©2024 – Seeq Corporation | 1©2024 – Seeq Corporation | 1

Data-Driven Condition Monitoring for Hydro Assets

Dr. Georgia Hunter

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

3

Who am I?

G Hunter| March 2026

Dr. Georgia Hunter Part of a team of engineers that has specialised in analytics/data science

Approach data problems with an engineer/science-backed mindset

Background

What I do at AGL Advanced Analytics Engineer

PhD in Materials Engineering

BMS/HVAC Analytics

Central team that works across all of the AGL energy assets

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

A Saberi | March 2025 5

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

G Hunter| March 2026 6

Hydro – How It Works

River

Dam/Reservoir

Generator

Turbine

Penstock

Main Inlet Valve

Circuit Breaker

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

G Hunter| March 2026 7

Condition Monitoring of Hydro Stations

What are the key components that can fail?

What data do we have for those components?

How can we use the data to monitor the condition of

these components?

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

G Hunter| March 2026 8

Analysis 1: Monitoring of MIV

What are we monitoring?

Previous Monitoring Method

Main Inlet Valve (MIV)

Manual monitoring with a stopwatch

What Data Do We Have

Limit Switches for MIV Open and MIV Closed

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

G Hunter| March 2026 9

Monitoring of MIV MIV Closed Limit Switch

MIV Open Limit Switch Raw Data 1

Create Capsules when MIV

Opening or Closing 2

Calculate Time for MIV to

Open or Close 3

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

G Hunter| March 2026 10

Monitoring of MIV

Define upper and lower

limits for expected timing 4

Contextualise data with

Unit ‘Mode’ 5

Identify when timing is

outside of limits AND unit is in Auto Mode

6

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

G Hunter| March 2026 11

Monitoring of MIV

Steadily increasing opening and closing time

Large step change in timing – general large variation in timing

Examples of condition degradation identified by the analysis

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

G Hunter| March 2026 12

Analysis 2: Monitoring of Sump Pump

What are we monitoring?

Previous Monitoring Method

Leakage in station (i.e., valve leakage)

None

What Data Do We Have

Sump Pumps On/Off switch

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

G Hunter| March 2026 13

Monitoring of Sump Pumps

Raw Data 1

Convert to Numerical Data2

Calculate proportion of

time pump is running on a rolling hourly basis (by taking hourly average of

numerical data every 5 minutes)

3

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

G Hunter| March 2026 14

Monitoring of Sump Pumps

Define Upper Limit for expected

time for sump pump to run 4

Create x-y plot to observe changes in

pump runtime over time relative to power

5

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

G Hunter| March 2026 15

Monitoring of Sump Pumps

Shaft seal replacement

Example of a leakage issue at one of the stations

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

G Hunter| March 2026 16

Analysis 3: Monitoring for Tunnel Collapse

Consequences of Failure

Previous Monitoring Method

Disrupts Power Generation

Requires long and expensive outages to repair

Geotechnical inspections and remote operating

vehicles

Challenges:

• Requires station outage

• Expensive

• Only retrospective (cannot identify a rock collapse as it happens)

17

Monitoring for Tunnel Collapse

𝐾𝑛 = 2𝑔 𝐻𝑠𝑡𝑎𝑡𝑖𝑐 −𝐻𝑛𝑒𝑡

𝑣2

Static Head

(dam level)

Net Head

Velocity (or flow)

Measured

Parameters

Gravity constant

Application of the Model:

• Steady State Only

• Units running at the same time

G Hunter| March 2026

18

Monitoring for Tunnel Collapse

G Hunter| March 2026

Raw Data 1

Identify steady state using 10mins rolling

stdev of head and flow

Identify when both units running within 1MW

4

2

3

Calculate loss coefficient when in steady state

and when both units are running within 1MW

19

Monitoring for Tunnel Collapse – Model Validation

Rockfall identified in geotechnical inspections in 2023

Blocked ~40% of the tunnel’s cross-sectional area

Last geotechnical inspection was 2013

No knowledge of WHEN the rockfall occurred

Can we retrospectively identify this rock collapse with our new data-modelling method?

G Hunter| March 2026

20

Monitoring for Tunnel Collapse – Model Validation Physical Identification of Rockfall

Proposed Time

of Rockfall

Δ𝐾 ≈ 1.2 → 𝐴𝑟𝑒𝑎 𝐵𝑙𝑜𝑐𝑘𝑎𝑔𝑒 ≈ 36%

Actual Observed Blockage ~ 40%

G Hunter| March 2026

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

G Hunter| March 2026 21

Making the Analyses Accessible

3rd Party

Platform

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

G Hunter| March 2026 22

How are these analyses used?

Notifications set up on all analyses to alert on deviation from

set limit

Dashboard review integrated into 6-weekly Condition

Monitoring rounds

Organiser Topic used to create dashboards -> Dashboards

embedded in Asset Intellect for easy accessibility

Engineers investigate any notifications or observed changes

and raise a work order with rectifying actions.

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

G Hunter| March 2026 23

Summary

Analyses are already showing high value by identifying issues that would have

otherwise gone unnoticed

Analyses have been embedded into engineering workflows

A series of Condition Monitoring analyses were set up using already available data

Analyses were set up using in-built SEEQ tools

Final analysis outputs include dashboards and automatic notifications

Edit the footer contents under Insert >

Header and Footer

TIP - FOOTER

Add or remove the slide number

under Insert > Slide Number

TIP – PAGE NUMBERS

Use this slide for unique content that

won’t fit the other slides

TIP

G Hunter| March 2026 24

Questions?

McKay Power Station penstock

Rocky Valley Dam, Falls Creek Eildon Spillway

©2024 – Seeq Corporation

Any Questions?

Please raise your hand and a team member will bring you a microphone.

OR

Scan the QR code to submit your questions using Slido.

©2024 – Seeq Corporation

Key Takeaways

• Choose Outcomes that Matter oIdentify 1 high value operational challenge

• Scale Today oMoving from a single application to enterprise wide can be done

today

• Build your Internal Champion Network oEmpower the team to drive change

• Start your Outcome Initiative with Seeq oDefine for measurable value for 90-180 days

AGL Slide 1 Slide 2: Data‑Driven Condition Monitoring for Hydro Assets Slide 3: Who am I? Slide 4 Slide 5 Slide 6: Hydro – How It Works Slide 7: Condition Monitoring of Hydro Stations Slide 8: Analysis 1: Monitoring of MIV Slide 9: Monitoring of MIV Slide 10: Monitoring of MIV Slide 11: Monitoring of MIV Slide 12: Analysis 2: Monitoring of Sump Pump Slide 13: Monitoring of Sump Pumps Slide 14: Monitoring of Sump Pumps Slide 15: Monitoring of Sump Pumps Slide 16: Analysis 3: Monitoring for Tunnel Collapse Slide 17: Monitoring for Tunnel Collapse Slide 18: Monitoring for Tunnel Collapse Slide 19: Monitoring for Tunnel Collapse – Model Validation Slide 20: Monitoring for Tunnel Collapse – Model Validation Slide 21: Making the Analyses Accessible Slide 22: How are these analyses used? Slide 23: Summary Slide 24: Questions?

James Higgie Slide 25

Closing Session Slide 26: Key Takeaways

End Slide Slide 27