invade exploitation workshop€¦ · use machine learning to predict production, consumption and...
TRANSCRIPT
This project has received funding from the European Union’s Horizon 2020 Research and Innovation programme under
Grant Agreement No 731148.
Smart system of renewable energy storage based on INtegrated EVs and bAtteries to
empower mobile, Distributed and centralised Energy storage in the distribution grid
The INVADE software platform
eSmart – Stig Ø. Ottesen, Knut H. H. Johansen
INVADE Exploitation Workshop
Copenhagen 06/03/2019
Agenda / Content
1. Role in INVADE
2. Achievements so far
3. What is eSmart Systems?
4. Ambitions in the project
5. Business case and target stakeholders
1
Role in project
2
Role in project
3
1. Have information about
flexibility sources, contracts
and flexibility services
2. Collect meter values every
15 minutes
3. Use machine learning to
predict production,
consumption and available
flexibility for near future
4. Use optimization
techniques to find optimal
ways to utilize flexibility to
obtain pre-defined
objectives
5. Send flexibility plans to
local systems
Achievements so far
4
1. Developed first version of
Integrated INVADE Platform
2. Ready to receive data and run
flexibility services for the pilots
3. First customer outside INVADE
consortium already up and
running
4. Lot of interesting business cases
from INVADE exploitation
What is eSmart Systems?
© www.esmartsystems.com
At eSmart our mission is to build digital intelligence to provide exceptional solutions to our customers and accelerate the transition to sustainable societies.
Proven Products with an Increasing Market Traction
Connected Drone: Analytics software for infrastructure inspections
Connected Prosumer: Analytics software for distributed energy resources
+ Improves infrastructure inspections through image recognition analysis
+ Automatic detection of grid anomalies (e.g. missing components, objects on power lines), ability to analyse 180,000+ images in less than one hour
+ Benefits the utility by more efficient inspections and analysis of data and in turn improved maintenance of the grid
+ Increases the utilization of distributed energy resources though AI-powered predictions of consumption and generation and sophisticated optimization models
+ Benefits the prosumer with higher utilization of energy resources and reduced power bills
+ Benefits the utility by enabling energy flexibility to relieve grid constraints and postpone investments
eSmart and AI - Background
8. February 2019
Institute for Energy Technology (IFE)NOK 10 billion invested in R&D focused on Applied
AI, Real-time Data and Advanced IoT
International co-operative effort affiliated to OECD (Organization for Economic Co-operation and
Developmentin in Paris) – hosted and run by IFE
Gunnar Randers together with Albert Einstein in 1939 Einstein’s home - Long Island, 1939.
Background – Applied AI in Critical Processes Started 1985 in Halden
eSmart - Focus on AI
Global attention #1 in Grid Analytics Competition in US Among top 4 Microsoft AI partners
out of 2600 AI companies
Knut H. H. Johansen © eSmart Systems
eSmart Systems & AI
AI & Machine LearningTop 100 Influencers and Brands
#30 @josephsirosh
eSmart Board of Directors
Source: Onalytica
An irreversible and dramatic change in human history
Three technologies converge into
extreme power
Exponential growth in volume of available data
Digital Technology Convergence
Exponential growth in computational power and storage
Exponential growth in development of sophisticated AI algorithms
a revolution
Alpha Zero
2017Learned by self-play alone …
…. in 9 hours !!
Tabula-rasa – No prior hard-coded knowledge
played 44 million games against itself
Artificial Intelligence – example – Alfa Zeroa chess-playing computer
If AI can create the worlds best chess player in 9 hours
what can AI then do for your business?
Knut H. H. Johansen © eSmart Systems
Artificial Neural Networks – Early Stage
Source: Cybercontrols
Knut H. H. Johansen © eSmart Systems
Deep Learning (Neural Networks) - Today
Source: NVIDIA
Knut H. H. Johansen © eSmart Systems
Deep Learning (Neural Networks) - Today
Input Output
Deep insight in utility operations and planning
Domain Insight
IoT, Storage, Graph,
Security, Microservices,
Data
EngineeringDeep learning, detection, prediction, segmentation, recommendation
Machine
Learning
Domain Insight
Data Engineering Machine Learning
© www.esmartsystems.com
Leveraging smart meter (AMI) data to improve operations and recover lost revenues
• Jacksonville, Florida – smart metering (AMI) of
water and electricity – 300 000 customers
• 7500 truck rolls a year to check zero
consumption meters - 80% are wasted
• Advanced analytics to estimate likelihood of
broken water meters
• 86.5 % prediction accuracy
• Annual net value for JEA $1.1 million
Ambitions in the project?
