def test_spill():
I = np.ones(N)*50
env, spill, Qreg_rel, Qreg_inf, s, E = Res_sys_sim(I, e, s_0, s_min, s_max, env_min, d, Qreg)
# Expected output
spill_expect = np.array([0.,0.,34.,38.,38.,38.,38.,38.,38.,38.]).transpose()
# Test
assert_array_equal(spill,spill_expect)
def update_operating_policy_1(s_ref_1,s_ref_2,u_ref):
if s_ref_1 > s_ref_2:
s_ref_1 = s_ref_2
x0 = [0, u_0]
x1 = [s_ref_1, u_ref]
x2 = [s_ref_2, u_ref]
x3 = [1, u_1]
param = [x0, x1, x2, x3, u_mean]
u_frac = four_points_policy(param)/u_mean
Qreg = {'releases' : {'file_name' : 'Reservoir_operating_policy.Operating_policy_functions',
'function' : 'four_points_policy',
'param': param},
'inflows' : [],
'rel_inf' : []}
Qenv, Qspill, u, I_reg, s, E = Res_sys_sim(I_hist, e_hist, s_0, s_min, s_max, env_min, d_hist, Qreg)
TSD = (np.sum((np.maximum(d_hist-u,[0]*N))**2)).astype('int')
fig_1b.title = 'Supply vs Demand - TSD = '+str(TSD)+' ML^2'
MSV = (np.sum((np.maximum(rc-s,[0]*(N+1))))).astype('int')
fig_1c.title = 'Reservoir storage volume - MSV = '+str(MSV)+' ML'
return u_frac, Qenv, Qspill, u, I_reg, s
### Inputs ###
N = 10
I = np.ones(N)*10
e = np.ones(N)*1
s_0 = 20
s_min = 5
s_max = 100
env_min = 2
d = np.ones(N)*9
Qreg = {'releases' : [],
'inflows' : [],
'rel_inf' : []}
env, spill, Qreg_rel, Qreg_inf, s, E = Res_sys_sim(I, e, s_0, s_min, s_max, env_min, d, Qreg)
### Testing functions ###
# Regulated releases
def test_Qreg_rel():
# Expected output
Qreg_rel_expect = np.array([9.,9.,9.,9.,9.,9.,9.,8.,7.,7.]).transpose()
# Test
assert_array_equal(Qreg_rel,Qreg_rel_expect)
# Storage
def test_s():
# Expected output
s_expect = np.array([20.,18.,16.,14.,12.,10.,8.,6.,5.,5.,5.]).transpose()
# Test
def Interactive_policy_auto(N,
I_hist, e_hist,
s_0, s_min, s_max,
u_0, u_1, u_mean, u_max,
env_min, d_hist,
rc,
results1_optim,results2_optim,sol_optim):
# Function to update the release policy when clicking on the points of the Pareto front
def update_operating_policy_2(i):
u_ref,s_ref_1,s_ref_2 = sol_optim[i]
x0 = [0, u_0]
x1 = [s_ref_1, u_ref]
x2 = [s_ref_2, u_ref]
x3 = [1, u_1]
param = [x0, x1, x2, x3, u_mean]
u_frac = four_points_policy(param)/u_mean
Qreg = {'releases' : {'file_name' : 'Reservoir_operating_policy.Operating_policy_functions',
'function' : 'four_points_policy',
'param': param},
'inflows' : [],
'rel_inf' : []}
Qenv, Qspill, u, I_reg, s, E = Res_sys_sim(I_hist, e_hist, s_0, s_min, s_max, env_min, d_hist, Qreg)
MSV = (np.sum((np.maximum(rc-s,[0]*(N+1))))).astype('int')
fig_2c.title = 'Reservoir storage volume - MSV = '+str(MSV)+' ML'
TSD = (np.sum((np.maximum(d_hist-u,[0]*N))**2)).astype('int')
fig_2b.title = 'Supply vs Demand - Total squared deficit = '+str(TSD)+' ML^2'
return u_frac, Qenv, Qspill, u, I_reg, s
# Function to update the figures when clicking on the points of the Pareto front
def update_figure_2(change):
policy_function.y = update_operating_policy_2(pareto_front.selected[0])[0]
releases.y = update_operating_policy_2(pareto_front.selected[0])[3]
storage.y = update_operating_policy_2(pareto_front.selected[0])[5]
# Fig_pf: Pareto front
x_sc_pf = LinearScale();y_sc_pf = LinearScale()
