Manipulate the Simulation¶

This example shows how to play with the simulation, such as contingency analysis and manipulate the constraints.

In [1]:

import ams

import datetime

In [2]:

print("Last run time:", datetime.datetime.now().strftime("%Y-%m-%d %H:%M:%S"))

print(f'ams:{ams.__version__}')

Last run time: 2024-06-18 20:16:41
ams:0.9.8

In [3]:

ams.config_logger(stream_level=20)

Manipulate the Simulation¶

Load Case¶

In [4]:

sp = ams.load(ams.get_case('5bus/pjm5bus_uced.xlsx'),
              setup=True,
              no_output=True,)

Parsing input file "/Users/jinningwang/Documents/work/mambaforge/envs/amsre/lib/python3.9/site-packages/ams/cases/5bus/pjm5bus_uced.xlsx"...
Input file parsed in 0.1065 seconds.
Zero line rates detacted in rate_b, rate_c, adjusted to 999.
System set up in 0.0025 seconds.

The system load are defined in model PQ.

In [5]:

sp.PQ.as_df()

Out[5]:

	idx	u	name	bus	Vn	p0	q0	vmax	vmin	owner	ctrl
uid
0	PQ_1	1.0	PQ 1	Bus_2	230.0	3.0	0.9861	1.1	0.9	None	1.0
1	PQ_2	1.0	PQ 2	Bus_3	230.0	3.0	0.9861	1.1	0.9	None	1.0
2	PQ_3	1.0	PQ 3	Bus_4	230.0	4.0	1.3147	1.1	0.9	None	1.0

In RTED, system load is referred as pd.

In [6]:

sp.RTED.pd.v

Out[6]:

array([3., 3., 4.])

Run Simulation¶

RTED can be solved and one can inspect the results as discussed in previous example.

In [7]:

sp.RTED.run(solver='CLARABEL')

<RTED> initialized in 0.0130 seconds.
<RTED> solved as optimal in 0.0120 seconds, converged in 10 iterations with CLARABEL.

Out[7]:

True

Power generation pg and line flow plf can be accessed as follows.

In [8]:

sp.RTED.pg.v

Out[8]:

array([2. , 2.1, 5.2, 0.7])

In [9]:

sp.RTED.plf.v

Out[9]:

array([ 0.70595331,  0.68616798,  0.00192539, -1.58809337,  0.61190663,
       -0.70192539,  0.70595331])

Change Load¶

The load values can be manipulated in the model PQ.

Note the difference between Model.set and Model.alter:

set: This method *WILL NOT* modify the input values from the case file that have not been converted to the system base. As a result, changes applied by this method *WILL NOT* affect the dumped case file.

alter: If the method operates on an input parameter, the new data *should be in the same base as that in the input file. This function will convert value to per unit in the system base whenever necessary. The values for storing the input data, i.e., the parameter's vin field, will be overwritten. As a result, altered values WILL BE* reflected in the dumped case file.

Besides, Group also has method set but has no alter.

In [10]:

sp.PQ.set(src='p0', attr='v', idx=['PQ_1', 'PQ_2'], value=[3.2, 3.2])

Out[10]:

True

According parameters need to be updated to make the changes effective in the optimization model. If not sure which parameters need to be updated, one can use update() to update all parameters.

In [11]:

sp.RTED.update('pd')

Out[11]:

True

After manipulation, the routined can be solved again.

In [12]:

sp.RTED.run(solver='CLARABEL')

<RTED> solved as optimal in 0.0030 seconds, converged in 10 iterations with CLARABEL.

Out[12]:

True

In [13]:

sp.RTED.pg.v

Out[13]:

array([2.        , 2.1       , 5.19999999, 1.10000001])

An alternative way is to alter the load through RTED.

As pd has owner StaticLoad and soruce p0, the parameter update through RTED actually happens to StaticLoad.p0.

In [14]:

sp.RTED.pd.owner

Out[14]:

StaticLoad (3 devices) at 0x142804dc0

In [15]:

sp.RTED.pd.src

Out[15]:

'p0'

Similarly, the load can be changed using set method.

In [16]:

sp.RTED.set(src='pd', attr='v', idx=['PQ_1', 'PQ_2'], value=[0.2, 0.2])

Out[16]:

True

Remember to update the optimization parameters after the change.

In [17]:

sp.RTED.update('pd')

Out[17]:

True

We can see that the original load is also updated.

