessibility to key information and the visualisation of impacts at regional level.

The enablers of AI are listed below: 

• The European Commission’s AI Act [15].
• Improved support and collaboration with institutional bodies in light of the European Commission’s AI Act [15], which will pose challenges in implementing all the measures in cases where AI is used for critical systems.
• Gathering appropriate data, the data sources are split into two large groups, operational and non-operational data. The power system operational data include all the measurement assets that collect power and energy data, including voltage, current, active and reactive power and grid status signals. On the other hand, non-operational data provide essential information that has a crucial role in supporting the energy services performance, such as weather conditions, electricity market data, social media, Geographic Information System (GIS) and known parameters given by customers. 
• Implementing efficient Edge Computation (EC): Edge Intelligence (EI) deployment aims to ensure the co-convergence of AI tools and EC. In one direction, the EI field highlights the optimisation methods that allow AI models to run and make the edge environment more efficient. In the other direction, EI reviews the various methods that help adapt the Edge to AI applications [17, 18].  
• Appropriate architectures to support Cloud Computing (CC) and EC: in Ref [18], three types of architectures are described for practical CC and EC implementations in power transmission networks. 
• Support from companies offering the utilisation of AI in energy sector [19];   
• The AI Pact launched by the European Commission.

There are several needs for R&D: 

• Dedicated software to handle data complexity: the EU-funded project VEDLIOT has developed a Deep Learning IoT platform. Instead of traditional algorithms, AI and DL enables the handling of large complexity of data handling [20]; 
• Standardisation of communication processes: EC–CC systems can be deployed with a unified communication adapter using any category of drivers. However, after the EC node receives all the information, the data fusion methods for sensors in EC among power equipment need to be improved in future EC–CC system for the transmission grid; 
• Cooperation in the TSO community for the seamless and effective implementation of the European AI Act in electricity systems; 
• Analyse procedures, support proponents and institutional bodies in designing plants and selecting the best technical solutions; and 
• The reuse of technology components developed in the first phase of AI adoption to launch new use cases.

The technology is in line with milestones “Integration of dynamic ratings and AI-based renewable power forecasts” and “Demonstration of innovative technologies for power flow control and increasing grid efficiency” under Mission 1, milestones “AI based reporting and analysis of system operation”, “AI based decision support system for system operation” and “Innovative training concepts and backup procedures” under Mission 4 and milestone “AI and ML solutions to boost horizontal and vertical system integration” under Mission 5 of the ENTSO-E RDI Roadmap 2024-2034.

TenneT:  Real-World Day-Ahead Congestion Management using Topological Remedial Actions (GridOptions Tool): 

• A tool developed on sub-net (Groningen-Drenthe) due to the challenges of shortages, the increasing requests of connections and the generation of renewable energy resources, and dynamic flow patterns; 
• Supportive tool, not full automation; and 
• Fully rolled-out solutions for entire day, in a dynamic grid model environment that is constantly changing. 

APG: Bing Chat for employees, chatbot for Law documents: 

• bing.com/ chat has been enabled for all APG employees; 
• The APG employees have been trained in chances and risks of Large Language models; and 
• Terraform has been used to deploy services on the Azure Cloud with Infrastructure as Code (IaC) principles. 

RTE: AI assistants for operators ORIented Grid Analysis by Machine Intelligence (ORIGAMI), chatbot work on technical documentation, news monitoring assistant 

• AI for the French TSO is focused on two areas: (i) Save operators time by automating time-consuming tasks, (ii) Provide decision-making tools to deal with the growing complexity of the electricity network; 
• The approach takes into account ethics, acceptability and sobriety; and 
• Cooperation with the AI community has been accommodated by exhibiting problems in a power network in a game-like manner to test the potential of AI to increase network management. 

ČEPS: Network Model Error Detection, Transmission Loss Prediction

TRL 7-8 for operation applications requiring very low latency.

1. International Renewable Energy Agency, “Artificial intelligence and big data innovation landscape brief”

2. S. Barja-Martinez et al. “Artificial intelligence techniques for enabling Big Data services in distribution networks: A review,” Ren. & Sust. En. Rev, vol. 150, p. 111459, 2021.

3. G. Shobha and S. Rangaswamy, “Chapter 8 - Machine learning,” Handbook of Statistics, Vol. 38, pp. 197–228, 2018.

4. M. S. Ibrahim et al.. “Machine learning driven smart electric power systems: Current trends and new perspectives,” Appl Energy, vol. 272, no. 1, p. 115237, Aug. 2020.

5. A. T. Perera and P. Kamalaruban, “Applications of reinforcement learning in energy systems,” Renew Sustain Energy Rev p. 137110618, 2021.

6. A. Moradzadeh and K. Pourhossein. Short Circuit Location in Transformer Winding Using Deep Learning of Its Frequency Responses. In Proceedings of the 2019 International Aegean Conference on Electrical Machines and Power Electronics, ACEMP 2019 and 2019 International Conference on Optimization of Electrical and Electronic Equipment, OPTIM 2019, Istanbul, Turkey, 27–29 August 2019; IEEE: New York, NY, USA, 2019; pp. 268–273. 

7. A. Brusaferri  et al., “Bayesian deep learning-based method for probabilistic forecast of day-ahead electricity prices,” Appl Energy, vol. 250, no. 4, pp. 1158–75, Sep. 2019.

8. D. Bzdok et al., “Statistics versus machine learning,” Nature Methods, vol. 14, pp. 232–4, 2018. 

9. T. Hastie et al., The Elements of Statistical Learning. Springer, 2014. 

10. Wikipedia., “Data mining.”

11. V. Franki et al., “A comprehensive review of artificial intelligence (AI) companies in the power sector,” Energies, vol. 16, no. 3, p. 1077, Nov. 2023. 

12. C. J. Khare et al., “Chapter 28 - Optimal power generation and power flow control using artificial intelligence techniques,” in Renewable Energy Systems: Modelling, Optimization and Control (Advances in Nonlinear Dynamics and Chaos - ANDC), 2021, pp. 607-631 

13. Equigy. “About equigy” 

14. S. R. Fahima et al., “A deep learning based intelligent approach in detection and classification of transmission line faults,” Int. J.  Elec. P & En. Syst., vol. 133, p. 107102, Dec. 2021. 

15. EU AI Act. “Recital 34, 2024 EU Artificial Intelligence Act,” 

16. F. Heymann et al., “Operating AI systems in the electricity sector under European’s AI Act – Insights on compliance costs, profitability frontiers and extraterritorial effects,” Energy Reports, vol. 10, pp. 4538–4555, Nov. 2023. 

17. W. J. Gross et al., “Hardware-aware design for edge intelligence,” IEEE Open J. Circuits Syst., vol. 2, pp. 113–127, 2021. 

18. Feng Cheng et al. “Smart grid encounters edge computing: opportunities and applications,” Adv.App. En., vol. 1. no. 23 100006, Feb. 2021. 

19. V. Franki et al. “A comprehensive review of artificial intelligence (AI) companies in the power sector,” Energies, vo. 16, p. 1077, 2023. 

20. CORDIS. “Very efficient deep learning in IOT.”