fastSE (power system state estimation)

A performant state estimator for power system

sparse matrix + jit + klu + custom improved ordering + python = efficient in computation and development!

Installation

To install, simply run pip install fastSE in your command prompt.

How to use

Here is one simple example. solve_se_lm is a high-level function which computes derivatives, assemble them as sparse matrix and then calculate the estimates using sparse matrix solver. All the low-level functions could also be imported and used individually.

from fastse import solve_se_lm, bdd_validation, StateEstimationInput
from scipy.sparse import csr_matrix
import numpy as np

import time
# A 5 bus example from Prof. Overbye's textbook
# node impedance
Ybus = np.array([[3.729 - 49.720j, 0.000 + 0.000j, 0.000 + 0.000j,
        0.000 + 0.000j, -3.729 + 49.720j],
       [0.000 + 0.000j, 2.678 - 28.459j, 0.000 + 0.000j,
        -0.893 + 9.920j, -1.786 + 19.839j],
       [0.000 + 0.000j, 0.000 + 0.000j, 7.458 - 99.441j,
        -7.458 + 99.441j, 0.000 + 0.000j],
       [0.000 + 0.000j, -0.893 + 9.920j, -7.458 + 99.441j,
        11.922 - 147.959j, -3.571 + 39.679j],
       [-3.729 + 49.720j, -1.786 + 19.839j, 0.000 + 0.000j,
        -3.571 + 39.679j, 9.086 - 108.578j]])
Ybus = csr_matrix(Ybus)

# branch impedance
Yf = np.array([[ 3.729-49.720j,  0.000 +0.000j,  0.000 +0.000j,  0.000 +0.000j,
    -3.729+49.720j],
   [ 0.000 +0.000j, -0.893 +9.920j,  0.000 +0.000j,  0.893 -9.060j,
     0.000 +0.000j],
   [ 0.000 +0.000j, -1.786+19.839j,  0.000 +0.000j,  0.000 +0.000j,
     1.786-19.399j],
   [ 0.000 +0.000j,  0.000 +0.000j,  7.458-99.441j, -7.458+99.441j,
     0.000 +0.000j],
   [ 0.000 +0.000j,  0.000 +0.000j,  0.000 +0.000j, -3.571+39.679j,
     3.571-39.459j]])
Yf = csr_matrix(Yf)

Yt = np.array([[-3.729+49.720j,  0.000 +0.000j,  0.000 +0.000j,  0.000 +0.000j,
     3.729-49.720j],
   [ 0.000 +0.000j,  0.893 -9.060j,  0.000 +0.000j, -0.893 +9.920j,
     0.000 +0.000j],
   [ 0.000 +0.000j,  1.786-19.399j,  0.000 +0.000j,  0.000 +0.000j,
    -1.786+19.839j],
   [ 0.000 +0.000j,  0.000 +0.000j, -7.458+99.441j,  7.458-99.441j,
     0.000 +0.000j],
   [ 0.000 +0.000j,  0.000 +0.000j,  0.000 +0.000j,  3.571-39.459j,
    -3.571+39.679j]])
Yt = csr_matrix(Yt)

# branch from and to bus
f = np.array([0, 3, 4, 2, 4])
t = np.array([4, 1, 1, 3, 3])

# slack, pv and pq buses
slack = np.array([0])  # The slack bus does not have to be the 0-indexed bus
pq = np.array([1, 3, 4])
pv = np.array([2])

# measurements
se_input = StateEstimationInput()

se_input.p_inj = np.array([ 3.948e+00, -8.000e+00,  4.400e+00, -6.507e-06, -1.407e-05])
se_input.p_inj_idx = np.arange(len(se_input.p_inj))
se_input.p_inj_weight = np.full(len(se_input.p_inj), 0.01)

se_input.q_inj = np.array([ 1.143e+00, -2.800e+00,  2.975e+00,  6.242e-07,  1.957e-06])
se_input.q_inj_idx = np.arange(len(se_input.q_inj))
se_input.q_inj_weight = np.full(len(se_input.q_inj), 0.01)

se_input.vm_m = np.array([0.834, 1.019, 0.974])
se_input.vm_m_idx = pq
se_input.vm_m_weight = np.full(len(se_input.vm_m), 0.01)

# First time will be slow due to compilation
start = time.time()
v_sol, err, converged, results = solve_se_lm(Ybus, Yf, Yt, f, t, se_input, slack, pq, pv)
print("compilation + execution time:", time.time() - start)
bdd_validation(results, m=len(se_input.measurements), n=Ybus.shape[0] + len(pq) + len(pv))

# But then it will be very performant
start = time.time()
v_sol, err, converged, results = solve_se_lm(Ybus, Yf, Yt, f, t, se_input, slack, pq, pv)
print("Execution time:", time.time() - start)

# False data injection
se_input.vm_m[1] -= 0.025
se_input.vm_m[2] += 0.025
v_sol, err, converged, results = solve_se_lm(Ybus, Yf, Yt, f, t, se_input, slack, pq, pv)
print("-------------After False Data Injection-------------")
bdd_validation(results, m=len(se_input.measurements), n=Ybus.shape[0] + len(pq) + len(pv))

Acknowledge

This work was supported by the U.S. Department of Energy (DOE) under award DE-OE0000895 and the Sandia National Laboratories’ directed R&D project #222444.