Lo que nunca te han contado sobre

Digitalización del Sector de la Energía Eléctrica

¡ Bienvenidos !
1
Agenda
10.00 Experiencias en el sector Energético y Energías Renovables (en inglés)
Shishir Shekhar - WW Utilities & Energy

10.25 Cómo predecir la demanda energética con Inteligencia Artificial

Lucas García, pHD –AE Data Analytics

11.25 Bio-break 5 min

11.30 Estudios de Sistemas de Energía mediante Gemelo Digital

María Fernández –AE Simulación, Control, Modelado Físico y Verificación

12.30 Conclusiones

2
 MathWorks Technology in the Utilities & Energy
Industry - An Overview

Shishir Shekhar, MSEE,MBM,SMIEEE

Worldwide Industry Manager - Utilities & Energy
Natick, USA
[email protected]

MathWorks Confidential – Do Not Distribute © 2020 The MathWorks, Inc.

1
 Introduction

▪ Education:
▪ Masters in Business Management and Economics
– Harvard University, USA
▪ Research Associate
– Massachusetts Institute of Technology (MIT)
▪ Masters in Electrical Engineering-Power Systems
– Northeastern University, USA
▪ Bachelor of Engineering in Electronics and Communications
– SRM University, India

▪ Work Experience:
– Worldwide Industry Manager, Utilities & Energy - MathWorks Inc, USA (Current)
– Senior Manager - Grid Innovation and Research, National Grid UK
– Senior Research Engineer - National Grid USA
– Engineering Analyst- Cognizant Technology Solutions, India
– Trainee Engineer- Essar Power and Steel Ltd, India 2
Motivators

▪ Connecting Measured Data to Simulation

▪ Applying Optimization to Simulation
▪ Generating Code from Simulation Models
▪ Grid Integration Studies
▪ Performing ‘Non-Traditional’ Power System Studies
▪ Digital Twins
▪ Big Data and IoT
▪ Artificial Intelligence (AI)

3
Core MathWorks Products

The Language of Technical Computing

In the Utilities and Energy sector, MATLAB is

used primarily for data analysis, optimization,
algorithm development, predictive modeling,
and application deployment.

4
Core MathWorks Products

Simulation and Model-Based Design

In the Utilities and Energy sector, Simulink is

used primarily for grid integration studies, grid
equipment design, and equipment parameter
estimation and optimization.

5
Headquarters
Natick, MA USA Europe Asia-Pacific
France Australia
Germany China
North America Ireland India
United States Italy Japan
Netherlands Korea
Spain
Sweden
Switzerland
UK

MathWorks Inc Today

5000+ Staff $1.2+ billion 50% of profits Privately held

Global Workforce in 2019 revenues reinvested into R&D and profitable every year

6
 Transfer learnings from customers across 15+ industries

Aerospace and Defense Automotive Biological Sciences Biotech and Pharmaceutical Communications

Electronics Utilities & Energy Energy Trading Industrial Machinery Medical Devices

Process Industries Neuroscience Railway Systems Semiconductors Software and Internet

7 7
The Landscape

8
 Power System Studies & Compliance
Perform Techno-Economic Feasibility and Compliance Studies

Renewable Energy Integration & Microgrids

• Design & simulate power electronics, mechanics, and control systems.
• Develop optimization algorithms for economic dispatch
• Integrate with other tools such as PSSE and PSCAD

Power Plant Model Validation (MOD26/27) & Digital Twins

• Build Digital Replicas (Digital Twins) of physical assets and systems
• Use Field Voltage/Frequency and Match Plant simulated P/Q Response
• Run multiple scenarios and perform automated parameter estimation

IEEE 1547 Compliance with PMU/SCADA Data

• Check compliance with Measured Data
• Design Controller/Protection Compliance for Grid Codes
9
 Grid Analytics PdM, APM,
Monitor Asset and Power System Events Inventory Mgmt
…?

Asset Management
▪ Use Signal Processing and AI techniques to monitor asset health
▪ Develop Predictive Maintenance and RUL (Remaining Useful Life) Algorithms

Helicopter view of a 60K overhead

line crossing a densely forested area.

Assess Grid Stability with OSIsoft PI + MATLAB

▪ Integrate with OSI PI to access system data and perform analytics using MATLAB
▪ Combine visualization and online analysis

Energy/Price Forecasting
▪ Use AI to forecast Load, Generation and Prices at multiple nodes
▪ Integrate forecast models with grid simulation to improve system reliability
10
 Enable IoT/OT Convergence
Deploy simulation models and algorithms on the hardware or in the cloud

Digitalization of Grid Infrastructure

▪ Access data from sensors and IoT devices on the grid.
▪ Accelerate computation with GPUs
▪ Deploy applications on the cloud using Microsoft Azure or AWS and perform analytics

Real-Time Simulation
▪ Develop electrical system simulation models and deploy them to Speedgoat
hardware using Simulink Real-Time
▪ Perform hardware-in-the-loop (HIL) testing of control systems

11
One platform for End-End Digital Solution
Private, Public and Hybrid Cloud
Active Directory

Assets Production System Analytics Development

MATLAB

Operationalized Packaged
Analytics Analytics
(Application Server) (CI/CD)

