sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22...
Transcript of sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22...
![Page 1: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/1.jpg)
Thursday 23rd of July 2020
Learning to Run a Power Network in a sustainable world
Antoine Marot – Lead AI Scientist @ RTE
A NeurIPS 2020 Competition https://l2rpn.chalearn.org/
![Page 2: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/2.jpg)
Introduction01
![Page 3: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/3.jpg)
3
AI is BOOMing
AI based on Deep Learning started to impact many fields & layers of society at a fast pace
Keys of success:• Large Benchmarks & leverage big data• Leverage computation• Open-source platform & papers• Deep Learning revival
NB: not much new algorithms & theory
ImageNet (2012) & NLP (2018) moments
![Page 4: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/4.jpg)
4
« Artificial Intelligence is the new Electricity », Andrew Ng
Are blackouts a story of the past ?
AI & Electricity
Is Electricity a 20th century
solved challenge?
![Page 5: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/5.jpg)
5
Quizz : How much time a grid could run without human intervention ?
1 minute 1 hour
Up to the next storm Forever
![Page 6: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/6.jpg)
6
Quizz : How much time a grid could run without human intervention ?
1 minute 1 hour
Up to the next storm Forever
![Page 7: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/7.jpg)
7
« Artificial Intelligence is the new Electricity », Andrew Ng
Are blackouts a story of the past ?
AI & Electricity
NOT REALLY!
![Page 8: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/8.jpg)
8
A control room today
Quite a complex work environment !
![Page 9: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/9.jpg)
9
Yet, Another system to consider : Earth!
⇒ Sustainability is the key, future smart grids have to do better with less
Renewables ressources & AI ? Fossil fuels & grid expansion
![Page 10: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/10.jpg)
10
AI & Electricity« Artificial Intelligence is the new Electricity », Andrew Ng
Is Electricity a 20th century
solved challenge?
Yet power grids are one of the most complex artificial systems
Will AI be the 21rst
century achievement ?
![Page 11: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/11.jpg)
11
« Artificial Intelligence is the new Electricity », Andrew Ng
How AI can be of any help here ?
AI & Electricity
![Page 12: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/12.jpg)
L2RPN challenge02
![Page 13: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/13.jpg)
13
« Learning To run a power network »L2RPN Challenge
1) Test the potential of AI to robustly operate a power grid in real-time given operational constraints.
![Page 14: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/14.jpg)
14
Different kind of flexibilitiesOn Grid Flexibilities
On Production and Load Balancing
Topology:• Branch switching• Node merging/splitting
Power devices: • Transformers• Phase shifter• HVDC• …
Redispatching Load shedding
All over the grid & 0$ ! Few & local & 0$
Many over the grid & $$$ Many over the grid & $$$$$
Lots of research
Some researchLittle research
2) Explore underutilized cheap topology controls with methods can deal with non-linear, non-convex and combinatorial complexity
![Page 15: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/15.jpg)
15
Motivations for a challenge
Exhibit important real-world problems to the research community
Large Benchmark for Reproducible Science:• Decouple the modeler (problem design) & the solver
Attract new communities, especially AI community, through an easy-to-use platform & a gameified problem
![Page 16: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/16.jpg)
16
Machine Learning & power systems
Power System Community AI Community
Mainly Today
![Page 17: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/17.jpg)
17
Artificial Intelligence & Power Grids
One small step for Electricity, one giant leap for Sustainability !
![Page 18: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/18.jpg)
Challenge overview02
![Page 19: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/19.jpg)
19
2019 Challenge overview
Website
• 6-week challenge• Participants from both power system & ML communities all around the world• To get the prize, winning teams had to open-source their code• Travel expanse Award for the best 2 teams (IJCNN conference or Paris at RTE)
![Page 20: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/20.jpg)
Gym openAI framework20
« Learning to run » inspirationHow to build a smart controller ?
Simulator + ModelPython API
openSim.py
Game Environment
Smart controler!
