Power Transformer Simulation Laboratory for Proactive ...

© Copyr i gh t 2014 O SIs o f t , LLC .

Presented by

Power Transformer

Simulation

Laboratory for

Proactive

Maintenance II

Nicolas Di Gaetano

& Luc Vouligny

© Copyr i gh t 2014 O SIs o f t , LLC .

Agenda

• An overview of Hydro-Québec

• Context of the project

• Objectives of the project

• Transformer monitoring

• Conclusion & Future Work

© Copyr i gh t 2014 O SIs o f t , LLC .

Snapshot

• Hydro-Québec is among the largest power

generator in North America

– 98% renewable energy

• Hydro-Québec is among the largest power

transmission companies in North America

• Hydro-Québec is the largest electric utility in

Canada

3

© Copyr i gh t 2014 O SIs o f t , LLC .

About Hydro-Québec

4

© Copyr i gh t 2014 O SIs o f t , LLC .

About Hydro-Québec’s Research Institute

(IREQ)

> Largest electric utility research centre in North America

• 500 employees of which half are researchers

• $100M invested each year on 100 projets

• 1000 patents

5

© Copyr i gh t 2014 O SIs o f t , LLC .

Hydro-Québec TransÉnergie

• Transmission assets: $17.6 B

• 33,630 km of power transmission

lines

– Including 11,422 km of 735 kV lines

• 514 transmission substations

• Annual investment: $1,3 B

6

© Copyr i gh t 2014 O SIs o f t , LLC .

17 interconnections with

neighboring markets

(> 6,000 MW)

7

© Copyr i gh t 2014 O SIs o f t , LLC .

Context of the project

• Network: reliable and available

• Data: high-quality, value-added and just-in-time

• Decisions: appropriate, timely

• Power transformers are critical assets of the network

– Average age: 33 years old

– Their life expectancy: 40-50 years

• Transition from systematic to proactive maintenance

8

© Copyr i gh t 2014 O SIs o f t , LLC .

Objectives of the project

• Maximize asset life

• Reduce the Risk & Cost of unexpected failure

• Drive maintenance and inspection by asset

condition

• Awareness of its condition and performance

• To maximize system availability

9

© Copyr i gh t 2014 O SIs o f t , LLC .

Transformer monitoring

Real-time monitoring

10

2 3 41 5 6MEET Project

Sensors,

IEDs,

Transducers

Communication

Infrastructure

Data Server

(PI System)Computational Intelligence

Diagnostics & PrognosticsWeb Interface

Sustainability

& Maintenance

Action

© Copyr i gh t 2014 O SIs o f t , LLC .

Transformer monitoring: From past to future

• Use remote monitoring for all growth and asset sustainment projects– More than 500 substations over a 15-year timeframe (2009–2025)

• Objective: 30 to 40 substations a year

– Monitoring of 240 strategic transformers by 2015

• Gas and moisture

• Temperature

• Bushing and tap changer monitors (future deployment) – OSIsoft PI System

• Automatic addition of points (PI APS, PI GenericCSV_APS connector)• Maximizes use of PI AF templates

• PI AF SDK

200820072006 2009

IT infrastructure

2010 2011 2025

Business

case

11

© Copyr i gh t 2014 O SIs o f t , LLC .

PI Webparts: Displays

12

© Copyr i gh t 2014 O SIs o f t , LLC .

PI Webparts: Displays (continued)

13

© Copyr i gh t 2014 O SIs o f t , LLC .

PI Webparts: Displays (continued)

14

© Copyr i gh t 2014 O SIs o f t , LLC .

PI Notifications: Alerts

Remote Maintenance Centres

Technical Support

Managers15

© Copyr i gh t 2014 O SIs o f t , LLC .

Transformer monitoring

Bushing

Dissolved

Gas

Thermal

Performance

Tap Changer

Moisture

and Aging

Mechanical state

16

2 3 41 5 6MEET Project

Sensors,

IEDs,

Transducers

Communication

Infrastructure

Data Server

PI SystemComputational Intelligence

Diagnostics & PrognosticsWeb Interface

Sustainability

& Maintenance

Action

© Copyr i gh t 2014 O SIs o f t , LLC .

Acquisition, Estimation & Detection

17

Microsoft StreamInsight

PI Server

Queries

TCP/SI Adapte

r

PI Input

Adapter

PI OutputAdapter

Simulated

data

PI

Points

Real

data

PI to PI

Microsoft StreamInsight

PI Server

Queries

TCP/SI Adapte

r

PI Input

Adapter

PI OutputAdapter

Microsoft StreamInsight

PI Server

Queries

TCP/SI Adapte

r

PI Input

Adapter

PI OutputAdapter

Estimation

data

Asset

maintenance

model

Detection

logic

JMS/SI

Input Conn.

© Copyr i gh t 2014 O SIs o f t , LLC .

Thermal prediction with physical models

• IEC 60076-7 International Standard

– Loading guide for oil-immersed power transformer

• IEEE Clause 7 Non-Linear

• IEEE Clause 7 Linear

• Swift

• Susa

18

© Copyr i gh t 2014 O SIs o f t , LLC .

IEC 60076-7 / Exponential Equations

19

– Load varies as a step function

– Used in the determination of heat

transfer parameters

– Increasing load (κ)

– Decreasing load (κ)

19

© Copyr i gh t 2014 O SIs o f t , LLC .

IEC 60076-7 / Difference EquationsTemporal variables

> κ : Load factor

> θo: Top-Oil temperature

> θa : Ambient temperature

> Dt: Time interval

• 1 minute

Constants:

> k11 : Model constant

• 0,8

> τO : Avg. oil time constant

• 150

> Δ θor : Top oil rise constant

• 50

> R: Load loss ratio

• 8,4

> x: Exponential power of current vs top oil rise

• 0,8

20

- Both the load and the ambient

temperature are time-varying

- Adapted for monitoring purposes

© Copyr i gh t 2014 O SIs o f t , LLC .