DevOps Algorithm
Developers
Gateway

Business Decisions
OPC
MQTT
AMQP

End Users
Blob, Data lake

12
Value companies obtain from partnering with us

Carnegie Wave Energy

13
We think differently about AI

Gartner recognizes MathWorks as a Visionary in its January 2019

Magic Quadrant for Data Science and Machine Learning Platforms
▪ Focus on asset-intensive industries
▪ Operationalize to enterprise IT/OT and edge systems
▪ Let domain experts do Data Science themselves

MathWorks is recognized as a 2019 Gartner

Peer Insights Customers’ Choice for Data
Science and Machine Learning Platforms

“Excellent technical package with excellent support structure”

Lead AI Architect in the
Manufacturing Industry

Disclaimer: Gartner does not endorse any vendor, product or service depicted in its research publications, and does not advise technology users to select only those vendors with
the highest ratings or other designation. Gartner research publications consist of the opinions of Gartner’s research organization and should not be construed as statements of fact.
Gartner disclaims all warranties, express or implied, with respect to this research, including any warranties of merchantability or fitness for a particular purpose. 14
 Tools Enabling Power System Application Development
Simulink Test Simulink Coverage
V&V
Polyspace

Embedded Coder Parallel Server Simulink Real-Time Connect/

Production Server Simulink Coder Deploy
Toolbox

Computer Vision Toolbox Deep Learning Toolbox

Text Analytics Toolbox Machine Learning Toolbox Algorithm

Development
Signal Processing Toolbox Optimization Toolbox

Image Processing Toolbox Predictive Maintenance Toolbox

Simscape Electrical Control System Toolbox

Platform
MATLAB Simulink Stateflow Simscape
15
The Landscape – Power Generation

MathWorks Confidential 16
 Hydro-Québec Models Wind Power Plant
Performance

Challenge
Plan the integration of new wind farms into the power
Turbines on a wind farm.
system, predict power output, and ensure safe, reliable
operation
Solution “Accurate modeling is essential not
Use MathWorks products to simulate individual wind only for planning investments but
turbines and wind farms and to generate C code for also to detect situations that can
multiprocessor simulation of entire power systems
cause an outage. With MathWorks
Results tools, we can simulate power
▪ Simulation speed increased to real time electronics, mechanics, and control
▪ Equipment needs accurately predicted systems in one environment, and
▪ Dynamic simulations enabled
our models respond like the turbines
we have in the field.”
Richard Gagnon
Hydro-Québec
Link to user story

17
 Horizon Wind Energy (EDP Renewables)
Develops Revenue Forecasting and Risk
Analysis Tools for Wind Farms

Challenge
Develop revenue forecasts and quantify risk for wind Horizon Wind Energy’s Wheat Field Wind Farm.

farms across multiple geographic locations

Solution “The tools that we developed with
Use MATLAB and MATLAB Compiler to develop and MATLAB and MATLAB Compiler are
deploy an automated production system that analyzes
much more reliable, scalable, and
historical, current, and forward-looking price and wind-
level data maintainable than our spreadsheet-
based approach. We can sleep at
Results
night because we know the tools will
▪ Core process automated
▪ Standalone program seamlessly integrated with work, we can add new capabilities
enterprise IT infrastructure and data inputs, and we can update
▪ Risk management improved, saving millions of the production system without getting
dollars IT involved.”
Manuel Arancibia
Horizon Wind Energy
Link to user story

18
Built-in Renewable Energy Models with Simscape Electrical
Wind, Solar, Energy Storage, etc.

19
Grid Oscillation Characterization – Prony Analysis

Impulse Response Behavior during Grid Events

• Grid Event – Typically faults (power lines, storms, etc.) that cause instant grid change
• Grid Oscillations – Generation systems/controls share loads and can oscillate with events

Goal: Fast identification of grid events to identify any unstable oscillations modes 20
Multi-Event Identification of Damping

Signal Frequency Damping

Number (Hz) Ratio
Event1 0.050 0.400
Event1 0.250 0.050
Event2 0.025 0.500
Event2 0.380 0.010
Event3 0.076 0.800
Event3 0.300 0.020
Event4 0.038 0.300
Event4 0.250 -0.005

21
Energy/Price Forecasting using AI

22
 RWE Develops and Deploys an Automated
System for Natural Gas and Power Trading
and Risk Management

Challenge
Automate business processes for quoting gas contracts
and hedging against price fluctuations
RWE headquarters in Essen.
Solution
Engage MathWorks Consulting to develop and deploy
to a production environment an automated pricing and “MathWorks consultants were well-
risk management system that fits within the company’s qualified, professional, and fast. They
existing IT infrastructure understood not only the technical
Results issues but also the business goals,
▪ Models created in minutes, not weeks which is essential when working on a
▪ 100% accurate results delivered core business system. We got more
▪ Technical expertise applied to core business goals than we expected from MathWorks
Consulting.”
Dr. Norbert Tönder
RWE
Link to user story