In [18]:

sp.PQ.as_df()

Out[18]:

	idx	u	name	bus	Vn	p0	q0	vmax	vmin	owner	ctrl
uid
0	PQ_1	1.0	PQ 1	Bus_2	230.0	0.2	0.9861	1.1	0.9	None	1.0
1	PQ_2	1.0	PQ 2	Bus_3	230.0	0.2	0.9861	1.1	0.9	None	1.0
2	PQ_3	1.0	PQ 3	Bus_4	230.0	4.0	1.3147	1.1	0.9	None	1.0

In [19]:

sp.RTED.run(solver='CLARABEL')

<RTED> solved as optimal in 0.0019 seconds, converged in 10 iterations with CLARABEL.

Out[19]:

True

As expected, the power generation also changed.

In [20]:

sp.RTED.pg.v

Out[20]:

array([1.99999999, 1.30000001, 0.5       , 0.6       ])

Trip a Generator¶

We can see that there are three PV generators in the system.

In [21]:

sp.PV.as_df()

Out[21]:

	idx	u	name	Sn	Vn	bus	busr	p0	q0	pmax	...	Qc2min	Qc2max	Ragc	R10	R30	Rq	apf	pg0	td1	td2
uid
0	PV_1	1.0	Alta	100.0	230.0	Bus_1	None	1.0000	0.0	2.1	...	0.0	0.0	999.0	999.0	999.0	999.0	0.0	0.0	0.5	0.0
1	PV_3	1.0	Solitude	100.0	230.0	Bus_3	None	3.2349	0.0	5.2	...	0.0	0.0	999.0	999.0	999.0	999.0	0.0	0.0	0.5	0.0
2	PV_5	1.0	Brighton	100.0	230.0	Bus_5	None	4.6651	0.0	6.0	...	0.0	0.0	999.0	999.0	999.0	999.0	0.0	0.0	0.5	0.0

3 rows × 33 columns

PV_1 is tripped by setting its connection status u to 0.

In [22]:

sp.StaticGen.set(src='u', attr='v', idx='PV_1', value=0)

Out[22]:

True

In AMS, some parameters are defiend as constants in the numerical optimization model to follow the CVXPY DCP and DPP rules. Once non-parametric parameters are changed, the optimization model will be re-initialized to make the changes effective.

More details can be found at CVXPY - Disciplined Convex Programming.

In [23]:

sp.RTED.update()

<RTED> reinit OModel due to non-parametric change.

Out[23]:

True

Then we can re-solve the model.

In [24]:

sp.RTED.run(solver='CLARABEL')

<RTED> solved as optimal in 0.0126 seconds, converged in 9 iterations with CLARABEL.

Out[24]:

True

We can see that the tripped generator has no power generation.

In [25]:

sp.RTED.pg.v.round(2)

Out[25]:

array([ 2. , -0. ,  1.8,  0.6])

Trip a Line¶

We can inspect the Line model to check the system topology.

In [26]:

sp.Line.as_df()

Out[26]:

	idx	u	name	bus1	bus2	Sn	fn	Vn1	Vn2	r	...	tap	phi	rate_a	rate_b	rate_c	owner	xcoord	ycoord	amin	amax
uid
0	Line_0	1.0	Line AB	Bus_1	Bus_2	100.0	60.0	230.0	230.0	0.00281	...	1.0	0.0	4.0	999.0	999.0	None	None	None	-6.283185	6.283185
1	Line_1	1.0	Line AD	Bus_1	Bus_4	100.0	60.0	230.0	230.0	0.00304	...	1.0	0.0	2.0	999.0	999.0	None	None	None	-6.283185	6.283185
2	Line_2	1.0	Line AE	Bus_1	Bus_5	100.0	60.0	230.0	230.0	0.00064	...	1.0	0.0	2.0	999.0	999.0	None	None	None	-6.283185	6.283185
3	Line_3	1.0	Line BC	Bus_2	Bus_3	100.0	60.0	230.0	230.0	0.00108	...	1.0	0.0	2.0	999.0	999.0	None	None	None	-6.283185	6.283185
4	Line_4	1.0	Line CD	Bus_3	Bus_4	100.0	60.0	230.0	230.0	0.00297	...	1.0	0.0	2.0	999.0	999.0	None	None	None	-6.283185	6.283185
5	Line_5	1.0	Line DE	Bus_4	Bus_5	100.0	60.0	230.0	230.0	0.00297	...	1.0	0.0	2.4	999.0	999.0	None	None	None	-6.283185	6.283185
6	Line_6	1.0	Line AB2	Bus_1	Bus_2	100.0	60.0	230.0	230.0	0.00281	...	1.0	0.0	4.0	999.0	999.0	None	None	None	-6.283185	6.283185

7 rows × 28 columns

Here line 2 is tripped by setting its connection status u to 0.

Note that in ANDES, dynamic simulation of line tripping should use model Toggle.

