I am modelling an MPC to control a fridge and keep the temperature within a given interval while minimizing the cost. I am using GEKKO to model my algorithm.
I wrote the following code. First, I identified my model using sensor data from my system (I used the function sysif from GEKKO). Then I built an ARX model (using the arx function in GEKKO) that becomes the results of sysid() as input.
I am trying to write a "dummy" algorithms to test is locally before implementing it to a Pi.
I get the following error :
KeyError Traceback (most recent call last)
<ipython-input-13-108148376700> in <module>
107 #Solve the optimization problem.
108
--> 109 m.solve()
~/opt/anaconda3/lib/python3.8/site-packages/gekko/gekko.py in solve(self, disp, debug, GUI, **kwargs)
2214 if timing == True:
2215 t = time.time()
-> 2216 self.load_JSON()
2217 if timing == True:
2218 print('load JSON', time.time() - t)
~/opt/anaconda3/lib/python3.8/site-packages/gekko/gk_post_solve.py in load_JSON(self)
48 vp.__dict__[o] = dpred
49 else: #everything besides value, dpred and pred
---> 50 vp.__dict__[o] = data[vp.name][o]
51 for vp in self._variables:
52 if vp.type != None: #(FV/MV/SV/CV) not Param or Var
KeyError: 'int_p6'
And this is my code
from gekko import GEKKO
import numpy as np
import matplotlib.pyplot as plt
m = GEKKO(remote = True)
#initialize variables
#Room Temprature:
T_external = [23,23,23,23,23.5,23.5,23.4,23.5,23.9,23.7,\
23,23.9,23.9,23.4,23.9,24,23.6,23.7,23.8,\
23,23,23,23,23]
# Temprature Lower Limit:
temp_low = 10*np.ones(24)
# Temprature Upper Limit:
temp_upper = 12*np.ones(24)
#Hourly Energy prices:
TOU_v = [39.09,34.93,38.39,40.46,40.57,43.93,25,11,9,24,51.28,45.22,45.72,\
36,35.03,10,12,13,32.81,42.55,8,29.58,29.52,29.52]
###########################################
#System Identification:
#Time
t = np.linspace(0,10,117)
#State of the Fridge
ud = np.append(np.zeros(78) ,np.ones(39),0)
#Temprature Data
y = [14.600000000000001,14.600000000000001,14.700000000000001,14.700000000000001,14.700000000000001,\
14.700000000000001,14.700000000000001,14.700000000000001,14.700000000000001,14.700000000000001,\
14.700000000000001,14.700000000000001,14.700000000000001,14.8,14.8,14.8,14.8,14.8,14.8,14.8,14.8,\
14.8,14.8,14.9,14.9,14.9,14.9,14.9,14.9,14.9,15,15,15,15,15,15,15,15,15,15,15,15,15.100000000000001,\
15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,\
15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,\
15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,\
15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,\
15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,\
15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,\
15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,
15,15,15,15,15,15,15,15,15,15,14.9,14.9,14.9,14.9,14.8,14.9,14.8,14.8,14.8,14.8,14.8,14.8,\
14.8,14.700000000000001,14.8,14.700000000000001,14.700000000000001,14.700000000000001,\
14.700000000000001,14.700000000000001,14.700000000000001,14.700000000000001,\
14.700000000000001,14.600000000000001,14.600000000000001,14.600000000000001,\
14.600000000000001,14.600000000000001,14.60]
na = 1 # output coefficients
nb = 1 # input coefficients
print('Identification')
yp,p,K = m.sysid(t,ud,y,na,nb,objf=10000,scale=False,diaglevel=1)
#create control ARX model:
y = m.Array(m.CV,1)
uc = m.Array(m.MV,1)
m.arx(p,y,uc)
# rename CVs
T= y[0]
# rename MVs
uc = uc[0]
# steady state initialization
m.options.IMODE = 1
m.solve(disp=True)
###########################################
#Parameter
P = m.Param(value =100) #power
TL = m.Param(value=temp_low)
TH = m.Param(value=temp_upper)
c = m.Param(value=TOU_v)
# Manipilated variable:
u = m.MV(lb=0, ub=1, integer=True)
u.STATUS = 1 # allow optimizer to change the variable to attein the optimum.
# Controlled Variable (Affected with changes in the manipulated variable)
T = m.CV(value=11) # Temprature will start at 11.
# Soft constraints on temprature.
eH = m.CV(value=0)
eL = m.CV(value=0)
eH.SPHI=0 #Set point high for linear error model.
eH.WSPHI=100 #Objective function weight on upper set point for linear error model.
eH.WSPLO=0 # Objective function weight on lower set point for linear error model
eH.STATUS =1 # eH : Error is considered in the objective function.
eL.SPLO=0
eL.WSPHI=0
eL.WSPLO=100
eL.STATUS = 1
#Linear error (Deviation from the limits)
m.Equations([eH==T-TH,eL==T-TL])
#Objective : minimize the costs.
m.Minimize(c*P*u)
#Optimizer Options.
m.options.IMODE = 6 # MPC mode in Gekko.
m.options.NODES = 2 # Collocation nodes.
m.options.SOLVER = 1 # APOT solver for mixed integer linear programming.
m.time = np.linspace(0,23,24)
#Solve the optimization problem.
m.solve()
The problem is with:
T = m.CV(value=11) # Temperature will start at 11.