23
 Gas Natural Fenosa Predicts Energy Supply
and Demand Using MathWorks Tools

Challenge
Maximize margins in energy trading by predicting Portomouros hydroelectric dam.
available supply and peak demand
Solution
Use MathWorks tools to build and optimize models “Because we need to rapidly respond
that incorporate historical data, weather forecasts, to shifting production constraints and
and regulatory rules changing demands, we cannot depend
Results on closed or proprietary solutions.
▪ Response time reduced by months With MathWorks tools we get more
▪ Productivity doubled accurate results — and we have the
▪ Program maintenance simplified flexibility to develop, update, and
optimize our models in response to
changing needs.”
Angel Caballero
Gas Natural Fenosa
Link to user story

24
The Landscape – Power Transmission

25
 Alstom Grid Develops High-Voltage Direct
Current Transmission Control System
Using Model-Based Design

Challenge
Accelerate control system development for high-voltage
direct current voltage source converters Alstom Grid’s HVDC demonstrator system
with power converter modules. The
Solution improved controllability of the VSC in this
system makes it well-suited for smart grid
Use Model-Based Design to model, simulate, verify, and applications.
generate code and documentation for the control and
protection systems “Using Model-Based Design we
Results developed a complex control system
▪ Quantifiable process improvements in significantly less time than our
▪ Rapid integration with power system simulation traditional process would have
software required. We eliminated months of
▪ Protection systems implemented in one week hand-coding by generating code from
our models, and we used simulations
to enable early design verification.”
Anthony Totterdell
Alstom Grid
Link to user story

26
 Transpower Ensures Reliability of New
Zealand National Grid with Reserve
Management Tool

Challenge
Calculate the amount of reserve power needed to
ensure New Zealand’s national grid can continue to
operate if a generator fails Transmission lines near Transpower’s
Benmore substation.
Solution
Use Simulink to run simulations of the entire grid,
including generators, loads, and HVDC links, every 30 “We record frequencies on the
minutes
grid, inject them into our Simulink
Results model, and compare the simulation
▪ Critical updates rapidly implemented results to the actual system
▪ Simulations verified using real data response. With Simulink we can
▪ Updates made in-house continually calibrate and improve
our model, and ultimately improve
the accuracy of our reserve
estimates.”
Heidi Heath
Link to user story Transpower

27
28
The Landscape – Working with PMU Data

29
Checking Compliance with Measured PMU Data
Voltage Grid Codes IEEE 1547

VPOC

Cascading Trips

Frequency Grid Codes

FPOC

MathWorks Confidential 30
MathWorks Confidential 31
“Replay” Field Measurements to Compare Model to Data
Use Field Voltage/Frequency and Match Plant P/Q Response

Voltage and Frequency Replay Block

Replaced “Simulated” Grid to Leverage Field Data
32
Assessing IEEE 1547 Compliance with PMU Data

33
 Protection/ Hardware Grid Code
Control Implementation Compliance Checks
Implementation

S C
Controller Development
Environment Protection System
or RTAC Controller

Prototype/
Actual System

34
 Protection/ Hardware Real-Time
Control Implementation Simulation
Implementation

S C
Controller Development Equipment Simulation
Environment Protection System
or RTAC Controller

35
The Landscape – Power Distribution

MathWorks Confidential 36
 Sandia National Laboratories Simulates
Hawaii Microgrid and Photovoltaic Systems

Challenge
Evaluate the battery capacity and control systems
required for reliable operation of a new solar power
generation facility
An array of solar panels.
Solution
Use Simulink and SimPowerSystems to model and
simulate microgrid distribution systems with photovoltaic “MATLAB, Simulink, and
sources
SimPowerSystems enable
Results mechanical, power, and controls
▪ Model development time cut by 80% engineers to work together using
▪ Costs reduced through battery right-sizing
the same tools, which helps in a
▪ Simulation accuracy verified with real data
multidisciplinary environment
like ours.”
Ben Schenkman
Sandia

Link to user story

37
 Shanghai Electric Builds and Deploys Cost-Saving
Enterprise Software for Planning and Designing
Distributed Energy Systems

Challenge
Shanghai Electric engineers wanted to model a broad range of
components, including different loads, the grid, batteries and other
energy storage units, and Develop web-accessible software for
planning and designing distributed energy systems
Solution
“My team’s expertise is in energy modeling or
Shanghai Electric used MATLAB, Simulink and MATLAB Production
algorithm development, not in deploying
Server to develop its distributed energy system planning and design
software into production. MATLAB saved us
platform.
months of development time on the models and
algorithms, and then made it easy to deploy
them as part of a stable, reliable web application
without recoding.”
Yunjiao Gu
Shanghai Electric

Plot of actual energy production by type

of power generation system.

