def test_interpolation_speed():
import xarray as xr
da = xr.DataArray(np.sin(0.3 * np.arange(20).reshape(5, 4)),
[('x', np.arange(5)), ('y', [0.1, 0.2, 0.3, 0.4])])
x = xr.DataArray([0.5, 1.5, 2.5], dims='z')
y = xr.DataArray([0.15, 0.25, 0.35], dims='z')
da.interp(x=x, y=y)
site = ParqueFicticioSite()
x, y = site.initial_position.T
X, Y, x_j, y_j, h_j = HorizontalGrid()(x, y, 70)
wd = [270] # site.default_wd
ws = site.default_ws
res1, t_lst = timeit(site.interp_funcs['A'])(
(x_j, y_j, h_j, x_j * 0 + 270))
print(res1.shape)
res2, t_lst = timeit(lambda x, y, z, sec: site._ds.A.interp(
x=xr.DataArray(x, dims='z'),
y=xr.DataArray(y, dims='z'),
z=xr.DataArray(z, dims='z'),
sec=xr.DataArray(sec, dims='z')).data)(x_j, y_j, h_j, x_j * 0 + 10)
npt.assert_array_almost_equal(res1, res2)
if 0:
c = plt.contourf(X, Y, res1.reshape(X.shape))
plt.colorbar(c)
plt.figure()
c = plt.contourf(X, Y, res2.reshape(X.shape))
plt.colorbar(c)
plt.show()
def test_multiprocessing_wd():
pool = multiprocessing.Pool(4)
aep1, t_lst1 = timeit(aep_wd, min_runs=1)((wt_x, wt_y, wd_lst))
aep2, t_lst2 = timeit(aep_all_multiprocessing, min_runs=1)(pool, wt_x, wt_y)
t1, t2 = np.mean(t_lst1), np.mean(t_lst2)
print("1 CPU: %.2fs, %d CPUs: %.2fs, speedup: %d%%" % (t1, pool._processes, t2, (t1 - t2) / t1 * 100))
npt.assert_almost_equal(aep1, aep2 / 36)
def compare_speed():
x = np.arange(0, 100)
y = np.arange(10, 100)
z = np.arange(30, 33)
wd = np.arange(360)
ws = np.arange(3, 26)
V = np.random.rand(len(x), len(y), len(z), len(wd), len(ws))
print(V.shape)
for x in [x, np.r_[x[:-1], 100]]:
da = xr.DataArray(V, coords=[('x', x), ('y', y), ('z', z), ('wd', wd), ('ws', ws)])
coords = xr.Dataset(coords={'wd': np.arange(360), 'ws': [9, 10], 'i': np.arange(16),
'x': ('i', np.linspace(0, 9, 16)),
'y': ('i', np.linspace(10, 20, 16)),
'z': ('i', np.linspace(30, 32, 16))})
x1, t1 = timeit(da.sel_interp_all, verbose=True)(coords)
I, L, K = len(coords.x), len(coords.wd), len(coords.ws)
def interp():
c = np.array([coords.x.data.repeat(L * K), coords.y.data.repeat(L * K), coords.z.data.repeat(L * K),
np.tile(coords.wd.data.repeat(K), I), np.tile(coords.ws.data, I * L)]).T
return GridInterpolator([x, y, z, wd, ws], da.values)(c)
x2, t2 = timeit(interp, verbose=True, line_profile=0, profile_funcs=[GridInterpolator.__call__])()
npt.assert_array_almost_equal(x1, x2.reshape(16, len(coords.wd), len(coords.ws)))
print(np.mean(t1) / np.mean(t2))
def test_multiprocessing_xy():
pool = multiprocessing.Pool(2)
arg_lst = [(np.array(wt_x) + i, wt_y) for i in range(4)]
aep1, t_lst1 = timeit(lambda arg_lst: [aep_xy(arg) for arg in arg_lst], min_time=0, min_runs=1)(arg_lst)
aep2, t_lst2 = timeit(lambda arg_lst: pool.map(aep_xy, arg_lst), min_time=0, min_runs=1)(arg_lst)
t1, t2 = np.mean(t_lst1), np.mean(t_lst2)
print("1 CPU: %.2fs, %d CPUs: %.2fs, speedup: %d%%" % (t1, pool._processes, t2, (t1 - t2) / t1 * 100))
npt.assert_almost_equal(aep1, aep2)
def test_multiprocessing_wfm_xy():
pool = multiprocessing.Pool(2)