In [27]:

sp.Line.set(src='u', attr='v', idx='Line_1', value=0)

Out[27]:

True

In [28]:

sp.RTED.update()

<RTED> reinit OModel due to non-parametric change.

Out[28]:

True

In [29]:

sp.RTED.run(solver='CLARABEL')

<RTED> solved as optimal in 0.0122 seconds, converged in 9 iterations with CLARABEL.

Out[29]:

True

Here we can see the tripped line has no flow.

In [30]:

sp.RTED.plf.v.round(2)

Out[30]:

array([-0.12,  0.  ,  0.24, -0.44,  1.16, -0.84, -0.12])

Disable Constraints¶

In addition to the system parameters, the constraints can also be manipulated.

Here, we load the case to a new system.

In [31]:

spc = ams.load(ams.get_case('5bus/pjm5bus_uced.xlsx'),
               setup=True,
               no_output=True,)

Parsing input file "/Users/jinningwang/Documents/work/mambaforge/envs/amsre/lib/python3.9/site-packages/ams/cases/5bus/pjm5bus_uced.xlsx"...
Input file parsed in 0.0398 seconds.
Zero line rates detacted in rate_b, rate_c, adjusted to 999.
System set up in 0.0027 seconds.

In [32]:

spc.RTED.init()

<RTED> initialized in 0.0115 seconds.

Out[32]:

True

In [33]:

spc.RTED.set(src='rate_a', attr='v', idx=['Line_3'], value=0.6)

Out[33]:

True

In [34]:

spc.RTED.update('rate_a')

Out[34]:

True

We can inspect the constraints status as follows. All constraints are turned on by default.

In [35]:

spc.RTED.constrs

Out[35]:

OrderedDict([('pglb', Constraint: pglb [ON]),
             ('pgub', Constraint: pgub [ON]),
             ('sba', Constraint: sbus [ON]),
             ('pb', Constraint: pb [ON]),
             ('plflb', Constraint: plflb [ON]),
             ('plfub', Constraint: plfub [ON]),
             ('alflb', Constraint: alflb [ON]),
             ('alfub', Constraint: alfub [ON]),
             ('rbu', Constraint: rbu [ON]),
             ('rbd', Constraint: rbd [ON]),
             ('rru', Constraint: rru [ON]),
             ('rrd', Constraint: rrd [ON]),
             ('rgu', Constraint: rgu [ON]),
             ('rgd', Constraint: rgd [ON])])

Then, solve the dispatch and inspect the line flow.

In [36]:

spc.RTED.run(solver='CLARABEL')

<RTED> solved as optimal in 0.0721 seconds, converged in 10 iterations with CLARABEL.

Out[36]:

True

In [37]:

spc.RTED.plf.v.round(2)

Out[37]:

array([ 1.2 ,  0.87, -1.17, -0.6 , -0.02, -1.15,  1.2 ])

In the next, we can disable specific constraints, and the parameter name takes both single constraint name or a list of constraint names.

In [38]:

spc.RTED.disable(['plflb', 'plfub'])

Turn off constraints: plflb, plfub

Out[38]:

True

Now, it can be seen that the two constraints are disabled.

In [39]:

spc.RTED.constrs

Out[39]:

OrderedDict([('pglb', Constraint: pglb [ON]),
             ('pgub', Constraint: pgub [ON]),
             ('sba', Constraint: sbus [ON]),
             ('pb', Constraint: pb [ON]),
             ('plflb', Constraint: plflb [OFF]),
             ('plfub', Constraint: plfub [OFF]),
             ('alflb', Constraint: alflb [ON]),
             ('alfub', Constraint: alfub [ON]),
             ('rbu', Constraint: rbu [ON]),
             ('rbd', Constraint: rbd [ON]),
             ('rru', Constraint: rru [ON]),
             ('rrd', Constraint: rrd [ON]),
             ('rgu', Constraint: rgu [ON]),
             ('rgd', Constraint: rgd [ON])])

In [40]:

spc.RTED.run(solver='CLARABEL')

Disabled constraints: plflb, plfub
<RTED> initialized in 0.0029 seconds.
<RTED> solved as optimal in 0.0120 seconds, converged in 9 iterations with CLARABEL.

Out[40]:

True

We can see that now the line flow limits are not in effect.

In [41]:

spc.RTED.plf.v.round(2)

Out[41]:

array([ 0.71,  0.69,  0.  , -1.59,  0.61, -0.7 ,  0.71])

Similarly, you can also enable the constraints again.