38
The Landscape – Electric Vehicles

MathWorks Confidential 39
MathWorks Confidential 40
 Optimizing Energy Management System
Denso
Generated and consumed power

¥
Electricity prices
Battery
command

EMS Controller Home

Predicted Stored energy

vehicle use

Model predictive control Simscape Electrical

Mixed integer linear programming
© 2018 The MathWorks, Inc.
The Landscape – Analyzing Smart Meter Data

MathWorks Confidential 42
43
The Landscape –Substation Automation & Controls

MathWorks Confidential 44
46
Asset Health Monitoring & Predictive Maintenance

Courtesy of General Cable

47
 Approach to Predictive Maintenance

Methods based on physical laws Method based on sensor data

First Principles Modeling Data-driven modeling

White-box Gray-box Black-box

Model-driven Data-driven

• Supplement data that is difficult to • Use the data obtained from the
measure equipment
• Lead to a highly accurate predictive • Create predictive models using
model statistical methods

48
 Albatroz Engineering Develops Automated,
Real-Time Power Line Inspection System
with MathWorks Tools

Challenge
Develop an automated, real-time system for
inspecting and maintaining overhead power lines
Helicopter view of a 60K overhead line
Solution crossing a densely forested area.

Use MathWorks tools to acquire, visualize,

synchronize, and programmatically interpret data
from multiple sources in real time “During the first test flight, we
Results discovered a bug in one of our
▪ Development time cut by at least 50% algorithms. We opened up MATLAB
▪ System accuracy exceeds 99% and solved the problem completely, in
▪ Inspection costs reduced by 90% flight. In 10 minutes our data
acquisition and algorithms were
working fine. It was very impressive.”
Alberto Vale
Albatroz Engineering

Link to user story

49
 Predictive Maintenance of Underground Cables

Courtesy of General Cable

▪ Convolutional Neural Network (CNN).

▪ Risk assessment of PD sites.
▪ Categorizing PD sites by risk, based on image data.
Data Analysis – ▪ Partial Discharge Severity (Risk).
• Interpreting PD signal waveforms ▪ Defect type classification with CNN.
• Time Series Data
50
 Predictive Maintenance of Wind Turbines

Wind speed
Measurements
wind_speed
pitch_angle
turbine_speed
Speed, torque
turbine_torque
Pitch angle
shaft_speed
Electrical Power
shaft_torque
electromagnetic_torque
P_gen
Q_gen Categories
Electromagnetic torque
P_out
Q_out
X 20 Normal
Increased_Friction
V_DC Generator_Rotor_Fault
Qref Feedback_Error

51
Building Larger Models

52
Classification Learner

53
 Vegetation Management using Deep Learning
Vegetation Management

Managing vegetation is one of the highest priorities

due to large customer impact and O&M spend
Vegetation related outages has significant customer
impacts, accounting for more than 30% of all outages

Risk identification using low

Innovative Analytics for Smart cost satellite imagery
Vegetation Management

• More efficient data-driven work planning

• Vegetation risk management & smart
contracting
• Reduced vegetation related outages
54
 Platform Architectural Diagram – Operationalizing Analytics

MATLAB Compiler MATLAB Compiler SDK MATLAB Coder Simulink

GPU HDL PLC Coder

Standalone MATLAB
Application
Excel
Add-in
Hadoop Spark C/C++ Java .NET Python
Production Coder Coder
Server

C, C++
CUDA HDL PLC

Desktop Users Enterprise IT Systems Embedded Systems (Including Edge Devices)

• Microcontrollers
• NVIDIA GPUs
• DSP chips
• FPGAs
• ARM-based
• Low-cost:
o Arduino
o Raspberry Pi
o BeagleBone
Denotes product dependency. 55
Achieve results faster with MathWorks Consulting Services
Having deep product knowledge and broad technical experience, we guide your team to
apply best practices to your development projects.

Transparent Approach Customized Engagements Return on Investment

Your team will have full We’ll work with you to develop a We’ll accelerate your learning
access to all our work customized project plan aligned curve, reducing development
throughout your project to your business goals time and cost while improving
quality and team collaboration
56
Advance your skills with MATLAB and Simulink courses
Get started for free with MATLAB Onramp, then build your skills with our self-paced
trainings and instructor-led courses.

Flexible Training Expert Trainers Proven Methods

Over 50 courses available Our instructors possess Participants benefit from
around the world, at your unparalleled knowledge of real-world examples and
work site, or on the web MathWorks products individualized attention
57
 MathWorks is your Engineering Partner!

The Platform Your People Our Expertise

MATLAB, Simulink, and over Helping you build an agile From onboarding and
100 add-on products for workforce today and implementation to solving
specialized applications preparing tomorrow’s advanced engineering
engineers challenges
58
Cómo predecir la demanda energética con
Inteligencia Artificial

Lucas García, PhD

Senior Application Engineer

© 2020 The MathWorks, Inc.

1
What is Data Analytics?
Turn large volumes of complex data into actionable information

Descriptive • What happened?

Diagnostics • Why did it happen?

Data Decisions
Predictive • What will happen?

Prescriptive • What should be done?