AI
![Page 21: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/21.jpg)
21
GridAlive ecosystem
Grid2Op: testbed platform to model real-time operations, run & benchmark control algorithms https://github.com/rte-france/Grid2Op
Data
ChroniX2Grid
https://github.com/mjothy/ChroniX2GridPower Grid time serie generation
Analysis
Grid2Viz
https://github.com/mjothy/grid2viz Scenario & agents study
AGENT
https://github.com/rte-france/l2rpn-baselines
L2RPN-BaselineRepo to share open-source models
![Page 22: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/22.jpg)
22
Modeling of power system operation worldObservation: flows, productions, consumptions, power grid topology, month, day, hour, etc
Action: connect/disconnect one transmission line or change the electrical configuration within a substation
Reward: penalize overflowed lines, distance to reference grid topology, number of disconnected loads/prods, etc.
Fig. 2 - Step-by-step evolution of the RL environment
Fig. 1 - Reinforcement Learning interaction loop
Maximize
![Page 23: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/23.jpg)
23
Agent evaluation – score functionControl the power flows to optimize the cost of operations on the power grid
while being robust to blackouts.
=> Participants are tested over 10-20 scenarios of different difficulties, with duration between 3-7 days.
&
Under N episodes, the final score to minimize is:
![Page 24: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/24.jpg)
Environment design03
![Page 25: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/25.jpg)
25
Game requirements
Criteria to Design and calibrate an environement:
• Be Realistic (represent real world caracteristics)
• Be Interesting (Diverse actions in diverse context to control the game )
• Be Feasible (existance of solutions to avoid game overs)
Line Capacities were the main design parameters to make a trade-off between feasible and interesting.
![Page 26: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/26.jpg)
26
Realistic: Production energy mixGenerate chronics & Localize generation units
![Page 27: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/27.jpg)
27
Realistic: Line Capacity Calibration
When building a line, its capacity should be properly sized according to the expected flows over a year.
The line capacity should be:
1) Close to the max flow we can observe.
2) Sometimes below• Because the line has been built few years ago, the system has changed
and will not be replaced soon. • This is when congestions can appear and needs to be controlled few % of
the time.
![Page 28: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/28.jpg)
28
Realistic & Interesting: Operational Constraints
The game should represent operational constraints:
1) A limited time to react to a congestion (2 ts)
2) A limited number of simultaneous action (1/ts)
3) A cooldown time before reusing an asset (3 ts)
There is hence a budget associated to the actions you take: they should be picked up carefully !
![Page 29: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/29.jpg)
29
Interesting & Feasible: existence of solutions
Minimal Grid: IEEE14 is the minimal grid to play with the topology because it is sufficiently meshed
Strutural feasability: Existence of Multiple electrical path to reroute power flows and avoid overloads.• Easier with larger meshed grids• But choice of the grid for a minimalist problem and facilitate
challenge analysis
Contextual feasability:Found an ensemble of simple baselines that solve a high percentage of overloads (85%)
2 electrical corridors cannot be overloaded at the same time
![Page 30: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/30.jpg)
30
Interesting - Selecting test scenarios
Our goal was to pick up 10 scenarios of different length (1 to 3 days) over different days of the week and with varying difficulties:
• Easy: 2 scenarios in which Do Nothing agent finishes the scenarios
• Medium: 6 scenarios were Overloads exists & Do Nothing fails, but which are solved by our baselines.
• Overloads could happen around peak consumption or under high renewable production.
• Hard: 2 scenarios our expert baselines could not solve
![Page 31: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/31.jpg)
L2RPN: competition series04
![Page 32: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/32.jpg)
Ongoing L2RPN serie competitions
Large Grid, 2 environments,Topology & redispatching actions
Participate and join us!https://l2rpn.chalearn.org/
![Page 33: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/33.jpg)
33
Sponsors & Collaborators
![Page 34: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/34.jpg)
34
Materials available
![Page 35: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/35.jpg)
2019 Results04
![Page 36: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/36.jpg)
36
Final Leaderboard
A mix of Machine Learning agents, expert systems and power system controllers were submitted
Explanatory videos of the winning teams describing their approach are available here: https://l2rpn.chalearn.org/competitions#h.p_6EGPeZwih5BD
![Page 37: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/37.jpg)
37
Agent behavior over time
Agent behavior over validation scenario 3 showing the depth of agent actions at time-step
BS ES
Expert system tries a lot more instantaneous actions while Leaning agent take less actions but with more anticipation. Overall, Learning agents showed to be more robust.