arg_lst = [(get_wfm(grad=False), np.array(wt_x) + i, wt_y) for i in range(4)]
aep1, t_lst1 = timeit(lambda arg_lst: [aep_wfm_xy(arg) for arg in arg_lst])(arg_lst)
aep2, t_lst2 = timeit(lambda arg_lst: pool.map(aep_wfm_xy, arg_lst))(arg_lst)
t1, t2 = np.mean(t_lst1), np.mean(t_lst2)
if debug:
print("1 CPU: %.2fs, %d CPUs: %.2fs, speedup: %d%%" % (t1, pool._processes, t2, (t1 - t2) / t1 * 100))
npt.assert_almost_equal(aep1, aep2)
def test_multiprocessing_wd(pool):
# compare result of vectorized call 12wd with result of multiprocessing 1wd/cpu
# Slow down is expected
aep1, t_lst1 = timeit(aep_wd, min_runs=1)((wt_x, wt_y, wd_lst))
aep2, t_lst2 = timeit(aep_all_multiprocessing, min_runs=1)(pool, wt_x, wt_y)
t1, t2 = np.mean(t_lst1), np.mean(t_lst2)
if debug:
print("1 CPU, 12wd/CPU: %.2fs, %d CPUs, 1wd/CPU: %.2fs, speedup: %d%%" %
(t1, pool._processes, t2, (t1 - t2) / t1 * 100))
npt.assert_almost_equal(aep1, aep2 / len(wd_lst))
def test_GCL_ex80():
site = Hornsrev1Site()
x, y = site.initial_position.T
windTurbines = V80()
wfm = PropagateDownwind(site,
windTurbines,
wake_deficitModel=GCLDeficit(),
superpositionModel=LinearSum())
if 0:
windTurbines.plot(x, y)
plt.show()
sim_res = timeit(wfm.__call__,
line_profile=0,
profile_funcs=[get_dU],
verbose=0)(x, y, ws=np.arange(10, 15))[0]
# test that the result is equal to previuos runs (no evidens that these number are correct)
aep_ref = 1055.956615887197
npt.assert_almost_equal(
sim_res.aep_ilk(normalize_probabilities=True).sum(), aep_ref, 5)
sim_res = wfm(x, y, ws=np.arange(3, 10))
npt.assert_array_almost_equal(
sim_res.aep_ilk(normalize_probabilities=True).sum(), 261.6143039016946,
5)
def test_fuga():
# move turbine 1 600 300
wt_x = [-250, 600, -500, 0, 500, -250, 250]
wt_y = [433, 300, 0, 0, 0, -433, -433]
wts = HornsrevV80()
path = tfp + 'fuga/2MW/Z0=0.03000000Zi=00401Zeta0=0.00E+0/'
site = UniformSite([1, 0, 0, 0], ti=0.075)
wake_model = Fuga(path, site, wts)
res, _ = timeit(wake_model.__call__, verbose=0, line_profile=0,
profile_funcs=[FugaDeficit.interpolate, LUTInterpolator.__call__, GridInterpolator.__call__])(x=wt_x, y=wt_y, wd=[30], ws=[10])
npt.assert_array_almost_equal(res.WS_eff_ilk.flatten(),
[10.002683492812844, 10.0, 8.413483643142389, 10.036952526815286,
9.371203842245153, 8.437429367715435, 8.012759083790058], 8)
npt.assert_array_almost_equal(res.ct_ilk.flatten(), [0.79285509, 0.793, 0.80641348, 0.79100456,
0.80180315, 0.80643743, 0.80601276], 8)
x_j = np.linspace(-1500, 1500, 500)
y_j = np.linspace(-1500, 1500, 300)
wake_model = Fuga(path, site, wts)
sim_res = wake_model(wt_x, wt_y, wd=[30], ws=[10])
flow_map70 = sim_res.flow_map(HorizontalGrid(x_j, y_j, h=70))
flow_map73 = sim_res.flow_map(HorizontalGrid(x_j, y_j, h=73))
X, Y = flow_map70.XY
Z70 = flow_map70.WS_eff_xylk[:, :, 0, 0]
Z73 = flow_map73.WS_eff_xylk[:, :, 0, 0]
if 0:
flow_map70.plot_wake_map(levels=np.arange(6, 10.5, .1))
plt.plot(X[0], Y[140])
plt.figure()
plt.plot(X[0], Z70[140, :], label="Z=70m")
plt.plot(X[0], Z73[140, :], label="Z=73m")
plt.plot(X[0, 100:400:10], Z70[140, 100:400:10], '.')