2
Sub-Applications of Energy Forecasting

Load Forecasting: Improve system operation, planning, and

market participation

Price Forecasting: Improve market participation and

financial hedging

Wind Forecasting: Understand uncertainty and risk for

system operation, planning, and market participation

Solar Forecasting: Understand uncertainty and risk for

system operation, planning, and market participation

3
Users of Energy Forecasts

▪ ISO
▪ Utilities
▪ Transmission
▪ Distribution
▪ Retail
▪ Traders

4
Renewables and Distributed Power Generation Increase Complexity

Source: Osceola Energy, The Case for Distributed Power Generation, June 30, 2015:
osceolaenergy.com/distributed-power-generation/
5
Challenges with Existing Forecasting Approaches

Not Accurate Enough

▪ Limited predictive modeling techniques in current tools
▪ Difficult to incorporate additional data sets into the model

Model is Outsourced
▪ Takes a long time to make updates
▪ Unable to keep pace with changing operational needs and changing
regulatory requirements

Current Tools Do Not Scale

▪ Spreadsheet-based tools not designed for complex calculations or workflow
automation, causing significant waste
6
Forecast Horizons

Short-Term: A few minutes to a few days ahead.

Important for day-to-day market operations.

Mid-Term: A few days to a few months ahead.

Important for balance sheet calculations, risk management, and
derivatives pricing.

Long-Term: Months, quarters, or years ahead.

Important for investment profitability analysis and planning.

7
Short-Term Forecasting Techniques1

Short-Term: A few minutes to a few days ahead.

Important for day-to-day market operations.

▪ Time Series Models

▪ ARMA, ARIMA, SARIMA
▪ Support Vector Machines
▪ Artificial Neural Networks
▪ etc.

Day-Ahead Forecasts at intervals of {5 mins, 15 mins, 30 mins, 1 hr}

1. An overview of energy demand forecasting methods published in 2005–2015 8

Data Analytics Workflow
Access and Explore Develop Predictive Integrate Analytics with
Preprocess Data
Data Models Systems

Files Working with Model Creation e.g. Desktop Apps

Messy Data Machine Learning

Databases Data Reduction/ Parameter Enterprise Scale

Transformation Optimization Systems

Sensors Feature Model Embedded Devices

Extraction Validation and Hardware

9
Case Study: Day-Ahead Energy Load Forecasting
Goal:
▪ Implement a tool for easy and accurate computation of day-ahead system load forecast

Requirements:
▪ Acquire and clean data from multiple
sources
▪ Accurate predictive model
▪ Easily deploy to production environment

10
The Data
NYISO Energy Load Data National Climatic Data Center Weather Data
mis.nyiso.com/public/ https://www.ncdc.noaa.gov/orders/qclcd/

11
Data Analytics Workflow
Access and Explore Develop Predictive Integrate Analytics with
Preprocess Data
Data Models Systems

Files Working with Model Creation e.g. Desktop Apps

Messy Data Machine Learning

Databases Data Reduction/ Parameter Enterprise Scale

Transformation Optimization Systems
1

Sensors Feature Model Embedded Devices

Extraction Validation and Hardware

13
Data Analytics Workflow
Access and Explore
Preprocess Data
Data

Files Business and Transactional Data Working with

Messy Data
▪ Repositories – SQL, NoSQL, etc.
▪ File I/O – Text, Spreadsheet, etc.
▪ Web Sources – RESTful, JSON, etc.
Databases Data Reduction/
Engineering, Scientific and Field Data Transformation

▪ Real-Time Sources – Sensors, GPS, etc.

▪ File I/O – Image, Audio, etc.

Sensors ▪ Communication Protocols – OPC (OLE for Feature

Extraction
Process Control), CAN (Controller Area
Network), etc.

14
Data Analytics Workflow
Access and Explore
Preprocess Data
Data

Files Challenges Working with

Messy Data
▪ Data aggregation
– Different sources (files, web, etc.)
– Different types (images, text, audio, etc.)

▪ Data clean up
Databases – Poorly formatted files Data Reduction/
– Irregularly sampled data Transformation
– Redundant data, outliers, missing data etc.

▪ Data specific processing

– Signals: Smoothing, resampling, denoising,
Sensors Wavelet transforms, etc. Feature
– Images: Image registration, morphological Extraction
filtering, deblurring, etc.

▪ Dealing with out of memory data (big data)

15
Access data in many formats from many locations

▪ Type
Simulation

▪ Structure
Databases

▪ Location
Images

MDF Files

Custom

16
Live Editor tasks for preprocessing data and manipulating tables and timetables

▪ Tasks are apps that can be included in scripts

▪ Preprocessing tasks allow you to:

– Interactively explore parameters and options
– Preview results based on those parameters and options

▪ Additional tasks for interactively

manipulating tables and timetables

▪ Automatically generate the

corresponding MATLAB code

▪ Save the task as part of the script

for subsequent use by others

17
tall Arrays

▪ New data type designed for data that doesn’t fit into memory

▪ Lots of observations (hence “tall”)

▪ Looks like a normal MATLAB array

– Supports numeric types, tables, datetimes, strings, etc…
– Supports several hundred functions for basic math, stats, indexing, etc.
– Statistics and Machine Learning Toolbox support
(clustering, classification, etc.), Database Toolbox

18
tall Arrays Single
tall array Single
Machine Machine
Memory Process Memory

▪ Automatically breaks data up into

small “chunks” that fit in memory

▪ Tall arrays scan through the

dataset one “chunk” at a time

▪ Processing code for tall arrays is

the same as ordinary arrays

19
tall Arrays Single
tall array Single
Machine Machine
Memory Process Memory

▪ With Parallel Computing Toolbox,

process several “chunks” at once Single
Machine
Process Memory

▪ Can scale up to clusters with

MATLAB Parallel Server Single
Cluster of Machine
Machines Process Memory
Memory
▪ Support for Spark and Hadoop
Single
Machine
Process Memory

