我要使用sysid
中的GEKKO
描述100个数据点。在某个时间点(在这种情况下,t = 50),数据发生明显变化,并且预测不再准确。我正在尝试包含一个if语句,该语句对实际值与预测值进行评估,如果预测值比模型大x倍,则生成一个新模型(new yp
)。这是我的示例代码。循环在每个时间点继续评估yp
,但现在应该评估yp_new
。
from gekko import GEKKO
import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
# load data and parse into columns
t = np.linspace(0,1,101)
u = np.linspace(0,1,101)
y = np.zeros(len(t))
y[:50] = np.sin(u[:50])
y[50:] = np.exp(u[50:]/500)
# generate time-series model
m = GEKKO(remote=False)
# system identification
na = 2 # output coefficients
nb = 2 # input coefficients
yp,p,K = m.sysid(t,u,y,na,nb,diaglevel=1)
print(yp)
for i in range(len(t)):
difference = np.abs((yp[i]-y[i])/max(0.01,y[i]))
if difference>=0.2: #If the difference is >20%
yp_new,p_new,K = m.sysid(t,u,y,na,nb,diaglevel=0)
print('Recalculating at i = ' + str(i))
print(yp_new)
plt.figure()
plt.subplot(2,1,1)
plt.plot(t,u)
plt.legend([r'$u_0$',r'$u_1$'])
plt.ylabel('MVs')
plt.subplot(2,1,2)
plt.plot(t,y)
plt.plot(t,yp)
plt.plot(t,y)
plt.plot(t,yp_new)
plt.show()
最佳答案
您需要将yp
更新为新的yp值yp = yp_new
,否则在进行下一个系统识别时只需返回yp
。但是,用于重做sysid
计算的数据与先前使用的数据相同,因此模型预测中没有变化。您是否仅尝试使用最新数据(例如yp_new,p_new,K = m.sysid(t[i:],u[i:],y[i:],na,nb)
)更新时间序列模型?
该模型当前更新的周期与原始时间序列模型的预测不一致。
Recalculating at i = 10
Recalculating at i = 11
Recalculating at i = 12
Recalculating at i = 13
Recalculating at i = 14
Recalculating at i = 15
Recalculating at i = 16
Recalculating at i = 17
Recalculating at i = 18
Recalculating at i = 19
Recalculating at i = 20
Recalculating at i = 21
Recalculating at i = 22
Recalculating at i = 23
Recalculating at i = 24
Recalculating at i = 25
Recalculating at i = 40
Recalculating at i = 41
Recalculating at i = 42
Recalculating at i = 43
Recalculating at i = 44
Recalculating at i = 45
Recalculating at i = 46
Recalculating at i = 47
Recalculating at i = 48
Recalculating at i = 49
Recalculating at i = 50
Recalculating at i = 51
Recalculating at i = 52
如果仅包括最新数据,则仅在周期10、42、46和48进行重新计算。
from gekko import GEKKO
import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
# load data and parse into columns
t = np.linspace(0,1,101)
u = np.linspace(0,1,101)
y = np.zeros(len(t))
y[:50] = np.sin(u[:50])
y[50:] = np.exp(u[50:]/500)
# generate time-series model
m = GEKKO(remote=False)
# system identification
na = 2 # output coefficients
nb = 2 # input coefficients
yp,p,K = m.sysid(t,u,y,na,nb)
yp_init = yp.copy()
print(yp)
j = 0
for i in range(len(t)):
difference = np.abs((yp[i-j]-y[i])/max(0.01,y[i]))
if difference>=0.2: #If the difference is >20%
j = i # get cycle where the last update occurred
yp,p,K = m.sysid(t[i:],u[i:],y[i:],na,nb)
print('Recalculating at i = ' + str(i))
plt.figure()
plt.subplot(2,1,1)
plt.plot(t,u)
plt.legend([r'$u_0$',r'$u_1$'])
plt.ylabel('MVs')
plt.subplot(2,1,2)
plt.plot(t,y)
plt.plot(t,yp_init)
plt.plot(t,y)
plt.plot(t[j:],yp)
plt.show()