Upload
docong
View
224
Download
9
Embed Size (px)
Citation preview
SYMPTOM Project SYnchrophasor-based Model calibration for Power systems and conTrol OptiMization
Maxime Baudette, L. Vanfretti
May 2015
Why “Model Validation”?
The quality of the models used by off-line and on-line tools will affect the result of any analysis for planning and operation • Good model: approximates the
simulated response as “close” to the “measured response” as possible
Validating models helps in having a model with better accuracy: • increasing the capability of reproducing
actual power system behavior 2 3 4 5 6 7 8 9
-2
-1.5
-1
-0.5
0
0.5
1
Δ P
(pu)
Time (sec)
Measured ResponseModel Response
WECC Break-up 1996
BAD Model for Dynamic Security Assessment!!!
15-05-08 SYMPTOM PROJECT PRESENTATION 2
Near RT Model Calibration App
Real-Time Measurements Pre-processing Mode Estimation
CIM Harmonized
Model
Initialization
Mode Estimation
Linear Analysis Tools
Pre-processing Simulation
Assessment
Performance Indicator
Criteria for parameter selection
Parameter Selection
Parameter Variation
Non-linear analysis
Linear Analysis
Critical Modes Optimization of
Controller Settings and
Control Actions
Performance indicator not satisfied
Real-time alarms from mode estimation
PMU / WAMS
SCADA / EMS
Critical Modes
15-05-08 SYMPTOM PROJECT PRESENTATION 3
Near RT Model Calibration
Source of information • PMU measurements (live or archived) • Harmonized model from SCADA
snapshot and dynamic components definitions
Calibration based on small signal response Calibrated models: • Validated representation of the power
system at any instant for further use
2
1
min max
min
s.t.
(Parameters ranges)
N
i ii
f
P P P
λ λ=
= −
≤ ≤
∑)
15-05-08 SYMPTOM PROJECT PRESENTATION 4
Near RT Control Redesign/Retuning
Control optimization: • Model reduction for control redesign • Tuning/ optimization of control schemes toward small
signal stability • Control always optimized for the current operating
conditions
à Better overview of the actual health conditions
15-05-08 SYMPTOM PROJECT PRESENTATION 5
Main Project Work
• Reworked Mode Estimation app to support real-time PMU measurements with SDK: • Magnitudes, angles, frequencies, angle differences can
be used with the method • Integration in WAMS Display with GUI elements
• Modelica / PacDyn comparison for linear analysis • Test for Statnett SDK Licensing
15-05-08 SYMPTOM PROJECT PRESENTATION 6
Collaboration STRONg2rid: • Assisting Almas with supervision of MSc
Student Gudrun iTesla: • RaPId Toolbox:
• Upcomming publication on RaPId Toolbox
• CIM2ModelicaFactory:
• Participation in specifications development
• Collaboration for interfacing the Simulation Engine
STINT • Visit at UFRJ, begin of work on
Modelica/PacDyn comparison
15-05-08 SYMPTOM PROJECT PRESENTATION 7
Mode Estimation Application
Implemented with S3DK to use RT- PMU data Implementation of two algorithms to increase precision • Ambient data (general case)
[Stochastic State Space Modeling or ARMA Modelling] • Ring-down data (post fault case)
[ERA]
PMU Data Preprocessing Mode Estimation
Real-time alarms from mode estimation
(PMU/WAMS)
Ambient data
Ring-down data
15-05-08 SYMPTOM PROJECT PRESENTATION 8
Mode Estimation Application
Switch logic between the two algorithms: • Detection of the presence of Ring-down
data • Detection of large discontinuities in the
measurements
Oscillation detector from previous work appeared well suited for the detection • Detecting oscillatory ”activity” within a
predefined frequency range [0.1 – 2 Hz]
Measurements
Ambiant Data Algo.
Ringdown Data Algo.
Merging
Faults ? NO YES
Combined Estimation
15-05-08 SYMPTOM PROJECT PRESENTATION 9
Mode Estimation Application Interface
The fault detection works well after calibration (Test with Nordic32) Both method show good results so far • ERA method gives more accurate estimates
-> Higher confidence! • How to “merge the results” and derive general confidence ?
-> Graphically for now Ongoing Work: • Test on synthetic data • Test with linearized Nordic 32 with synthetized data using knowledge
from dominant paths to carefully select measurement points • Test on PMU data (RT simulation and ambient from STRONg2rid
PMUs and/or Statnett PMUs). To be done.
