Smart Cities + Energy + AI : EDF Labs perspective

Size: px

Start display at page:

Download "Smart Cities + Energy + AI : EDF Labs perspective"

Virginia Day
5 years ago
Views:

1 Titre Smart Cities + Energy + AI : EDF Labs perspective Stéphane Tanguy, EDF Labs CIO IA Symposium

2 Summary 1. EDF Labs introduction 2. EDF Smart City platform 3. Singapore HDB project results 4. Electric mobility 5. Smart charging and AI IA Symposium

3 Major Forces disrupting the utility industry ELECTRIC Transition CLIMATE Change DIGITAL & SOCIAL Transition IA Symposium

4 EDF Labs 4 strategic priorities ELECTRIC Transition CLIMATE Change DIGITAL & SOCIAL Transition IA Symposium

5 Urban planning as a foundation of Smart Cities Why energy providers should get involved from planning stage Source: Newman&Kenworthy

6 Smart cities : energy & comfort Air pollution Noise Urban heat Island Results from Centrale Lyon, Study for EIFER &EDF, 2015 Results from CEREMA, Study for EIFER &EDF, 2015 Image LANDSAT des températures à Strasbourg, le 14 juillet Capture d écran ADEUS

7 Smart cities : supporting urban project introducing energy planning

Evolution of oil price, population, interest rates, temperatures Initiatives: parameters, time, location, describing investments,

8 Prospective Time T1 Building refurbishment applied on neighborhood A Time T2 PV investments applied on buildings X1+X2+ +X22 Time T3 District heating applied on neighborhood B+C+D Exogenous scenarios User + + Simulation = scenarios Prospective for urban developement Evolution of oil price, population, interest rates, temperatures Initiatives: parameters, time, location, describing investments, incentives, or urban policies Integrated model with multiple interactions at each time step Results corresponding to each scnenario: Decision support

9 2012 THOMAS SINGAPOUR SINGAPORE HDB PROJECT I EDF I 2013

SINGAPORE HDB PROJECT Singapour, Housing & Development Board

KPI definition Technology choices Scenarios Long term

Precise scenarios Impact simulation Results analysis 3D

10 SINGAPORE HDB PROJECT Singapour, Housing & Development Board Refurbishment New neighborhoods Strategic recommendations KPI definition Technology choices Scenarios Long term impacts Decision support platform Initiative parameters Precise scenarios Impact simulation Results analysis 3D visualisation Yuhua (refurbish.) 38 buildings (new) 800 buildings 10 I EDF I 2013

PROJECT RESULTS Implement Greenery to reduce cooling

of appliances Install PV panels to produce local

11 PROJECT RESULTS Implement Greenery to reduce cooling needs Green Home Package to improve energy efficiency of appliances Install PV panels to produce local energy Energy Efficiency Green Energy Up to -12% Energy Cons. ~ 2,700 MWh/yr prod. CO2-30 % Improve technology of Light in common areas with LED Costs SGD xxxk investment

12 Electric Mobility : opportunity and challenges for cities and utilities and where IA can help

13 Why Electric Vehicle charging should better be smart! France 2035 scenario : 9 Million Electric Vehicles 30% of national vehicles fleet 5% of national electricity consumption Without charging optimization : 20% of peak consumption without charging optimization : 5% of peak consumption!! Smart management of EV charging offers multiple services opportunity for the electric system : Supply and demand equilibrium Frequency and voltage regulation Absorption of renewable production surplus Different smart charging strategies to accompany each step of EV roll-out: Plug & charge Basic control with price signals Advanced control at the building level Advanced control at the electric system level Advanced control with bidirectional power flow

14 Leveraging AI to turn charging smart! Problem statement : How to optimize district EV charging in order to take into account electric network constraints? Proposal : leverage reinforcement learning to train an EV charging algorythm (home controller) taking into account home electricity consumption, mobility habits, home presence, price signal.. Results after 300 iterations : vehicle always charged, cost optimized, 1 charge period Perspectives : - Adapt algorythm to specific EV owner requirement (tipical schedule, charging rish aversion, cost aversion..) - Fleet management with Pmax constraint (charging operator, 1 central controller) - Complete district optimization with PowerFlow constraint (DSO use case) - Multi agent modelisation at district scale IA Symposium

An Agent-Based Information System for Electric Vehicle Charging Infrastructure Deployment

An Agent-Based Information System for Electric Vehicle Charging Infrastructure Deployment Timothy Sweda, Diego Klabjan Northwestern University June 7, 2013 1 Outline Background on EVs Proposed model Implementation