2019 Brake Tests: Initial Analysis - 2019 Brake Tests Initial Analysis.pdfOBAT also includes VARPRO and Matrix Pencil –results should be validated in future work Results presented

2019 Brake Tests:Initial Analysis

Jim Follum

September 19, 2019

PNNL-SA-147531

2

• BPA has performed system tests over the past decade

• Large disturbances excite system dynamics, providing valuable data

▪ Insertion of Chief Joseph Dynamic Braking Resistor

▪ Probing signal injection modulates flow on the PDCI

Background

Chief Joseph Dynamic Braking Resistor. Image courtesy of BPA.

3

Test Series A: Calibration Checks on PDCI Probing Signals

Step A0 [16:10] Celilo instrumentation check using +20MW waveform (10 seconds) and -20 MW (10 seconds). Check proper function of PSG using Celilo/Sylmar DC metering.

Step A1 [16:15] Calibration check on MSF-1/5/2/100 for ±20 MW noise probing for a duration of one period (100 seconds). Adjust PSG scaling if needed.

Step A4 [16:30] Apply MSF-0.1/4x for ±20 MW single frequency sine wave for four cycles. (1 period)




Test Series B: Noise Probing

Step B1 [17:10] Measurement of ambient noise conditions

Step B2 [17:30] Apply a +20 MW MSF-1/5/2/100 for a duration of 12 periods (20 minutes).

Test Series C: Chief Joseph brake insertion (Afternoon)

Step C1 [21:14] Apply Chief Joseph braking resistor

Step C2 [21:20] Apply a +20 MW MSF-1/5/2/100 for a duration of 12 periods (20 minutes).

Step C3 [21:44] Apply Chief Joseph braking resistor

Notes:

• (Times are in UTC and reflect the testing plan)

• PSG = Probing Signal Generator

• MSF = Multi-Sine Fitted

• Tests conducted May 7 and August 21, 2019

• Presented results focus on Test Series C

Testing Procedure

4

Big Eddy – Celilo MW FlowSteps A0-A1: Instrumentation/Calibration Checks

5

Big Eddy – Celilo MW FlowSteps A4-A7: Single Frequency Sinusoids

6

Big Eddy – Celilo MW FlowStep B2: Noise Probing

Full Signal

(12 periods)

Zoomed In

Frequency Domain

7

Derived Frequency at all SubstationsSteps C1 and C3: Brake Insertions

8

• Archive Walker

▪ Ingests, processes, and exports PMU data

▪ Detects ringdowns

▪ Implements mode meter

• Oscillation Baselining and Analysis Tool (OBAT)

▪ Performs ringdown analysis

▪ Displays mode shape geographically

Tools

9

Setup

• Input Signals

▪ Linear combinations of voltage angles

▪ Voltage angles from 21 substations in BPA

▪ Voltage angles are used to derive frequency deviations using first-order derivative filter

• Modes

▪ North-South A (NS-A)

➢ Typical frequency ~ 0.25 Hz

➢ Alberta vs System

➢ British Columbia and Pacific Northwest swing with Alberta

▪ North-South B (NS-B)

➢ Typical frequency ~ 0.38 Hz

➢ Alberta vs British Columbia and Northern US vs Southern US

10

Methods

• Ringdown

▪ Prony analysis

▪ OBAT also includes VARPRO and Matrix Pencil – results should be validated in future work

▪ Results presented for linear combinations and all substations

▪ Consistent results from two brake insertions provide validation

• Mode Meter

▪ Yule-Walker algorithm for ambient data analysis

➢ Not suitable for use during PDCI probing

➢ Care required to include ringdowns

➢ Estimates are only reported during ambient conditions

▪ Input signals and model orders match BPA setup

▪ Results provide validation for ringdown results

11

May 7 Results

Notes:

• MM = Mode Meter

• Lin Combo = Prony applied to linear combination of signals

• All Sigs = Prony applied to all 21 signals

North-South A North-South B

12

August 21 Results

Notes:

• MM = Mode Meter

• Lin Combo = Prony applied to linear combination of signals

• All Sigs = Prony applied to all 21 signals

North-South A North-South B

Typical OBAT Results: Fit and Shape for NS-A and NS-B

14

Discussion

Observations

• Mode meter estimates generally agree with ringdown analyses

• As expected, modes were well-damped during tests

• North-South A mode not sufficiently excited by brake insertions for reliable ringdown (Prony) analysis

• Good data quality

Recommended Additional Steps

• Ringdown analysis with VARPRO, Matrix Pencil, etc.

• Apply mode meter algorithms that:

▪ Incorporate probing signals to improve estimates

▪ Handle ringdowns in robust implementation

▪ Estimate mode shape

• Evaluate mode shape using non-parametric methods, e.g., Empirical Transfer Function Estimation (ETFE)

• Include PMUs from a wider area

Thank you

15

2019 Brake Tests: Initial Analysis - 2019 Brake Tests Initial Analysis.pdfOBAT also includes VARPRO and Matrix Pencil –results should be validated in future work Results presented

Documents