PhD - AI-based demand-response strategies for an energy-consuming operator F/M
Orange
- Châtillon, Hauts-de-Seine
- CDI
- Temps-plein
Your role is to carry out a PhD on AI-based "demand-response" strategies for a prosumer energy operator in the context of the energy and ecological transition: electrification, proliferation of renewable energies, storage optimisation and energy flexibility.
The latter can be achieved in particular by implementing "demand-response" strategies, which involve aligning consumption with renewable energy production.
These demand-response strategies can be implemented in a marketplace made up of consumers and producers/prosumers and governed by the law of supply and demand.
Through its energy assets (antennas, batteries, solar panels, electric cars, buildings, etc.), the operator can take part in demand-response contracts.
These strategies encompass a number of techniques, including the implementation of global or partial network standby modes and the sharing of infrastructure between operators, while maintaining a good level of service.Scientific objective - results and challenges
The aim of the PhD is to optimise demand-response strategies for a telecom operator in a context of uncertainty (variable production of renewable energy, fluctuations in prices and consumption, volume of network traffic) and faced with adversarial or cooperative agents participating in the exchange.
The function to be optimised will take into account the carbon footprint, energy sale/purchase prices, quality of service, etc. And the decision-making policy will be based on storage management, electricity trading, putting base stations on standby, etc.The main obstacles to be overcome :
1. Modelling of a multi-agent system capable of integrating demand-response strategies.
2. Multi-objective optimisation: reducing the carbon footprint, reducing electricity bills, preserving quality of service.
3. Alignment of time scales between decisions applied to the electricity network and the telecoms network.
4. Heterogeneous nature of actions (continuous, discrete).
5. Complexity of implementing deep reinforcement learning algorithms: convergence problem, safe exploration.
6. Robustness of the model in an environment different from that of the simulator.Key deliverables:
1. Simulator of an exchange integrating radio sites as a reinforcement learning environment covering several scenarios.
2. Design and development of multi-agent deep reinforcement learning algorithms to optimise decision-making.
3. Design risk-free exploration strategies.
4. Study robustness with analytical and experimental results.
5. Scientific publications.about youSkills and personal qualities required by the post- In-depth knowledge of artificial intelligence, in particular reinforcement learning and deep learning,- Skills in optimisation theory and game theory will be highly appreciated,- Knowledge of mobile access networks is desirable,- Fluency in Python programming, in particular PyTorch.Education requiredMaster's degree or engineering degree specialising in AI.additional information- Development in a stimulating environment within a department specialising in artificial intelligence and data science and staffed by experts in the field.- Participating in Orange's involvement in CSR (Corporate Social Responsibility) issues, particularly the energy transition.- Strengthening Orange's ambitions in the field of energy.- Research work requiring an interdisciplinary approach, crossing several areas of computer science, mathematics and telecommunications such as data prediction, reinforcement learning, multi-agent systems, neural networks, mobile access network (MAN) architecture and modelling the energy consumption of 4G/5G base stations and beyond, etc...- Potential contribution to collaborative projects.departmentThe Innovation Division's ambition is to take Orange's innovation even further and strengthen its technological leadership, by mobilizing our research capabilities to nurture responsible innovation at the service of people, inform the Group's long-term strategic choices and influence the global digital ecosystem.
We train the technology experts of today and tomorrow, and ensure continuous improvement in the performance of our services and our efficiency. The Innovation Division employs 6,000 people worldwide dedicated to research and innovation, including 740 researchers. Carrying a global vision with a wide range of profiles (researchers, engineers, designers, developers, data scientists, sociologists, graphic designers, marketers, cybersecurity experts, etc.), the men and women of Innovation listen to and serve countries, regions and business units to make Orange a trusted multiservice operator.The Data & AI division's objectives are to define Orange's Data & AI standards and to develop use cases, products and services based on AI and data science.
The reporting team is DREAMS (Data science, REsearch, Algorithms & ModelS). Its projects concern the application of AI to network and service issues, the development of voicebots, and research into sustainable territories and healthcare.contractThesis