![Page 38: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/38.jpg)
38
Performance comparison
Agent
BS
Machine Learning
Expert System / OPF
![Page 39: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/39.jpg)
39
Room for improvements
Oracle to find the best topology actions aposteriori: • Run many topology in parallel• Create edges between topologies you can transition to• Find the best scoring path (longest path)
Oracle actions over time Normalized Agent scores (Oracle=100)ES DQN
Beside overcomming overloads, not yet learning to optimize the flows continuously
Assessing how good agents did given an oracle
![Page 40: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/40.jpg)
40
Best models are now open-source baselineshttps://github.com/rte-france/l2rpn-baselines
![Page 41: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/41.jpg)
2020 WCCI first results04
![Page 42: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/42.jpg)
42
WCCI competition is over
Strong scores for top 3 participants:
1) South Corean working at Naver
2) Chinese working at Baidu3) PHD at Singapour (Data Science et
Power system)4) Chinese Researcher (working in cyber
physical power systems) - but double account
=> We check the leaderboard and give prizes to those who share their code and complied with the rules
All from Asia ..!
![Page 43: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/43.jpg)
43
Detailed winner results on test set
Do Nothing Winner
![Page 44: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/44.jpg)
44
Smart but risky strategy
- When a line is overloaded, the agent disconnects its extremity before it gets disconnected automatically (=> it remains available )
- To avoid cooldown time, it uses the multiple ways to connect and disconnect lines
![Page 45: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/45.jpg)
45
It learns interesting Topologies
Changing the topologies (in green) of 3 substations (16,23,26) leads to a robust grid topology
![Page 46: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/46.jpg)
![Page 47: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/47.jpg)
![Page 48: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/48.jpg)
More at : https://l2rpn.chalearn.org/competitions#h.p_X7vG2_hacIL_
![Page 49: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/49.jpg)
Join us now!05
![Page 50: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/50.jpg)
50
L2RPN NEURIPS 2020 - Robustness TrackDevelop your agent to be robust to unexpected events and keep delivering reliable electricity everywhere even in difficult circumstances. You will have to overcome an opponent attacks and keep operating the grid safely.
https://competitions.codalab.org/competitions/25426
![Page 51: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/51.jpg)
51
L2RPN NEURIPS 2020 - Adaptability TrackIn this track, develop your agent to adapt to new energy productions in the grid with an increasing share of less controllable renewable energies over years. You are given a multi-mix environment, that is a set of environments with varying amount of renewables (varying from 1x to 3x).
https://competitions.codalab.org/competitions/25427
![Page 52: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/52.jpg)
52
Conclusions
1. Machine Learning approaches are promising for control problems and here showed to generalize better than Expert Systems
2. It is now possible to learn topology controllers, opening a new field for research and smart grid flexibilities.
3. Challenge helps develop benchmarks and enforce reproducibility to make faster and stronger progress as a community
4. We should keep working on attracting AI researchers and collaborating with them on power system related problems
![Page 53: sustainable world Antoine Marot – Lead AI Scientist @ RTE ...yweng2/Tutorial5/pdf/antoine...22 Modeling of power system operation world Observation: flows, productions, consumptions,](https://reader033.fdocuments.us/reader033/viewer/2022060920/60abfb1713c6d75c7e393989/html5/thumbnails/53.jpg)
Ongoing L2RPN serie competitions
Large Grid, 2 environments,Topology & redispatching actions
Participate and join us!https://l2rpn.chalearn.org/