In [42]:

spc.RTED.enable(['plflb', 'plfub'])

Turn on constraints: plflb, plfub

Out[42]:

True

In [43]:

spc.RTED.constrs

Out[43]:

OrderedDict([('pglb', Constraint: pglb [ON]),
             ('pgub', Constraint: pgub [ON]),
             ('sba', Constraint: sbus [ON]),
             ('pb', Constraint: pb [ON]),
             ('plflb', Constraint: plflb [ON]),
             ('plfub', Constraint: plfub [ON]),
             ('alflb', Constraint: alflb [ON]),
             ('alfub', Constraint: alfub [ON]),
             ('rbu', Constraint: rbu [ON]),
             ('rbd', Constraint: rbd [ON]),
             ('rru', Constraint: rru [ON]),
             ('rrd', Constraint: rrd [ON]),
             ('rgu', Constraint: rgu [ON]),
             ('rgd', Constraint: rgd [ON])])

In [44]:

spc.RTED.run(solver='CLARABEL')

<RTED> initialized in 0.0005 seconds.
<RTED> solved as optimal in 0.0224 seconds, converged in 10 iterations with CLARABEL.

Out[44]:

True

In [45]:

spc.RTED.plf.v.round(2)

Out[45]:

array([ 1.2 ,  0.87, -1.17, -0.6 , -0.02, -1.15,  1.2 ])

Alternatively, you can also force init the dispatch to rebuild the system matrices, enable all constraints, and re-init the optimization models.

In [46]:

spc.RTED.disable(['plflb', 'plfub', 'rgu', 'rgd'])

Turn off constraints: plflb, plfub, rgu, rgd

Out[46]:

True

In [47]:

spc.RTED.init(force=True)

<RTED> initialized in 0.0026 seconds.

Out[47]:

True

In [48]:

spc.RTED.constrs

Out[48]:

OrderedDict([('pglb', Constraint: pglb [ON]),
             ('pgub', Constraint: pgub [ON]),
             ('sba', Constraint: sbus [ON]),
             ('pb', Constraint: pb [ON]),
             ('plflb', Constraint: plflb [ON]),
             ('plfub', Constraint: plfub [ON]),
             ('alflb', Constraint: alflb [ON]),
             ('alfub', Constraint: alfub [ON]),
             ('rbu', Constraint: rbu [ON]),
             ('rbd', Constraint: rbd [ON]),
             ('rru', Constraint: rru [ON]),
             ('rrd', Constraint: rrd [ON]),
             ('rgu', Constraint: rgu [ON]),
             ('rgd', Constraint: rgd [ON])])

Alter Config¶

Routines have an config attribute as configuration settings.

In [49]:

spf = ams.load(ams.get_case('5bus/pjm5bus_uced.xlsx'),
               setup=True,
               no_output=True,)

Parsing input file "/Users/jinningwang/Documents/work/mambaforge/envs/amsre/lib/python3.9/site-packages/ams/cases/5bus/pjm5bus_uced.xlsx"...
Input file parsed in 0.0340 seconds.
Zero line rates detacted in rate_b, rate_c, adjusted to 999.
System set up in 0.0024 seconds.

In RTED, the default interval is 5/60 [hour], and the formulations has been adjusted to fit the interval.

In [50]:

spf.RTED.config

Out[50]:

OrderedDict([('t', 0.08333333333333333)])

In [51]:

spf.RTED.run(solver='CLARABEL')

<RTED> initialized in 0.0109 seconds.
<RTED> solved as optimal in 0.0118 seconds, converged in 10 iterations with CLARABEL.

Out[51]:

True

In [52]:

spf.RTED.obj.v

Out[52]:

0.1953750000267141

We can update the interval to 1 [hour] and re-solve the dispatch.

Note that in this senario, compared to DCOPF, RTED has extra costs for pru and prd.

In [53]:

spf.RTED.config.t = 60/60

Remember to update the parameters after the change.

In [54]:

spf.RTED.update()

<RTED> reinit OModel due to non-parametric change.

Out[54]:

True

In [55]:

spf.RTED.run(solver='SCS')

<RTED> solved as optimal in 0.0602 seconds, converged in 325 iterations with SCS.

Out[55]:

True

We can then get the objective value.

In [56]:

spf.RTED.obj.v

Out[56]:

2.3444999975679632

Note that in this build-in case, the cru and crd are defined as zero.

In [57]:

spf.RTED.cru.v

Out[57]:

array([0., 0., 0., 0.])

In [58]:

spf.RTED.crd.v

Out[58]:

array([0., 0., 0., 0.])

As benchmark, we can solve the DCOPF.

In [59]:

spf.DCOPF.run(solver='SCS')

<DCOPF> initialized in 0.0075 seconds.
<DCOPF> solved as optimal in 0.0079 seconds, converged in 225 iterations with SCS.

Out[59]:

True

As expected, the DCOPF has a similar objective value.

In [60]:

spf.DCOPF.obj.v

Out[60]:

2.3445000000033347