State-estimation results of pandapower and power-grid-model are different

[Milena1511]

Hello!
I have been trying to convert a pandapower-network into power-grid-model and run a state-estimation on both networks. The conversion with PandaPowerConverter from power-grid-model-io was successful. I added the measurements at the pandapower-buses as voltage and power-sensors to the network modelled in power-grid-model. These networks are attached. There is also a json-file "idx_lookup.json" attached which is a serialized version of the idx_lookup-attribute of the PandaPowerConverter and maps the power-grid-model indices to the indices of the elements in the pandapower network tables. Then I executed the following code:

import pandapower as pp
from power_grid_model import PowerGridModel, ComponentType
from power_grid_model.utils import json_deserialize_from_file

pp_net = pp.from_json("pp_net.json")
pgm_input = json_deserialize_from_file("pgm_net.json")
pgm_net = PowerGridModel(pgm_input)

pp.estimation.estimate(pp_net)
state_estimation_results = pgm_net.calculate_state_estimation(error_tolerance=1e-3)

The results of the both networks are very different. For example the first line (the power-grid-model id 118 matches the id 0 in pandapower):

In [41]: res_line_est_pgm = pd.DataFrame(state_estimation_results[ComponentType.line])

In [42]: pp_net.res_line_est.loc[0, "p_to_mw"]
Out[42]: np.float64(-29.727469274206157)

In [43]: res_line_est_pgm.loc[0, "p_to"]*1e-6
Out[43]: np.float64(-1.040860668477377)

In [44]: pp_net.res_line_est.loc[0, "i_from_ka"]
Out[44]: np.float64(0.16288191139708483)

DEBUG:asyncio:Using selector: EpollSelector
In [45]: res_line_est_pgm.loc[0, "i_from"]*1e-3
Out[45]: np.float64(0.03254555759419108)

Also the result-values for the voltages at the nodes are unrealistic:

In [59]: res_node_est_pgm = pd.DataFrame(state_estimation_results[ComponentType.node])

In [60]: res_node_est_pgm.u_pu.describe()
Out[60]: 
count    118.000000
mean       0.177706
std        0.026815
min        0.115898
25%        0.147870
50%        0.183951
75%        0.203110
max        0.220805
Name: u_pu, dtype: float64

For pandapower, they look okay:

In [61]: pp_net.res_bus_est.vm_pu.describe()
Out[61]: 
count    118.000000
mean       0.927139
std        0.060537
min        0.793122
25%        0.882525
50%        0.930570
75%        0.962631
max        1.079446

Name: vm_pu, dtype: float64

Do you have any suggestions on why the results are so different?

idx_lookup.json
pgm_net.json
pp_net.json

[petersalemink95]

Hi @Milena1511, thanks for asking the question here. I think they use p.u. voltage at PandaPower, looking at the voltages in your voltage sensors. At Power Grid Model u_measured should be provided in V. I think this is the reason your results are different. Can you check if this is indeed the case?

[petersalemink95]

Just checked and PandaPower voltages are indeed in p.u. so when converting your voltage measurements from PandaPower to PGM you need to convert p.u. to Volt as well :)

[Milena1511]

I corrected the values in the voltage sensors (and the power-sensors as well, they were in MW) and now it looks a lot better:

>>> pd.DataFrame(state_estimation_results[ComponentType.node]).u_pu.describe()
count    118.000000
mean       0.931406
std        0.089459
min        0.786863
25%        0.831020
50%        0.945723
75%        1.020064
max        1.068462
Name: u_pu, dtype: float64
>>> res_line_est_pgm = pd.DataFrame(state_estimation_results[ComponentType.line])
>>> res_line_est_pgm.loc[0, "p_to"]*1e-6
np.float64(-31.50631338908761)
>>> res_line_est_pgm.loc[0, "i_from"]*1e-3
np.float64(0.13403518169615117)

Thanks a lot for the quick help!

[petersalemink95]

No problem! Happy to help