15-05-08 SYMPTOM PROJECT PRESENTATION 10
Sample results from statistical analysis Synthetic Model
15-05-08 11 SYMPTOM PROJECT PRESENTATION
Am
bien
t
ER
A
Sample results from statistical analysis Synthetic Data from Linearized Nordic32
15-05-08 12 SYMPTOM PROJECT PRESENTATION
Am
bien
t
ER
A
0.47 0.48 0.49 0.5 0.51 0.52 0.53 0.54 0.55 0.56
0
10
20
30
40
50
60
70
80
90
100
AMB mode1 frequency after zero removal
Prob
abilit
y De
nsity
Probability Density Function
empiricaltlocationscaleloglogisticlogisticlognormal
0 0.02 0.04 0.06 0.08 0.1 0.12
10
20
30
40
50
60
AMB mode1 damping after zero removal
Prob
abilit
y De
nsity
Probability Density Function
empiricaltlocationscalelogisticnormalgeneralized extreme value
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.90
2
4
6
8
10
12
ERA mode1 frequency after zero removal
Prob
abilit
y De
nsity
Probability Density Function
empiricaltlocationscalelogisticloglogisticnormal
-0.05 0 0.05 0.1 0.15 0.20
5
10
15
20
25
ERA mode1 damping after zero removal
Prob
abilit
y De
nsity
Probability Density Function
empiricalgeneralized extreme valuegeneralized paretotlocationscalelogistic
Modelica / PacDyn
Comparison of Modelica for linear analysis to the PacDyn reference software: • SMIB • 2 areas system • AKD (SS properties?) • Nordic44 If good results, Modelica will be used for linear analysis
15-05-08 13 SYMPTOM PROJECT PRESENTATION
Modelica / PacDyn SMIB system
15-05-08 14 SYMPTOM PROJECT PRESENTATION
• Preliminary comparisons show large difference between the Modelica and PacDyn results.
à Modification of the initialization of the Modelica models (the linearization is performed at initialization, from the powerflow data.)
• New results more satisfactory for the critical mode, but the Modelica models seem to still require additional checking.
# f (Hz) ζ (%) f (Hz) ζ (%)
1 0.8494 -5.2 0.8498 -7.34 2 1.2147 78.03 1.9385 79.28
# f (Hz) ζ (%) f (Hz) ζ (%)
1 0.8494 -5.2 0.8498 -7.34 2 1.2147 78.03 1.9385 79.28
No PSS PSS
Development Status
15-05-08 15 SYMPTOM PROJECT PRESENTATION
Real-Time Measurements Pre-processing Mode
Estimation
CIM Harmonized
Model
Initialization
Mode Estimation
Linear Analysis Tools
Pre-processing Simulation
Assessment
Performance Indicator
Criteria for parameter selection
Parameter Selection
Parameter Variation
Non-linear analysis
Linear Analysis
Critical Modes Optimization of Controller Settings and
Control Actions
Performance indicator not satisfied
Real-time alarms from mode estimation
PMU / WAMS
SCADA / EMS
Critical Modes
Orange: Developed, not completely tested yet Red: Not yet developed, work for 2015
Future Work
• Wrapping of the script for model building and model simulation to interface with LabVIEW when necessary
• Develop the interface for linear analysis with Modelica
• Develop and implement the performance indicator
15-05-08 16 SYMPTOM PROJECT PRESENTATION
THANK YOU
15-05-08 SYMPTOM PROJECT PRESENTATION 17
Future Development
• Develop user interface of the mode meter module to fit requirements
• Implementation of data export in HDF5 for collaboration within iTesla
• Test the tools at Statnett control center, Alta
15-05-08 SYMPTOM PROJECT PRESENTATION 18
iTesla: Power System Library
15-05-08 19 SYMPTOM PROJECT PRESENTATION
• The FP7 iTESLA project develops a high level library for modeling power grid components
• Generators, • Governors, • Excitation Systems • Stabilizers • Branches, • Loads, • Buses, • Events • Connectors • Non Electrical Components
iTesla / Mango: Power System Library
E.g.: Model in modelica validated against a PSSE model
15-05-08 20 SYMPTOM PROJECT PRESENTATION
iTesla: Model Execution Engine
Open-source software for cyber-physical system simulation Plug-in different compilers and solvers: Use of Python Scripts Simulation of .fmu using JModelica.
• JModelica contains Python scripts for compile models into .fmu and simulate .fmu
Simulation of .mo with OpenModelica / Dymola • OpenModelica comes with JAVA libs (.jar files) leading calls to OMC (Open
Modelica Compiler) • Dymola script needs run-time license for executing simulations
15-05-08 21 SYMPTOM PROJECT PRESENTATION
Pro
perti
es
Res
ults
HDF5
JM OMC
PYTHON
JAVA
Dymola
iTesla: Model Execution Engine
15-05-08 22 SYMPTOM PROJECT PRESENTATION