x_ax_pf = Axis(label='Total squared deficit [ML^2]', scale=x_sc_pf)
y_ax_pf = Axis(label='Minimum storage violation [ML]', scale=y_sc_pf, orientation='vertical')
pareto_front = plt.scatter(results1_optim[:],results2_optim[:],
scales={'x': x_sc_pf, 'y': y_sc_pf},
colors=['deepskyblue'],
interactions={'hover':'tooltip','click': 'select'})
pareto_front.unselected_style={'opacity': 0.4}
pareto_front.selected_style={'fill': 'red', 'stroke': 'yellow', 'width': '1125px', 'height': '125px'}
def_tt = Tooltip(fields=['index','x', 'y'],
labels=['index','Water deficit', 'Min storage'],
formats=['.d','.1f', '.1f'])
pareto_front.tooltip=def_tt
fig_pf = plt.Figure(marks = [pareto_front],title = 'Interactive Pareto front',
axes=[x_ax_pf, y_ax_pf],
layout={'width': '400px', 'height': '400px'}, animation_duration=1000)
if pareto_front.selected == []:
pareto_front.selected = [0]
pareto_front.observe(update_figure_2,'selected')
# Initial simulation applting the point of the Pareto Fron selected by default
u_ref,s_ref_1,s_ref_2 = sol_optim[pareto_front.selected[0]]
x0 = [0, u_0]
x1 = [s_ref_1, u_ref]
x2 = [s_ref_2, u_ref]
x3 = [1, u_1]
param = [x0, x1, x2, x3, u_mean]
u_frac = four_points_policy(param)/u_mean
Qreg = {'releases' : {'file_name' : 'Reservoir_operating_policy.Operating_policy_functions',
'function' : 'four_points_policy',
'param': param},
'inflows' : [],
'rel_inf' : []}
Qenv, Qspill, u, I_reg, s, E = Res_sys_sim(I_hist, e_hist, s_0, s_min, s_max, env_min, d_hist, Qreg)
# Fig 2a: Policy function
s_frac = np.arange(0,1.01,0.01)
x_sc_2a = LinearScale(min=0,max=1); y_sc_2a = LinearScale(min=0,max=u_max/u_mean);
x_ax_2a = Axis(label='Storage fraction', scale=x_sc_2a);
y_ax_2a = Axis(label='Release fraction', scale=y_sc_2a, orientation='vertical')
policy_function = Lines(x = s_frac,
y = u_frac ,
colors = ['blue'],
scales = {'x': x_sc_2a, 'y': y_sc_2a})
fig_2a = plt.Figure(marks = [policy_function],
title = 'Policy function',
axes=[x_ax_2a, y_ax_2a],
layout={'width': '400px', 'height': '375px'},
animation_duration=1000,
scales={'x': x_sc_2a, 'y': y_sc_2a})
policy_function.observe(update_figure_2, ['x', 'y'])
# Fig 2b: Releases vs Demand
x_sc_2b = LinearScale(min=0,max=N); y_sc_2b = LinearScale(min=0,max=u_max);
x_ax_2b = Axis(label='week', scale=x_sc_2b); y_ax_2b = Axis(label='ML/week', scale=y_sc_2b, orientation='vertical')
demand = Bars(x = np.arange(1,N+1),
y = d_hist,
colors = ['gray'],
scales = {'x': x_sc_2b, 'y': y_sc_2b})
releases = Bars(x = np.arange(1,N+1),
y = u,
colors = ['green'],
scales = {'x': x_sc_2b, 'y': y_sc_2b})
TSD = (np.sum((np.maximum(d_hist-u,[0]*N))**2)).astype('int')
fig_2b = plt.Figure(marks = [demand, releases],
title = 'Supply vs Demand - TSD = '+str(TSD)+' ML^2',
axes=[x_ax_2b, y_ax_2b],
layout={'width': '950px', 'height': '250px'},
animation_duration=1000,
scales={'x': x_sc_2b, 'y': y_sc_2b})
releases.observe(update_figure_2, ['x', 'y'])
# Fig 2c: Storage
x_sc_2c = LinearScale(); y_sc_2c = LinearScale(min=0,max=200);
x_ax_2c = Axis(label='week', scale=x_sc_2c); y_ax_2c = Axis(label='ML', scale=y_sc_2c, orientation='vertical')
storage = Lines(x = np.arange(0,N+1),