print(list(np.round(Z70.data[140, 100:400:10], 4)))
print(list(np.round(Z73.data[140, 100:400:10], 4)))
plt.legend()
plt.show()
npt.assert_array_almost_equal(
Z70[140, 100:400:10],
[10.0467, 10.0473, 10.0699, 10.0093, 9.6786, 7.8589, 6.8539, 9.2199, 9.9837, 10.036, 10.0796,
10.0469, 10.0439, 9.1866, 7.2552, 9.1518, 10.0449, 10.0261, 10.0353, 9.9256, 9.319, 8.0062,
6.789, 8.3578, 9.9393, 10.0332, 10.0183, 10.0186, 10.0191, 10.0139], 4)
npt.assert_array_almost_equal(
Z73[140, 100:400:10],
[10.0463, 10.0468, 10.0688, 10.0075, 9.6778, 7.9006, 6.9218, 9.228, 9.9808, 10.0354, 10.0786,
10.0464, 10.0414, 9.1973, 7.3099, 9.1629, 10.0432, 10.0257, 10.0344, 9.9236, 9.3274, 8.0502,
6.8512, 8.3813, 9.9379, 10.0325, 10.018, 10.0183, 10.019, 10.0138], 4)
def test_neighbour_farm_speed():
# import and setup site and windTurbines
site = IEA37Site(16)
# setup current, neighbour and all positions
wt_x, wt_y = site.initial_position.T
neighbour_x, neighbour_y = wt_x - 4000, wt_y
all_x, all_y = np.r_[wt_x, neighbour_x], np.r_[wt_y, neighbour_y]
windTurbines = WindTurbines.from_WindTurbines([IEA37_WindTurbines(), IEA37_WindTurbines()])
windTurbines._names = ["Current wind farm", "Neighbour wind farm"]
types = [0] * len(wt_x) + [1] * len(neighbour_x)
wf_model = PropagateDownwind(site, windTurbines,
wake_deficitModel=BastankhahGaussianDeficit(use_effective_ws=True),
superpositionModel=LinearSum())
# Consider wd=270 +/- 30 deg only
wd_lst = np.arange(240, 301)
sim_res, t = timeit(wf_model, verbose=False)(all_x, all_y, type=types, ws=9.8, wd=wd_lst)
if 1:
ext = 100
flow_box = wf_model(neighbour_x, neighbour_y, wd=wd_lst).flow_box(
x=np.linspace(min(wt_x) - ext, max(wt_x) + ext, 53),
y=np.linspace(min(wt_y) - ext, max(wt_y) + ext, 51),
h=[100, 110, 120])
wake_site = XRSite.from_flow_box(flow_box)
wake_site.save('tmp.nc')
else:
wake_site = XRSite.load('tmp.nc')
wf_model_wake_site = PropagateDownwind(wake_site, windTurbines,
wake_deficitModel=BastankhahGaussianDeficit(use_effective_ws=True),
superpositionModel=LinearSum())
sim_res_wake_site, _ = timeit(wf_model_wake_site, verbose=False)(wt_x, wt_y, ws=9.8, wd=wd_lst)
npt.assert_allclose(sim_res.aep().sel(wt=np.arange(len(wt_x))).sum(), sim_res_wake_site.aep().sum(), rtol=0.0005)