Acquisition, Estimation & Detection

21

Microsoft StreamInsight

PI Server

Queries

TCP/SI Adapte

r

PI Input

Adapter

PI OutputAdapter

Simulated

data

PI

Points

Real

data

PI to PI

Microsoft StreamInsight

PI Server

Queries

TCP/SI Adapte

r

PI Input

Adapter

PI OutputAdapter

Microsoft StreamInsight

PI Server

Queries

TCP/SI Adapte

r

PI Input

Adapter

PI OutputAdapter

Estimation

data

Asset

maintenance

model

Detection

logic

Exponential

Equation

Difference

Equation

© Copyr i gh t 2014 O SIs o f t , LLC .

Difference equation performance

Simulation

Estimation

22

© Copyr i gh t 2014 O SIs o f t , LLC .

Real data

• 3 years of data

23

Load from 0 to 1.35 pu

Ambient Temp. from -31°C to 33°C

© Copyr i gh t 2014 O SIs o f t , LLC .

Performance with real data

24

Top-Oil

Ambient

Estimated

Difference

© Copyr i gh t 2014 O SIs o f t , LLC .

Performance over a year

25

© Copyr i gh t 2014 O SIs o f t , LLC .

Results with physical models – 2009

26

© Copyr i gh t 2014 O SIs o f t , LLC .

Results with physical models – 2011

27

© Copyr i gh t 2014 O SIs o f t , LLC .

Results with clustering

28

© Copyr i gh t 2014 O SIs o f t , LLC .

Neural networks

• Neural networks are nonlinear black-box

structures with “interesting” properties

– general architecture

– universal approximator

– non-sensitive to over-parametrization

– have learning algorithm to acquire knowledge from

their environment, using examples

– have recall algorithm to use the learned knowledge

29

© Copyr i gh t 2014 O SIs o f t , LLC .

Basic Regressor Neural Networks

Current Inputs

(Static Networks)

Current Inputs

and Past Outputs

(Dynamic Networks)

30

© Copyr i gh t 2014 O SIs o f t , LLC .

Matlab results for Neural Network

top-oil temperature prediction

31

2011 data test (6 months)– Mean Absolute Error: 0.25%

– Mean Error : 0.1 °C

– Max Error : 1 °C

© Copyr i gh t 2014 O SIs o f t , LLC .

Abnormal behaviour (data cleaning)

AmbT load Actual ANN

9.68 36 19.52 19.49

9.68 36 10.85 10.89

9.68 11739 146.90 15.81

9.68 12339 13.86 16.88

9.56 3 24.86 28.95

9.58 36 10.79 27.59

9.37 3 24.75 15.70

9.38 3 10.70 10.71

9.37 3 10.72 10.69

9.38 3 10.72 10.69

9.37 3 10.67 10.66

© Copyr i gh t 2014 O SIs o f t , LLC .

Neural Network Estimator

33

Microsoft StreamInsight

PI Server

Queries

TCP/SI Adapte

r

PI Input

Adapter

PI OutputAdapter

Simulation

data

PI

Points

Real

data

PI to PI

Microsoft StreamInsight

PI Server

Queries

TCP/SI Adapte

r

PI Input

Adapter

PI OutputAdapter Microsoft StreamInsight

PI Server

Queries

TCP/SI Adapte

r

PI Input

Adapter

PI OutputAdapter

Estimation

data

Detection

logic

.dll

© Copyr i gh t 2014 O SIs o f t , LLC .

Neural Network prediction results with real data

34

© Copyr i gh t 2014 O SIs o f t , LLC .

Future work

• Gradual implementation of models (2014-2018)– Thermal performance

– Dissolved gas

– Moisture and aging

– Tap changer and bushing

– Mechanical state

• Each model are developed with the following steps

35

Specification Development Coding Testing Documentation Implementation

© Copyr i gh t 2014 O SIs o f t , LLC .

Proactive maintenance Advantages

• Transformers– Avoiding major failures

– Avoiding unavailability and captive power

• Equipment– Awareness about equipment condition as it

ages

• Sensors – Detection of malfunctioning equipment

– Improvements on many equipment settings

36

© Copyr i gh t 2014 O SIs o f t , LLC .

- Transformers

– Avoiding major failures

- Equipment

– Awareness about equipment

condition as it ages

Solution Results and Benefits

Power Transformer Simulation Laboratory for

Proactive Maintenance

- Maximization of asset life

- Risk & Cost Reduction of

unexpected failure

- Maintenance and Inspection

driven by asset condition

37

- Set up the infrastructure &

organization needed to ensure

real-time monitoring of

transformer condition

- Real-time data analysis with

predictive model

« Now that we have on-line monitoring possibilities

for maintenance purposes on power transformers,

our simulation laboratory will allow us to elaborate

and test many possibilities of proactive

maintenance models & technics before

implementing them in the field. »

Business Challenge

© Copyr i gh t 2014 O SIs o f t , LLC .

Nicolas Di Gaetano• DiGaetano.Nicolas@hydro.qc.ca

• Fiability Engineer

• TransEnergie, Hydro-Québec

38

Arnaud Zinflou• Zinflou.Arnaud@ireq.ca

• Researcher

• IREQ, Hydro-Québec

Patrick Picher• Picher.Patrick@ireq.ca

• Researcher & project manager

• IREQ, Hydro-Québec

Luc Vouligny• Vouligny.Luc@ireq.ca

• Researcher

• IREQ, Hydro-Québec

© Copyr i gh t 2014 O SIs o f t , LLC .

Brought to you by