-
Notifications
You must be signed in to change notification settings - Fork 4
/
helm_as_a_function_v3.py
188 lines (160 loc) · 6.61 KB
/
helm_as_a_function_v3.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from scipy.sparse import csc_matrix
from scipy.sparse.linalg import spsolve, factorized
def conv(A, B, c, i, tipus):
"""
3 types of convolution needed
:param A: vector 1
:param B: vector 2
:param c: current depth
:param i: selected bus
:param tipus: kind of convolution
:return: result of convolution
"""
if tipus == 1:
suma = [np.conj(A[k, i]) * B[c - k, i] for k in range(1, c + 1)]
return sum(suma)
elif tipus == 2:
suma = [A[k, i] * B[c - 1 - k, i] for k in range(1, c)]
return sum(suma)
elif tipus == 3:
suma = [A[k, i] * np.conj(B[c - k, i]) for k in range(1, c)]
return sum(suma)
def helm_josep(n, Y, vec_Y0, V0, S0, vec_shunts, pq, pv, vd, n_coeff=30):
"""
:param n: number of buses, including the slack bus (expected index 0)
:param Y: Admittance matrix
:param vec_Y0: vector of series admittances of the branches connected to the slack bus (length n-1)
:param V0: vector of set voltages (length n)
:param S0: vector of set power injections (length n)
:param vec_shunts: vector of nodal shunts (length n)
:param pq: list of PQ node indices
:param pv: list of PV bus indices
:param vd: list of SLack bus indices
:param n_coeff: number of coefficients
:return: HELM voltage
"""
pqpv = np.r_[pq, pv]
np.sort(pqpv)
# --------------------------- PREPARING IMPLEMENTATION
vec_V = np.abs(V0[pqpv]) - 1.0 # data of voltage magnitude
vec_W = vec_V * vec_V # voltage magnitude squared
U = np.zeros((n_coeff, n - 1), dtype=complex) # voltages
U_re = np.zeros((n_coeff, n - 1), dtype=complex) # real part of voltages
U_im = np.zeros((n_coeff, n - 1), dtype=complex) # imaginary part of voltages
X = np.zeros((n_coeff, n - 1), dtype=complex) # X=1/conj(U)
X_re = np.zeros((n_coeff, n - 1), dtype=complex) # real part of X
X_im = np.zeros((n_coeff, n - 1), dtype=complex) # imaginary part of X
Q = np.zeros((n_coeff, n - 1), dtype=complex) # unknown reactive powers
npq = len(pq)
npv = len(pv)
V_slack = V0[vd]
G = Y.real
B = Y.imag
vec_P = S0.real
vec_Q = S0.imag
dimensions = 2 * npq + 3 * npv # number of unknowns
# .......................GUIDING VECTOR
lx = 0
index_Ure = []
index_Uim = []
index_Q = []
for i in range(n - 1):
index_Ure.append(lx)
index_Uim.append(lx + 1)
if i + 1 in pq:
lx = lx + 2
else:
index_Q.append(lx + 2)
lx = lx + 3
# .......................GUIDING VECTOR. DONE
# .......................CALCULATION OF TERMS [0]
Y = csc_matrix(Y)
U[0, :] = spsolve(Y, vec_Y0)
X[0, :] = 1 / np.conj(U[0, :])
U_re[0, :] = np.real(U[0, :])
U_im[0, :] = np.imag(U[0, :])
X_re[0, :] = np.real(X[0, :])
X_im[0, :] = np.imag(X[0, :])
# .......................CALCULATION OF TERMS [0]. DONE
# .......................CALCULATION OF TERMS [1]
valor = np.zeros(n - 1, dtype=complex)