npt.assert_array_almost_equal(sim_res.aep().sel(wt=np.arange(len(wt_x))), sim_res_wake_site.aep(), 2)
def test_fuga():
# move turbine 1 600 300
wt_x = [-250, 600, -500, 0, 500, -250, 250]
wt_y = [433, 300, 0, 0, 0, -433, -433]
wts = HornsrevV80()
path = tfp + 'fuga/2MW/Z0=0.03000000Zi=00401Zeta0=0.00E+00/'
site = UniformSite([1, 0, 0, 0], ti=0.075)
wake_model = Fuga(path, site, wts)
res, _ = timeit(wake_model.__call__,
verbose=0,
line_profile=0,
profile_funcs=[
FugaDeficit.interpolate, LUTInterpolator.__call__,
GridInterpolator.__call__
])(x=wt_x, y=wt_y, wd=[30], ws=[10])
npt.assert_array_almost_equal(res.WS_eff_ilk.flatten(), [
10.00669629, 10., 8.47606501, 10.03143097, 9.37288077, 8.49301941,
8.07462708
], 8)
npt.assert_array_almost_equal(res.ct_ilk.flatten(), [
0.7926384, 0.793, 0.80647607, 0.79130273, 0.80177967, 0.80649302,
0.80607463
], 8)
x_j = np.linspace(-1500, 1500, 500)
y_j = np.linspace(-1500, 1500, 300)
wake_model = Fuga(path, site, wts)
sim_res = wake_model(wt_x, wt_y, wd=[30], ws=[10])
flow_map70 = sim_res.flow_map(HorizontalGrid(x_j, y_j, h=70))
flow_map73 = sim_res.flow_map(HorizontalGrid(x_j, y_j, h=73))
X, Y = flow_map70.XY
Z70 = flow_map70.WS_eff_xylk[:, :, 0, 0]
Z73 = flow_map73.WS_eff_xylk[:, :, 0, 0]
if 0:
flow_map70.plot_wake_map(levels=np.arange(6, 10.5, .1))
plt.plot(X[0], Y[140])
plt.figure()
plt.plot(X[0], Z70[140, :], label="Z=70m")
plt.plot(X[0], Z73[140, :], label="Z=73m")
plt.plot(X[0, 100:400:10], Z70[140, 100:400:10], '.')
print(list(np.round(Z70.data[140, 100:400:10], 4)))
print(list(np.round(Z73.data[140, 100:400:10], 4)))
plt.legend()
plt.show()
npt.assert_array_almost_equal(Z70[140, 100:400:10], [
10.0407, 10.0438, 10.0438, 10.013, 9.6847, 7.8787, 6.9561, 9.2251,
9.9686, 10.0382, 10.0498, 10.0569, 10.0325, 9.1787, 7.4004, 9.1384,
10.0329, 10.0297, 10.0232, 9.9265, 9.3163, 8.0768, 6.8858, 8.3754,
9.9592, 10.0197, 10.0118, 10.0141, 10.0118, 10.0095
], 4)
npt.assert_array_almost_equal(Z73[140, 100:400:10], [
10.0404, 10.0435, 10.0433, 10.0113, 9.6836, 7.9206, 7.0218, 9.2326,
9.9665, 10.0376, 10.0494, 10.0563, 10.0304, 9.1896, 7.4515, 9.15,
10.0317, 10.0294, 10.0226, 9.9245, 9.3252, 8.1192, 6.9462, 8.3988,
9.9574, 10.0194, 10.0117, 10.014, 10.0117, 10.0094
], 4)