You are redefining the T variable but it stores both internally. If you need to re-initialize to 11 then use T.value=11. Also, I added the eH and eL variables before the steady state initialization. Here is a complete script that runs successfully.
from gekko import GEKKO
import numpy as np
import matplotlib.pyplot as plt
m = GEKKO(remote = True)
#initialize variables
#Room Temprature:
T_external = [23,23,23,23,23.5,23.5,23.4,23.5,23.9,23.7,\
23,23.9,23.9,23.4,23.9,24,23.6,23.7,23.8,\
23,23,23,23,23]
# Temprature Lower Limit:
temp_low = 10*np.ones(24)
# Temprature Upper Limit:
temp_upper = 12*np.ones(24)
#Hourly Energy prices:
TOU_v = [39.09,34.93,38.39,40.46,40.57,43.93,25,11,9,24,51.28,45.22,45.72,\
36,35.03,10,12,13,32.81,42.55,8,29.58,29.52,29.52]
###########################################
#System Identification:
#Time
t = np.linspace(0,10,117)
#State of the Fridge
ud = np.append(np.zeros(78) ,np.ones(39),0)
#Temprature Data
y = [14.600000000000001,14.600000000000001,14.700000000000001,14.700000000000001,14.700000000000001,\
14.700000000000001,14.700000000000001,14.700000000000001,14.700000000000001,14.700000000000001,\
14.700000000000001,14.700000000000001,14.700000000000001,14.8,14.8,14.8,14.8,14.8,14.8,14.8,14.8,\
14.8,14.8,14.9,14.9,14.9,14.9,14.9,14.9,14.9,15,15,15,15,15,15,15,15,15,15,15,15,15.100000000000001,\
15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,\
15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,\
15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,\
15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,\
15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,\
15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,\
15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,15.100000000000001,
15,15,15,15,15,15,15,15,15,15,14.9,14.9,14.9,14.9,14.8,14.9,14.8,14.8,14.8,14.8,14.8,14.8,\
14.8,14.700000000000001,14.8,14.700000000000001,14.700000000000001,14.700000000000001,\
14.700000000000001,14.700000000000001,14.700000000000001,14.700000000000001,\
14.700000000000001,14.600000000000001,14.600000000000001,14.600000000000001,\
14.600000000000001,14.600000000000001,14.60]
na = 1 # output coefficients
nb = 1 # input coefficients
print('Identification')
yp,p,K = m.sysid(t,ud,y,na,nb,objf=10000,scale=False,diaglevel=1)
#create control ARX model:
y = m.Array(m.CV,1)
uc = m.Array(m.MV,1)
m.arx(p,y,uc)
# rename CVs
T= y[0]
# rename MVs
uc = uc[0]
###########################################
#Parameter
P = m.Param(value =100) #power
TL = m.Param(value=temp_low[0])
TH = m.Param(value=temp_upper[0])
c = m.Param(value=TOU_v[0])
# Manipilated variable:
u = m.MV(lb=0, ub=1, integer=True)
u.STATUS = 1 # allow optimizer to change the variable to attein the optimum.
# Controlled Variable (Affected with changes in the manipulated variable)
# Soft constraints on temprature.
eH = m.CV(value=0)
eL = m.CV(value=0)
eH.SPHI=0 #Set point high for linear error model.
eH.WSPHI=100 #Objective function weight on upper set point for linear error model.
eH.WSPLO=0 # Objective function weight on lower set point for linear error model
eH.STATUS =1 # eH : Error is considered in the objective function.
eL.SPLO=0
eL.WSPHI=0
eL.WSPLO=100
eL.STATUS = 1
#Linear error (Deviation from the limits)
m.Equations([eH==T-TH,eL==T-TL])
#Objective : minimize the costs.
m.Minimize(c*P*u)
#Optimizer Options.
# steady state initialization
m.options.IMODE = 1
m.solve(disp=True)
TL.value = temp_low
TH.value = temp_upper
c.value = TOU_v
T.value = 11 # Temprature starts at 11
m.options.IMODE = 6 # MPC mode in Gekko.
m.options.NODES = 2 # Collocation nodes.
m.options.SOLVER = 1 # APOT solver for mixed integer linear programming.
m.time = np.linspace(0,23,24)
#Solve the optimization problem.
m.solve()
Here is the controller output:
--------- APM Model Size ------------
Each time step contains
Objects : 1
Constants : 0
Variables : 9
Intermediates: 0
Connections : 2
Equations : 3
Residuals : 3
Number of state variables: 1035
Number of total equations: - 1012
Number of slack variables: - 0
---------------------------------------
Degrees of freedom : 23
----------------------------------------------
Dynamic Control with APOPT Solver
----------------------------------------------
Iter: 1 I: 0 Tm: 0.07 NLPi: 3 Dpth: 0 Lvs: 0 Obj: 6.76E+03 Gap: 0.00E+00
Successful solution
---------------------------------------------------
Solver : APOPT (v1.0)
Solution time : 8.319999999366701E-002 sec
Objective : 6763.77971670735
Successful solution
---------------------------------------------------