20
Data Analytics Workflow
Access and Explore MATLAB Analytics work 1 Preprocess Data
Data with business and
engineering data

Files Working with

Messy Data

Databases Data Reduction/

Files Databases Transformation

Signals Images

Sensors ▪ Point and click tools to access ▪ Built-in algorithms for data Feature
Extraction
variety of data sources preprocessing including sensor,

▪ High-performance environment image, audio, video and other

for big data real-time data

21
Data Analytics Workflow
Access and Explore Develop Predictive Integrate Analytics with
Preprocess Data
Data Models Systems

Files Working with Model Creation e.g. Desktop Apps

Messy Data Machine Learning

Databases Data Reduction/ Parameter Enterprise Scale

Transformation Optimization Systems
1 2

Sensors Feature Model Embedded Devices

Extraction Validation and Hardware

22
Data Analytics Workflow
Develop Predictive
Preprocess Data
Models

Working with Challenges Model Creation e.g.

Messy Data Machine Learning
▪ Lack of data science expertise

▪ Feature Extraction – How to transform

data to best represent the system?
– Requires subject matter expertise
Data Reduction/ – No right way of designing features Parameter
Transformation Optimization
▪ Feature Selection – What attributes or
subset of data to use?
– Entails a lot of iteration – Trial and error
– Difficult to evaluate features

Feature ▪ Model Development Model

Extraction Validation
– Many different models
– Model Validation and Tuning

▪ Time required to conduct the analysis

23
Predictive Maintenance Toolbox
Diagnostic Feature Designer App

▪ Extract, visualize, and rank

features from sensor data

▪ Use both statistical and

dynamic modeling methods

▪ Work with out-of-memory data

▪ Explore and discover

techniques without writing
MATLAB code

24
Machine Learning “I would have never attempted machine
learning if this app was not available.”

MATLAB makes machine learning easy and accessible

for everyone, even if you’re not an expert

▪ Use apps to interactively explore data, choose

algorithms to train and validate models, and
compare results
– Classification Learner app
– Regression Learner app

▪ Automated Machine Learning (AutoML)

– Automatically select the best model
and associated hyperparameters for
classification (fitcauto)

25
Deep Network Designer app to build, visualize, and edit Deep
Learning Networks

26
Data Analytics Workflow
MATLAB enables
2 Develop Predictive
Preprocess Data
domain experts to Models
do Data Science

Working with Apps Language Model Creation e.g.

Messy Data Machine Learning

Data Reduction/ Parameter

Transformation Optimization

Feature ▪ Easy to use apps ▪ Automatic MATLAB code Model

Extraction Validation
▪ Wide breadth of tools to facilitate generation
domain specific analysis ▪ High speed processing of large
▪ Examples/videos to get started data sets

28
Data Analytics Workflow
Access and Explore Develop Predictive Integrate Analytics with
Preprocess Data
Data Models Systems

Files Working with Model Creation e.g. Desktop Apps

Messy Data Machine Learning

Databases Data Reduction/ Parameter Enterprise Scale

Transformation Optimization Systems
1 2 3

Sensors Feature Model Embedded Devices

Extraction Validation and Hardware

29
Data Analytics Workflow
Develop Predictive Integrate Analytics with
Models Systems

Model Creation e.g. Challenges Desktop Apps

Machine Learning
▪ End user: Operators, Analysts,
Administrative Staff, customers etc.

▪ Different target platforms:

– Cluster or Cloud environment
Parameter – Standalone desktop applications Enterprise Scale
Optimization Systems
– Server based Web and enterprise systems
– Embedded hardware

▪ Different Interfaces: C/C++, Java, Python,

.NET etc.
Model Embedded Devices
Validation ▪ Need to translate analytics to production and Hardware
environment

30
Integrate analytics with systems 3
MATLAB Analytics
run anywhere

Embedded Hardware Enterprise Systems

Standalone
Application Excel Hadoop/ MATLAB
C, C++ CUDA HDL PLC or C/C++ Java ++ Python .NET Production
Add-in Spark
Web App Server

MATLAB
Runtime

32
MATLAB Production Server is an application server that publishes
MATLAB code as APIs that can be called by other applications
Analytics Development

MATLAB MATLAB
Compiler SDK Deploy

Package Code / test Access

Enterprise
Application
Integrate
MATLAB Production Server
Data sources /
Worker processes
applications
<>
Request
Broker Mobile / Web
Application

3rd party
Scale and secure
dashboard

33
Deployed Analytics

Web MATLAB MATLAB

Application Production Desktop
Server Server
Train in
MATLAB MATLAB
Production Server

Request Broker
Predictive
Models

Weather
CTF
Data

Energy
Data
34
How about integration PI System

with OSIsoft PI System?

