overlapping plot in time domain simulation when Fault model added

[seedpower]

import numpy as np
import andes

ss = andes.run(andes.get_case("ieee14/ieee14_full.xlsx"), setup=False, default_config=True)
ss.add("Fault", dict(bus=9, tf=0.4, tc=0.7))
ss.setup()
ss.PFlow.run()

ss.TDS.config.tstep = 0.1
ss.TDS.config.shrinkt = 1

for tf in np.linspace(0.1, 1.2, 12):
    ss.TDS.config.tf = tf
    ss.TDS.run()

# ss.TDS.config.tf = 1.2
# ss.TDS.run()

ss.TDS.load_plotter()
ss.TDS.plotter.plot(ss.GENROU.delta)

The diagram (P1) is:

Now, I change code to this:

# for tf in np.linspace(0.1, 1.2, 12):
#     ss.TDS.config.tf = tf
#     ss.TDS.run()

ss.TDS.config.tf = 1.2
ss.TDS.run()

The diagram (P2) is:

If I set ss.TDS.config.shrinkt = 0，use below code:

ss.TDS.config.tstep = 0.1
ss.TDS.config.shrinkt = 0

for tf in np.linspace(0.1, 1.2, 12):
    ss.TDS.config.tf = tf
    ss.TDS.run()

# ss.TDS.config.tf = 1.2
# ss.TDS.run()

The diagram (P3) is:

Existing problems:

1. compare P1 with P2, TDS resume, overlapping segments 0.4~0.5 and 0.7~0.8.
2. compare P1 with P3, setting ss.TDS.config.shrinkt=0 , overlapping segment become shorter.

[cuihantao]

Thanks for reporting. I will fix it asap. It might have to do with the timing logic again.

[cuihantao]

I pushed a fix to my develop branch. Please test and let me know. Thanks!

[cuihantao]

I just pushed a change. Mind testing agin? I probably need more time to test if more thoroughly as well.

[seedpower]

The bug seems to be fixed.

The diagrams are slightly different using two methods:

ss = andes.run(andes.get_case("ieee14/ieee14_full.xlsx"), setup=False, default_config=True)
ss.add("Fault", dict(bus=9, tf=0.4, tc=0.7))
ss.setup()
ss.PFlow.run()

ss.TDS.config.tstep = 0.1
ss.TDS.config.shrinkt = 1

# Method 1:
for tf in np.linspace(0.1, 1.2, 12):
    ss.TDS.config.tf = tf
    ss.TDS.run()

ss.TDS.load_plotter()
ss.TDS.plotter.plot(ss.GENROU.omega, title="ss.TDS.config.tf = np.linspace(0.1, 1.2, 12)")

# Method 2:
# ss.TDS.config.tf = 1.2
# ss.TDS.run()

# ss.TDS.load_plotter()
# ss.TDS.plotter.plot(ss.GENROU.omega, title="ss.TDS.config.tf = 1.2")

[seedpower]

And more, a gap plot bug:

import numpy as np
import andes
from andes.utils import get_case

ss = andes.run(get_case(“ieee14_1.00_Line_4_0.25_1.00_1.83.xlsx”), default_config=True)

ss.TDS.config.criteria = 0

# Method 1:
for tf in np.linspace(0.1, 5.0, 50):
    print("t={t:.2f}".format(t=tf))
    ss.TDS.config.tf = tf
    ss.TDS.run()

ss.TDS.load_plotter()
ss.TDS.plotter.plot(ss.GENROU.delta)

# Method 2:
# ss.TDS.config.tf = 5.0
# ss.TDS.run()
# ss.TDS.load_plotter()
# ss.TDS.plotter.plot(ss.GENROU.delta)

Method 1 diagram:

Method 2 diagram:

Existing problem:

1. discontinuity occurs in t ≈ 3.5s

case file:
ieee14_1.00_Line_4_0.25_1.00_1.83.xlsx

[cuihantao]

ss = andes.run(andes.get_case("ieee14/ieee14_full.xlsx"), setup=False, default_config=True)
ss.add("Fault", dict(bus=9, tf=0.4, tc=0.7))
ss.setup()
ss.PFlow.run()

ss.TDS.config.tstep = 0.1
ss.TDS.config.shrinkt = 1

# Method 1:
for tf in np.linspace(0.1, 1.2, 12):
    ss.TDS.config.tf = tf
    ss.TDS.run()

ss.TDS.load_plotter()
ss.TDS.plotter.plot(ss.GENROU.omega, title="ss.TDS.config.tf = np.linspace(0.1, 1.2, 12)")

# Method 2:
# ss.TDS.config.tf = 1.2
# ss.TDS.run()

# ss.TDS.load_plotter()
# ss.TDS.plotter.plot(ss.GENROU.omega, title="ss.TDS.config.tf = 1.2")

This one is expected since the discretization is taken at different time steps. All these points should be on the same trajectory.

[cuihantao]

And more, a gap plot bug:

import numpy as np
import andes
from andes.utils import get_case

ss = andes.run(get_case(“ieee14_1.00_Line_4_0.25_1.00_1.83.xlsx”), default_config=True)

ss.TDS.config.criteria = 0

# Method 1:
for tf in np.linspace(0.1, 5.0, 50):
    print("t={t:.2f}".format(t=tf))
    ss.TDS.config.tf = tf
    ss.TDS.run()

ss.TDS.load_plotter()
ss.TDS.plotter.plot(ss.GENROU.delta)

# Method 2:
# ss.TDS.config.tf = 5.0
# ss.TDS.run()
# ss.TDS.load_plotter()
# ss.TDS.plotter.plot(ss.GENROU.delta)

Method 1 diagram:

Method 2 diagram:

Existing problem:

1. discontinuity occurs in t ≈ 3.5s

case file: ieee14_1.00_Line_4_0.25_1.00_1.83.xlsx

Again, thank you for reporting. I pushed a fix to my develop branch.

The issue is due to the non-convergence at t=3.9xx. It was a bug in the step size control where the shrinked step size was not properly applied.

[cuihantao]

On a side note, if you don't need the progress bar during your reinforcement learning studies, you can turn it off by setting ss.TDS.config.no_tqdm = 1.

[seedpower]

Thank you, it's very useful.

Need some help, My goal is to keep transient stability:

In practical engineering, the generators generally are cut off as a whole.

Q1: At present, I have two ways to achieve this. One is to modify the u status of the generator, and the other is to set tm0=0 of the generator . Which one is better?

loads usually are cut off in acertain proportion.

Q2: Can I achieve this by setting Ppf * (1-proportion) of PQ load?