MAKING RENEWABLES SMARTER · 3 DNV GL © 18 April 2019 Road Map 01 What is AI 02 Current applications in Renewables 03 Visions of the Future 04 Panel Discussions

DNV GL © 18 April 2019 SAFER, SMARTER, GREENER DNV GL ©

Elizabeth Traiger, Ph.D. M.Sc.

18 April 2019

MAKING RENEWABLES SMARTER

1

Artificial Intelligence and Machine Learning Applications in the Wind Industry

DNV GL © 18 April 2019

Data = New Oil… AI = New Electricity

2

DNV GL © 18 April 2019 3

Road Map

What is AI 01

Current applications in Renewables 02

Visions of the Future 03

Panel Discussions 04


01 What is AI? Perceptions and Definitions

4


Buzz Words Need Definitions

5


DS, ML & AI

Data science produces insights

Machine learning produces predictions

Artificial intelligence produces actions

6


Data Science produces Insights

7

Statistical inference

Data visualization

Experiment design

Domain knowledge

Communication


Machine learning produces predictions

8

Machine Learning

Pattern Recognition

Separation

Estimates

Generalization


Machine Learning produces Predictions

9

Statistics

Primary Data Analysis

‘Top Down’

Hypothesis Testing

Model Driven

Confirmatory Analysis

Machine Learning

Include Secondary

Observational Data

Hypothesis Generation

Data Driven

Knowledge Discovery


Machine Learning Based On Data

10

Supervised

Classification Regression

Unsupervised

Clustering Dimension Reduction

Other

Reinforcement Learning

Adversarial Networks


Common Supervised Machine Learning Algorithms

11 https://machinelearningmastery.com/blog/


Common Unsupervised Algorithms & Other

12

Reinforcement Learning


Natural Language Processing (NLTK) - LSTM

13


Artificial intelligence produces actions

14


Automation is not Artificial Intelligence

15

≢

https://www.serbot.ch/en/solar-panels-cleaning/gekko-solar-farm-robot


Machine learning underpins advancements in AI

Properties of AI – human-like capabilities Converting human-like capabilities to data

science and ML

16

See: Image and video recognition

Hear: Understand input via text or

spoken language

Speak: Respond meaningfully to

input (from ‘hear’)

Make human-like decisions:

Offer advice or new knowledge

Learn: Change its behavior based

on environment changes

Move: Move and interact with

physical objects

Image processing:

Convolutional NNs (CNNs)

Natural Language Processing:

Recurrent NNs (RNNs), e.g. LSTM

Question Answering Machines

(Ontologies, e.g. IBM Watson)

Unsupervised learning:

Generative Adversarial NNs (GAN)

Reinforcement learning

(e.g. by RNNs)

Robotics

D


Big Data in the Renewables Industry – Potential for AI

17

SCADA

Atmospheric Performance

Demand Response

Temperature

Grid

Market

IoT


02 Applications in Renewables

18


Generative Design & Computational Chemistry

19 https://www.autodesk.com/solutions/generative-design


Data QC - Cleaning

20


Automating Preconstruction and Operational Assessments

21

Outputs


CFD Flow Modelling

22

NREL using Gaussian Processes in model

corrections

Autodesk replacing CFD numerical simulations

and solvers with CNNs


Optimization via Reinforcement Learning

Need to carefully define utility

23


Construction & Decommissioning – Autonomous Robotics

24

Remote-controlled machines begin dismantling a cooling tower at the Mülheim-Kärlich nuclear power station on the banks of the Rhine Photograph: Thomas Frey/AFP/Getty Images

Built autonomous dozer

Canrig Robotic Technologies autonomous robotic drilling

rig for unmanned drilling operations


Autonomous Drone Inspections & CV Analysis

25

E Smart Systems & SkySpecs AI software


Computer Vision - Object Recognition within Energy

Environmental Permitting

GIS & Satellite Surveys

Field Inspection Services

Laboratory Services

Flow Modelling

26


Forecasting – Resource, Demand, Price, etc.

28


Non-Traditional Industry Players

29


Power Curve Specifications – Change in Standards

PCWG & IEA Task 32– Regression Forest Ensemble Models

University of Strathclyde – Gaussian Process Models

30


Condition Monitoring for Predictive Maintenance

31

Time

Anomaly KPI

Normal Behaviour Older Threshold Methods


Production Performance Monitoring

32


AI - Wind Farm Control – Solar Dual Axis Panel Control

33


Grid Balancing

34


NLTK in Renewables – Reduce barriers to entry

Maintenance Diagnosis

Independent Engineering Contract Review

Site Inspection Log Analysis

35


Change in the Air - Financials, Market Analysis & Consumer Behaviour

36

Financing Incentives in Area Consumption Recommendations


04 Visions of the Future

37


ETIP Wind(.eu) thoughts

38


DNV GL’s Vision of the Future

39


SAFER, SMARTER, GREENER

www.dnvgl.com

The trademarks DNV GL®, the Horizon Graphic and Det Norske Veritas® are the properties of companies in the Det Norske Veritas group. All rights reserved

Thank You

40

Elizabeth Traiger, Ph.D., M.Sc.

[email protected]

Senior Researcher, Group Technology & Research – Power & Renewables

MAKING RENEWABLES SMARTER · 3 DNV GL © 18 April 2019 Road Map 01 What is AI 02 Current applications in Renewables 03 Visions of the Future 04 Panel Discussions

Documents