PI Asset
Algorithm Framework Production
Development Deployment

AF SDK
MATLAB Production
Server Interface for
OSIsoft PI System
Configuration Tool

MATLAB
MATLAB Interface
Production Server MATLAB
MATLAB for OSIsoft PI
Interface for
System
OSIsoft PI System
Production Server

Code Deployment

35
Enterprise-Scale Data Analytics

Data Analytics Business System

Databases Dashboards
MATLAB
Production Server
Cosmos DB

Cloud Storage Azure Web

Blob Request
Broker

MATLAB
Runtime
Streaming AWS Custom Apps
Kinesis
Azure
PI System IoT Hub

MATLAB Parallel Server

Platform
36
 is a Leader in the Gartner Magic
Quadrant for 2020 Data Science
and Machine Learning Platforms

*Gartner Magic Quadrant for Data Science and Machine Learning Platforms, Peter Krensky, Erick Brethenoux, Jim Hare, Carlie Idoine, Alexander Linden, Svetlana Sicular, 11 February 2020 .

This graphic was published by Gartner, Inc. as part of a larger research document and should be evaluated in the context of the entire document. The Gartner document is available upon request from MathWorks.
Gartner does not endorse any vendor, product or service depicted in its research publications, and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner research
publications consist of the opinions of Gartner’s research organization and should not be construed as statements of fact. Gartner disclaims all warranties, express or implied, with respect to this research, including any
warranties of merchantability or fitness for a particular purpose.
37
EDP Renováveis Improves Wind Energy
Forecasting with Machine Learning
Daniel Cabezón, EDP Renováveis
“ We were able to significantly reduce
deviation costs between the wind
power forecast and the real production,
EDP Renováveis is a global leader in the renewable energy sector and the world’s resulting in millions of euros saved “
fourth-largest wind energy producer. As wind farms are installed in increasingly
per year.
complex locations, resource assessment and energy production becomes more and
more challenging.
To overcome these challenges, EDP Renováveis used MATLAB® to find patterns
between meteorological data and measurements recorded in the wind turbine. They
developed a suite of applications to apply machine learning to these patterns that
enabled them reduce the error in wind power forecast, estimate potential losses due
to ice in cold climates, and study the effect of weather on wind and production in
large regions of the planet.

Advantages of using MATLAB:

• Access a wide collection of algorithms for machine learning
• Estimate energy lost due to ice accretion in blades to find wind farm locations
• Safely install of wind turbines in locations prone to strong storms
• Save millions of euros by reducing deviation costs

Presented at MATLAB EXPO Spain 2019

Explore MATLAB and Simulink for the utilities and energy industry
Watch video (22:27 – in Spanish) 38
How can I continue learning?

Onramp Courses Continue building your skills:

▪ Free 2-hours each interactive tutorials ▪ MATLAB Fundamentals

▪ Statistical Methods in MATLAB

▪ Signal Preprocessing and Feature
Extraction for Data Analytics with
MATLAB
▪ Machine Learning with MATLAB

▪ Deep Learning with MATLAB

For more details visit mathworks.com/training 39

Key Takeaways

MATLAB Analytics work 1

with business and ▪ Utilize all of your data
engineering data

2
MATLAB enables ▪ Apply advanced analytics techniques
domain experts to do
Data Science using Machine Learning and Deep
Learning

MATLAB Analytics
3
run anywhere ▪ Operationalize analytics to enterprise
systems and embedded devices

40
Estudios de Sistemas de Energía mediante Gemelo
Digital

María Fernández
Application Engineer

© 2020 The MathWorks, Inc.

1
Agenda

▪ Modelizado de sistemas de energías renovables y almacenamiento de

energía

▪ Validación del modelo de planta mediante la reproducción de datos de

campo

▪ Integración en red y Códigos de Red

▪ Simulaciones en Tiempo Real – Hardware-in-the-Loop (HIL)

2
Agenda

▪ Modelizado de sistemas de energías renovables y almacenamiento de

energía

▪ Validación del modelo de planta mediante la reproducción de datos de

campo

▪ Integración en red y Códigos de Red

▪ Simulaciones en Tiempo Real – Hardware-in-the-Loop (HIL)

3
Example Solar Inverter Simulation – For Equipment Design
Inverter Designers Leverage Proprietary Control

4
Model Using Fundamental Approaches

First Principles
Simulink
Physical Components
Simscape
Complete Components
Simscape Electrical
5
Poll no. 1

▪ ¿Qué herramientas usas actualmente para el modelizado?

a) Simulink y Simscape exclusivamente
b) Co-simulación Simulink + Otras
c) Otras exclusivamente (indica cuál en el chat)
d) No modelizo mi sistema

6
 Simscape Electrical Simulation Modes
Different Fidelities for Different Tasks
System-Level Behavior Component Validation Component Design
Electrical Determine power requirements Stay within design envelope Losses during switching events
Evaluate system-level response Effect of switching dynamics Analyze & predict fault behavior
Measure efficiency

System AC Power AC Converter

D D D
AC-DC DC-AC
Power Motor Pump
Converter Converter
D D D
DC Link

Power

Hours Minutes Microseconds 7

Agenda

▪ Modelizado de sistemas de energías renovables y almacenamiento de

energía

▪ Validación del modelo de planta mediante la reproducción de datos de

campo

▪ Integración en red y Códigos de Red

▪ Simulaciones en Tiempo Real – Hardware-in-the-Loop (HIL)