y = s ,
colors = ['blue'],
scales = {'x': x_sc_2c, 'y': y_sc_2c},
fill = 'bottom',fill_opacities = [0.8],fill_colors = ['blue'])
max_storage = plt.plot(x=np.arange(0,N+1),
y=[s_max]*(N+1),
colors=['red'],
scales={'x': x_sc_2c, 'y': y_sc_2c})
max_storage_label = plt.label(text = ['Max storage'],
x=[0],
y=[s_max+15],
colors=['red'])
min_storage = plt.plot(np.arange(0,N+1),rc,
scales={'x': x_sc_2c, 'y': y_sc_2c},
colors=['red'],opacities = [1],
line_style = 'dashed',
fill = 'bottom',fill_opacities = [0.4],fill_colors = ['red'], stroke_width = 1)
min_storage_label = plt.label(text = ['Min storage'],
x=[0],
y=[rc[0]-10],
colors=['red'])
MSV = (np.sum((np.maximum(rc-s,[0]*(N+1))))).astype('int')
fig_2c = plt.Figure(marks = [storage,max_storage,max_storage_label,
min_storage,min_storage_label],
title = 'Reservoir storage volume - MSV = '+str(MSV)+' ML',
axes=[x_ax_2c, y_ax_2c],
layout={'width': '950px', 'height': '250px'},
animation_duration=1000,
scales={'x': x_sc_2c, 'y': y_sc_2c})
storage.observe(update_figure_2, ['x', 'y'])
return fig_pf, fig_2a,fig_2b,fig_2c
def Interactive_policy_manual(N,
I_hist, e_hist,
s_0, s_min, s_max,
u_0, u_1, u_mean, u_max,
env_min, d_hist,
rc):
#Function to update the release policy when changing the parameters with the sliders
def update_operating_policy_1(s_ref_1,s_ref_2,u_ref):
if s_ref_1 > s_ref_2:
s_ref_1 = s_ref_2
x0 = [0, u_0]
x1 = [s_ref_1, u_ref]
x2 = [s_ref_2, u_ref]
x3 = [1, u_1]
param = [x0, x1, x2, x3, u_mean]
u_frac = four_points_policy(param)/u_mean
Qreg = {'releases' : {'file_name' : 'Reservoir_operating_policy.Operating_policy_functions',
'function' : 'four_points_policy',
'param': param},
'inflows' : [],
'rel_inf' : []}
Qenv, Qspill, u, I_reg, s, E = Res_sys_sim(I_hist, e_hist, s_0, s_min, s_max, env_min, d_hist, Qreg)
TSD = (np.sum((np.maximum(d_hist-u,[0]*N))**2)).astype('int')
fig_1b.title = 'Supply vs Demand - TSD = '+str(TSD)+' ML^2'
MSV = (np.sum((np.maximum(rc-s,[0]*(N+1))))).astype('int')
fig_1c.title = 'Reservoir storage volume - MSV = '+str(MSV)+' ML'
return u_frac, Qenv, Qspill, u, I_reg, s
# Function to update the figures when changing the parameters with the sliders
def update_figure_1(change):
policy_function.y = update_operating_policy_1(s_ref_1.value,s_ref_2.value,u_ref.value)[0]
releases.y = update_operating_policy_1(s_ref_1.value,s_ref_2.value,u_ref.value)[3]
storage.y = update_operating_policy_1(s_ref_1.value,s_ref_2.value,u_ref.value)[5]
# Definition of the sliders
u_ref = widgets.FloatSlider(min=0.5, max=u_1, value=1, step=0.05,
description = 'u_ref: ',
continuous_update = False)
u_ref.observe(update_figure_1,names = 'value')
s_ref_1 = widgets.FloatSlider(min=0, max=1, value=0.25, step=0.05,
description = 's_ref_1: ',
continuous_update=False)
s_ref_1.observe(update_figure_1,names = 'value')
s_ref_2 = widgets.FloatSlider(min=0, max=1, value=0.75, step=0.05,
description = 's_ref_2: ',
continuous_update=False)
s_ref_2.observe(update_figure_1,names = 'value')
# Initial simulation applying the default slider values of the parameters
x0 = [0, u_0]
x1 = [s_ref_1.value, u_ref.value]
x2 = [s_ref_2.value, u_ref.value]