New application of energy flexibility
Large scale demand response for Norwegian TSO Statnett
Connected ProsumerDemand side management to solve transmission-level challenges in northern Norway
+ A solution for Statnett, the Norwegian TSO
+ Delivery: A system providing Statnett the possibility to manually disconnect load through a dashboard at Statnett’sown control center
+ Project headed by eSmart Systems
+ Project management
+ Software development
+ Cooperated with several sub-contractors:
+ Load provided by balance responsible parties
+ Statkraft and Ishavskraft
+ Load from industrial, public and residential sectors
+ Hardware provided by E2U Systems
SWEDEN
420 kV power line300/220 kV power line132 kV power linePower stationSub station
NORWAY
Polar Circle
Connected ProsumerDemand side management for commercial prosumers
Energy assets+ Solar PV generation+ Thermal storage+ Battery storage+ Flexible forklift charging
eSmart deliveries+ Intelligent energy management dashboard
+ Predictions+ Optimization+ Activity based energy control
Pain relief: energy management to reduce energy costs and increase energy resource efficiency
Energy assets+ Solar PV generation+ Thermal storage
eSmart deliveries+ Intelligent energy management dashboard
+ Solar monitoring+ Predictions+ 100 kW regulatory limit
Pain relief: avoid costly peak power tariffs, and optimize the utilization of local PV production
Energy assets+ Solar PV’s+ HVAC systems with heat recovery+ Large electric boilers+ Thermal storage
eSmart deliveries+ Visualization of energy flow (public)+ Predictions for loads and available
flexibility+ Dynamic “smart” EV charging
Pain relief: avoid costly peak power tariffs and to provide a common energy platform across a diversified portfolio of shopping malls and equipment
DER monitoringDER and activity data streams
monitored real-time
DER schedulingDemand side flex utilized for
optimization scenarios
Demand side flexibility market bidding
Aggregated flex portfolios optimized for market bididng
Monitor IoT – energy assets
+ Collect data
+ Aggregate
+ Monitor
+ Visualize
Intelligent decision-support for
scheduling flexibility portfolios to
optimize value according to
+ ToU
+ kWmax control
+ Self-balancing
+ EV charging capacity
optimization
The Connected Prosumer Product
Intelligent decision-support for
bidding flexibility portfolios
according to
+ Multi-market bidding strategy
(“Maybid”)
+ Offered by many
+ IoT and HW-integrations
+ Data management
+ Solved without AI
+ Offered by many
+ Can be solved with AI
and optimization, but in
many cases it is not
needed
+ Few if any players at the
moment
+ Complex flexibility
modelling
+ AI and optimization
needed
Main processes of Connected Prosumer
Business case and target stakeholders
Product Adaption Life Cycle – Connected Prosumer
The Chasm – ideal point to enter the market
Source: Geoffrey Moore - Crossing the Chasm
Early market Mainstream market
enthusiasts visionaries pragmatists conservatives skeptics
It is the ability to quickly segment and activate reference customers –putting the right customer in front of the right prospect – that is the key to crossing the chasm and solidifying a foothold in the mainstream market.
Push from eSmart
No Pull from Prosumers
First Pull from
Prosumers
Target segments with current customer examples
+ DSO’s
+ TSO’s
+ I&C prosumers
+ Energy retailers
+ Power infrastructure operators + Middle-sized utilities
+ I&C prosumers
Business model – Software as a Service
Specification SaaS
1-4 weeks 3+ years
Direct
SaaS
PoC to SaaS Specification Customization PoC SaaS
1-2 weeks 4-6 weeks 3 months 1-3 years
Development project SaaS
1-3 years 1-3 years
R&D to SaaS
+ The direct way to SaaS contract
+ Period of work prior to start of SaaS relatively short with specification and customization to customer needs necessary before starting SaaS contract
+ Typically takes 3-6 weeks combined
+ Recurring revenue from start of SaaS. SaaS contract length initially 1-3 years
+ Proof of concept (“PoC”) contract in which a customization and proof of concept phase is added
+ PoC period provides extra comfort to customer when purchasing new solutions
+ Option to start SaaS subscription with initial length of 1-3 years given successful PoCphase
+ Nature of R&D or project is less commercial than two examples above
+ Development project phase naturally takes up majority of project time
+ Billing in both phases of project (development is paid)
• Note: Black arrows/lines indicate timing of signature/commercial agreement
This project has received funding from the European Union’s Horizon 2020 Research and Innovation programme under
Grant Agreement No 731148.
Any question or comment?
Thank you!