9
10
What is Model Validation?
Renewables Field Equipment Standardized Models of Equipment
ALTERNATE
VOEL OEL INPUTS VOEL

ALTERNATE
UEL INPUTS
VUEL VUEL
VP
VRMAX
VS
VRMAX/KA
KA VR
_ LV 1
+_ HV   EFD
+ KI sKD GATE + sTE
GATE 1+sTA
VC  KP + + -
s 1+sTD VEMIN
+ VRMIN
_ +
VREF VRMIN/KA  KE

VX VX=VESE(EFD)

VF
sKF

(1 + sTF)

Active Power Response Reactive Power Response

Measured Field Data

Simulated Response
11
Link to Reliability Guideline document

12
Power Plant Model Validation Workflow

1 2 3 4 5
Develop Power Model Validation Leverage
Access Test Data Preprocess Data
System Model for Power Systems Calibrated Models

Files Format Data for Build Renewable Manual Tuning of Export Parameters
Easy Import Model Parameters for Other Tools

Databases Clean up Voltage/Frequency Setup Parameter Regulatory

Noisy Data Replay for Data Estimation Compliance

PMUs Import Current Setup and Run Parameters for Digital Twin of
Parameters Loadflow Automated Tuning Renewable System

13
Central Station Solar Plant One-line Diagram and Controls
For Grid-scale System-Studies and Validation

Pad-mounted
Transformer Interconnection
Transmission Line

PV System Station
Equivalent Transformer

Plant-level Reactive
Compensation

System-Level Study
14
 Central Station Solar Plant One-line Diagram and Controls
IEEE Standardized Controls for System-Studies and Validation

Plant Electrical Grid

Controller Controller Connection

Plant Level Qref Iq

Q Control
V/Q Control Current
Setpoints Generator
Limit
Model
Pref Logic Id
Plant Level
P Control
P Control

REPC_A REEC_A REGC_A

15
Central Station Solar Plant in Simscape Electrical
IEEE Standardized Controls for System-Studies and Validation
Plant Mode

Grid PCC

Solar Farm Controls/Dynamics

16
Poll no. 2

▪ ¿Estás interesado en crear un Gemelo Digital del sistema?

a) Sí, para estudiar degradación
b) Sí, para Mantenimiento Predictivo
c) Sí, para simulación de fallos
d) Todas las anteriores
e) No, solo quiero validar mi sistema
f) No

17
Parameter Estimation for Multiple Grid Tests

18
Agenda

▪ Modelizado de sistemas de energías renovables y almacenamiento de

energía

▪ Validación del modelo de planta mediante la reproducción de datos de

campo

▪ Integración en red y Códigos de Red

▪ Simulaciones en Tiempo Real – Hardware-in-the-Loop (HIL)

19
Renewable and Distributed Resource Grid Integration

20
What are Grid Codes?
Voltage Grid Codes

VPOC

POC Cascading Trips

Frequency Grid Codes

FPOC

21
Common Voltage/Frequency Grid Codes

22
Assessing Grid Code Compliance with PMU Data

23
Assessing Controls + Protection with PMU Data Replay

24
Assess Grid Codes in Parallel

25
Agenda

▪ Modelizado de sistemas de energías renovables y almacenamiento de

energía

▪ Validación del modelo de planta mediante la reproducción de datos de

campo

▪ Integración en red y Códigos de Red

▪ Simulaciones en Tiempo Real – Hardware-in-the-Loop (HIL)

26
 Protection/ Hardware Grid Code
Control Implementation Compliance Checks
Implementation

S C
Controller Development
Environment Protection System
or RTAC Controller

Prototype/
Actual System

27
 Protection/ Hardware Grid Code
Control Implementation Compliance Checks
Implementation

S C
Controller Development Equipment Simulation
Environment Protection System
or RTAC Controller

28
Testing Grid Controls with Hardware-in-the-Loop
S C

29
Poll no. 3

▪ ¿En qué tipo de aplicaciones estás interesada/o?

a) Gemelo Digital del sistema real
b) Validación de la planta de acuerdo a mod 26/27 (u otros)
c) Estudios de Integración en Red y/o códigos de red
d) Simulaciones en Tiempo Real de red y/o sistema

30
Additional Material

▪ MATLAB and Simulink for the Utilities and Energy Industry

▪ White Paper: PPMV with MATLAB and Simulink

▪ Videos:
– Simulation-Based Grid Code Compliance
– Energy Storage Optimization
– Renewable Grid Integration Studies with Simscape Electrical
– Renewable Energy Power Plant Model Validation
– Performing Power System Studies Video Series
– Modeling, Simulating, and Generating Code for a Solar Inverter

32
Training

▪ Simulink On-Ramp (free)

▪ Stateflow On-Ramp (free)

▪ Simulink for System and Algorithm Modeling

▪ Simulink Model Management and Architecture
▪ Control System Design with MATLAB and Simulink

▪ Modeling Electrical Power Systems with Simscape

▪ Simulink-Based Testing with Simulink

▪ Real-Time Testing with Simulink Real-Time and Speedgoat Hardware

33