x3 = [1, u_1]
param = [x0, x1, x2, x3, u_mean]
u_frac = four_points_policy(param)/u_mean
Qreg = {'releases' : {'file_name' : 'Reservoir_operating_policy.Operating_policy_functions',
'function' : 'four_points_policy',
'param': param},
'inflows' : [],
'rel_inf' : []}
Qenv, Qspill, u, I_reg, s, E = Res_sys_sim(I_hist, e_hist, s_0, s_min, s_max, env_min, d_hist, Qreg)
### Figures ###
# Fig 1a: Policy function
s_frac = np.arange(0,1.01,0.01)
x_sc_1a = LinearScale(min=0,max=1); y_sc_1a = LinearScale(min=0,max=u_1);
x_ax_1a = Axis(label='Storage fraction', scale=x_sc_1a);
y_ax_1a = Axis(label='Release fraction', scale=y_sc_1a, orientation='vertical')
policy_function = Lines(x = s_frac,
y = u_frac,
colors = ['blue'],
scales = {'x': x_sc_1a, 'y': y_sc_1a})
fig_1a = plt.Figure(marks = [policy_function],
title = 'Policy function',
axes=[x_ax_1a, y_ax_1a],
layout={'width': '400px', 'height': '375px'},
animation_duration=1000,
scales={'x': x_sc_1a, 'y': y_sc_1a})
policy_function.observe(update_figure_1, ['x', 'y'])
# Fig 1b: Releases vs Demand
x_sc_1b = LinearScale(min=0,max=N); y_sc_1b = LinearScale(min=0,max=u_max);
x_ax_1b = Axis(label='week', scale=x_sc_1b); y_ax_1b = Axis(label='ML/week', scale=y_sc_1b, orientation='vertical')
demand = Bars(x = np.arange(1,N+1),
y = d_hist,
colors = ['gray'],
scales = {'x': x_sc_1b, 'y': y_sc_1b})
releases = Bars(x = np.arange(1,N+1),
y = u,
colors = ['green'],
scales = {'x': x_sc_1b, 'y': y_sc_1b})
TSD = (np.sum((np.maximum(d_hist-u,[0]*N))**2)).astype('int')
fig_1b = plt.Figure(marks = [demand, releases],
title = 'Supply vs Demand - TSD = '+str(TSD)+' ML^2',
axes=[x_ax_1b, y_ax_1b],
layout={'width': '950px', 'height': '250px'},
animation_duration=1000,
scales={'x': x_sc_1b, 'y': y_sc_1b})
releases.observe(update_figure_1, ['x', 'y'])
# Fig 1c: Storage
x_sc_1c = LinearScale(); y_sc_1c = LinearScale(min=0,max=200);
x_ax_1c = Axis(label='week', scale=x_sc_1c); y_ax_1c = Axis(label='ML', scale=y_sc_1c, orientation='vertical')
storage = Lines(x = np.arange(0,N+1),
y = s ,
colors = ['blue'],
scales = {'x': x_sc_1c, 'y': y_sc_1c},
fill = 'bottom',fill_opacities = [0.8],fill_colors = ['blue'])
max_storage = plt.plot(x=np.arange(0,N+1),
y=[s_max]*(N+1),
colors=['red'],
scales={'x': x_sc_1c, 'y': y_sc_1c})
max_storage_label = plt.label(text = ['Max storage'],
x=[0],
y=[s_max+15],
colors=['red'])
min_storage = plt.plot(np.arange(0,N+1),rc,
scales={'x': x_sc_1c, 'y': y_sc_1c},
colors=['red'],opacities = [1],
line_style = 'dashed',
fill = 'bottom',fill_opacities = [0.4],fill_colors = ['red'], stroke_width = 1)
min_storage_label = plt.label(text = ['Min storage'],
x=[0],
y=[rc[0]-10],
colors=['red'])
MSV = (np.sum((np.maximum(rc-s,[0]*(N+1))))).astype('int')
fig_1c = plt.Figure(marks = [storage,max_storage,max_storage_label,
min_storage,min_storage_label],
title = 'Reservoir storage volume - MSV = '+str(MSV)+' ML',
axes=[x_ax_1c, y_ax_1c],
layout={'width': '950px', 'height': '250px'},
animation_duration=1000,
scales={'x': x_sc_1c, 'y': y_sc_1c})
storage.observe(update_figure_1, ['x', 'y'])
return fig_1a,fig_1b,fig_1c, u_ref,s_ref_1,s_ref_2