valor[pq - 1] = (V_slack - 1) * vec_Y0[pq - 1, 0] + (vec_P[pq - 1, 0] - vec_Q[pq - 1, 0] * 1j) * X[0, pq - 1] + U[
0, pq - 1] * vec_shunts[pq - 1, 0]
valor[pv - 1] = (V_slack - 1) * vec_Y0[pv - 1, 0] + (vec_P[pv - 1, 0]) * X[0, pv - 1] + U[0, pv - 1] * vec_shunts[
pv - 1, 0]
RHSx = np.zeros((3, n - 1), dtype=float)
RHSx[0, pq - 1] = valor[pq - 1].real
RHSx[1, pq - 1] = valor[pq - 1].imag
RHSx[2, pq - 1] = np.nan # to later delete
RHSx[0, pv - 1] = valor[pv - 1].real
RHSx[1, pv - 1] = valor[pv - 1].imag
RHSx[2, pv - 1] = vec_W[pv - 1, 0] - 1
rhs = np.matrix.flatten(RHSx, 'f')
rhs = rhs[~np.isnan(rhs)] # delete dummy cells
mat = np.zeros((dimensions, 2 * (n - 1) + npv), dtype=complex) # constant matrix
k = 0 # index that will go through the rows
for i in range(n - 1): # fill the matrix
lx = 0
for j in range(n - 1):
mat[k, lx] = G[i, j]
mat[k + 1, lx] = B[i, j]
mat[k, lx + 1] = -B[i, j]
mat[k + 1, lx + 1] = G[i, j]
if (j == i) and (i + 1 in pv) and (j + 1 in pv):
mat[k + 2, lx] = 2 * U_re[0, i]
mat[k + 2, lx + 1] = 2 * U_im[0, i]
mat[k, lx + 2] = -X_im[0, i]
mat[k + 1, lx + 2] = X_re[0, i]
lx = lx + 2 if (j + 1 in pq) else lx + 3
k = k + 2 if (i + 1 in pq) else k + 3
mat_factorized = factorized(csc_matrix(mat))
lhs = mat_factorized(rhs)
U_re[1, :] = lhs[index_Ure]
U_im[1, :] = lhs[index_Uim]
Q[0, pv - 1] = lhs[index_Q]
U[1, :] = U_re[1, :] + U_im[1, :] * 1j
X[1, :] = (-X[0, :] * np.conj(U[1, :])) / np.conj(U[0, :])
X_re[1, :] = np.real(X[1, :])
X_im[1, :] = np.imag(X[1, :])
# .......................CALCULATION OF TERMS [1]. DONE
# .......................CALCULATION OF TERMS [>=2]
for c in range(2, n_coeff): # c defines the current depth
valor[pq - 1] = (vec_P[pq - 1, 0] - vec_Q[pq - 1, 0] * 1j) * X[c - 1, pq - 1] + U[c - 1, pq - 1] * vec_shunts[
pq - 1, 0]
valor[pv - 1] = conv(X, Q, c, pv - 1, 2) * (-1) * 1j + U[c - 1, pv - 1] * vec_shunts[pv - 1, 0] + \
X[c - 1, pv - 1] * vec_P[pv - 1, 0]
RHSx[0, pq - 1] = valor[pq - 1].real
RHSx[1, pq - 1] = valor[pq - 1].imag
RHSx[2, pq - 1] = np.nan # per poder-ho eliminar bé, dummy
RHSx[0, pv - 1] = valor[pv - 1].real
RHSx[1, pv - 1] = valor[pv - 1].imag
RHSx[2, pv - 1] = -conv(U, U, c, pv - 1, 3)
rhs = np.matrix.flatten(RHSx, 'f')
rhs = rhs[~np.isnan(rhs)]
lhs = mat_factorized(rhs)
U_re[c, :] = lhs[index_Ure]
U_im[c, :] = lhs[index_Uim]
Q[c - 1, pv - 1] = lhs[index_Q]
U[c, :] = U_re[c, :] + U_im[c, :] * 1j
X[c, range(n - 1)] = -conv(U, X, c, range(n - 1), 1) / np.conj(U[0, range(n - 1)])
X_re[c, :] = np.real(X[c, :])
X_im[c, :] = np.imag(X[c, :])
# .......................CALCULATION OF TERMS [>=2]. DONE
# sum the coefficients
V = V0.copy()
V[pqpv] = U.sum(axis=0)
return V
if __name__ == '__main__':
V2 = helm_josep(n,
Y,
vec_Y0,
V0,
S0,
vec_shunts,
pq,
pv,
vd,
n_coeff=30)