the energy transition & the electricity network
TRANSCRIPT
![Page 1: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/1.jpg)
16-01-2019
Simulations, data science and machine learning
Werner van Westering MSc.
Senior data scientist & PhD candidateData & Insight Alliander
The energy transition & the electricity network
![Page 2: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/2.jpg)
Index
2
1. Introduction
2. Our context
3. Highlights
4. Case I: Network planning
5. Case II: Real time control
6. Outlook
Liander is the largest DNO in the
Netherlands and serves over 3 million
customers
• 3 million small-scale consumers
• 13,000 large-scale consumers
Key figures asset base Liander
• 50,000 km low voltage cable
• 35,000 km medium voltage cable
• 37,000 transformers
• 230 substations
Electricity & GasElectricity only
Key FiguresAlliander Service Area
![Page 3: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/3.jpg)
Our context
3
Challenge:
• Peak electricity load to increase three times
• 2.5% maximum replacement per year
Tools:
• 25 data scientists
• Data sets of:
- The entire Alliander grid
- Electricity consumption (anonomized)
- Sociographic data on neighbourhood level
• R/Pyhton programming language
• 512 GB RAM machines
![Page 4: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/4.jpg)
A few highlights
4
1. Asset health analytics
• ~200 variables
• Random forest
• Result: Priority list
2. Network recovery plans
- Real time data
- Linearized physics, brute force
- Result: Recovery plan in seconds
3. Smart meter deployment
- ~150 variables
- SVM for clustering
- Result: Risk map
![Page 5: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/5.jpg)
5
Case I: Network planning
![Page 6: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/6.jpg)
The ANDES model determines the impact of the energy transition
REGIONAL ADOPTIONLOCAL
ADOPTIONPROJECT ON GRID
TOPOLOGY
CALCULATE LOAD PROFILE PER
ASSET
DETERMINE # OF OVERLOADS
1 2 3 4 5
![Page 7: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/7.jpg)
ANDES technology adoption model
Input: 150 demographical aspects Output: The adoption is predicted at zip-code level per technology up to 2050
• Multiple regression techniques were studied.
• The result is an absolute number of EVs, HPs, and PV systems per zip code for every year.
…in the LianderService Area
Local adoptionfor each household…
Analysis: Probability of adoption is determined
Socio-demographic, e.g.:• income• education• Life phase
House properties, e.g.:• Type of house• Value house• Owned/rented
Financial info, e.g.:• Savings• insurance• Other financial info
Vehicle information, e.g.:• Number owned• Segment• Age
Media, e.g.:• Internet behaviour• Magazines• TV channel preference
Buying habits, e.g.:• Clothing segment• Holidays• Charity
Etcetera
Regression analysis
![Page 8: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/8.jpg)
Geographic representation of the technology adoption
Technology adoption percentage
PV 2030 High scenario
PV
EV
HP
20302023 2050
![Page 9: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/9.jpg)
An example of determining network overload
HV
Substation Transformer
MV
Distribution Transformer
LV
Sample power grid Load profile calculation
Customer load
Electric Vehicle charging
Solar feed-in
Heat pump
Large customer load
Total load at transformer
+
0
200
0
50
0
50
-50
0
0
50
0
500
Based on the analytical model with approximately 250 billion data points peak loads per asset are determined
![Page 10: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/10.jpg)
In development: INP AI Portfolio Optimization System
10
Optimization algorithm
Asset dataLoadConditionCustomer requirementsStrategy
Investment simulation
Investment model
Asset dataLoadConditionCustomer requirements
Project portfolio1. Replace cable X2. Service station Y3. New station Z4. Inspecti cable A5. Etc.
Content by Matthijs Danes and Willem van Doesburg
![Page 11: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/11.jpg)
11
Case II: Real time Control
![Page 12: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/12.jpg)
Case II: Buurtbatterij
![Page 13: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/13.jpg)
Algorithm design goal priority
13
Network
stability
Customers
self-sufficient
Energy
trading
![Page 14: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/14.jpg)
Predicting solar power generation
14
Variability:
- Time: 15 minutes
-Spatial resolution: 100 m
Results Research WUR & Alliander
So
lar
inte
nsi
ty[W
/m2]
State-of-the-art
Resolution of KNMI model:
- Time : ~1 hour
-Spatial resolution : ~2 km
Content by Frank Kreuwel
![Page 15: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/15.jpg)
Determining the optimal charge path
Optimization problem formulation:
![Page 16: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/16.jpg)
Buurtbatterij real time control
![Page 17: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/17.jpg)
17
Final remarks
![Page 18: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/18.jpg)
New insights warrant new decisions
18
Dilemma I: Is everybody equal or are some more equal than
others?
Dilemma II: Lower CO2 emissions or more reliability?
![Page 19: The energy transition & the electricity network](https://reader033.vdocument.in/reader033/viewer/2022052905/628ffb8011c4627b5e34f229/html5/thumbnails/19.jpg)
Final remarks
19
• Feel free to contact me for:
- Model/data exchange
- Co-authorship