Artificial intelligence applied to energy attracts many announcements and few concrete assessments. This study, conducted jointly with AI specialist Visium Technologies, examines five use cases from both technical and economic angles, to gauge their real feasibility and the value they create.
What machine learning can do, and under what conditions
Machine learning means teaching computers to perform three types of task: predict, classify and cluster. These correspond respectively to regressions, classifiers and partitioning algorithms.
The added value over classical methods lies in the ability to handle dozens or hundreds of variables drawn from data of every type. One condition outweighs all others: the quantity and quality of training data determine model performance. Training a classifier on a limited or poor-quality dataset sharply degrades its results.
The distinction between supervised and unsupervised learning is decisive for feasibility. Supervised learning, regression and classification, requires labelled data, meaning concrete examples whose outcomes could be collected. Unsupervised learning, or clustering, divides data into groups without those groups being predefined, which allows rare items to be detected or useful groupings to be formed.
Five use cases examined
- Predictive maintenance, applied to generation assets and the electricity network.
- Customer insight tools, for personalised marketing to end consumers.
- Customer management tools, within the same commercial scope.
- Smart consumption management and energy efficiency.
- Potential identification algorithms, whether for grouping electricity consumption, biogas production, or deploying infrastructure such as EV charging points.
A performance tool, not a replacement for jobs
This is the study’s most useful conclusion, and it runs against the prevailing narrative. Over the medium term, AI should not cause any job to disappear in the energy sector. It will however be a key element of operational performance, in the face of growing sector complexity.
For electricity distributors and suppliers, with full retail market liberalisation in prospect, these applications fall into two distinct categories.
Cost reduction potential, driven by predictive maintenance and personalised customer management tools.
A positioning opportunity in services, through energy flow optimisation, smart consumption management, and the handling of groupings and new interfaces between consumption and generation.
It is this second category that carries the greater strategic stake, since it does not merely improve an existing business but opens new territory.
To go further, download our full analysis below or contact our experts.
