Thesis---Batteries/CLF_Training_Sequential.py at main · hovsept/Thesis---Batteries · GitHub

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
# -*- SPMcoding: utf-8 -*-
"""
Created on Wed Nov 15 14:21:03 2023

@author: Hovsep Touloujian
"""

import numpy as np
from SPM_Params import *
from tqdm import tqdm
import matplotlib.pyplot as plt
import scipy.optimize
from functools import *

import torch
import torch.nn as nn
import torch.nn as F
from torchsummary import summary

n = p['nrn']+p['nrp']+p['n_sei']+4

class CLF(nn.Module):

    def __init__(self):
        super(CLF,self).__init__()

        self.linear1 = nn.Linear(n, 20,bias=False)
        self.linear2 = nn.Linear(20,10,bias=False)
        self.linear3 = nn.Linear(10,1,bias=False)


    def forward(self,x):
        V = torch.relu(self.linear1(x))
        V = torch.relu(self.linear2(V))
        V = self.linear3(V)
        return V

class U(nn.Module):
    def __init__(self):
        super(U,self).__init__()

        self.linear4 = nn.Linear(n,10,bias=False)
        self.linear5 = nn.Linear(10,10,bias=False)
        self.linear6 = nn.Linear(10,1,bias=False)

    def forward(self,x):
        u = torch.relu(self.linear4(x))
        u = torch.relu(self.linear5(u))
        # u = 10*torch.sigmoid(0.005*self.linear6(u))-5
        u = 2.5*(2/torch.pi) * torch.atan(0.05*self.linear6(u))-2.5
        # u = self.linear6(u)
        return u


clf = CLF()
u = U()

#Generate Samples in State Space
max_i_s = 2e-5
x_I = torch.transpose(torch.tensor(set_sample(max_i_s, 0, 0.3, 0),dtype=torch.float32, requires_grad=True),0,1)
x_SnG = torch.transpose(torch.tensor(set_sample(max_i_s, 0, 0.9, 0),dtype=torch.float32, requires_grad=True),0,1)
x_dS = torch.transpose(torch.tensor(set_sample(max_i_s, 0, 1, 1),dtype=torch.float32, requires_grad=True),0,1)
for i in range(200):
    x_I = torch.vstack((x_I,torch.transpose(torch.tensor(set_sample(2e-5, 0, 0.3, 0),dtype=torch.float32, requires_grad=True),0,1)))
    x_SnG = torch.vstack((x_SnG,torch.transpose(torch.tensor(set_sample(2e-5, 0, 0.9, 0),dtype=torch.float32, requires_grad=True),0,1)))
    x_dS = torch.vstack((x_dS,torch.transpose(torch.tensor(set_sample(2e-5, 0, 1, 1),dtype=torch.float32, requires_grad=True),0,1)))

a = 1e-3
loss_fn = nn.LeakyReLU(a)
loss_k = []
dVf_max_k = []
lr = 5e-4
optimizerV = torch.optim.RMSprop(clf.parameters(), lr=lr, weight_decay=0.01)
optimizerU = torch.optim.RMSprop(u.parameters(), lr=lr, weight_decay=0.01)

eps = 0.01
for t in tqdm(range(5000)):
    L_I = torch.sum(loss_fn(clf(x_I) + eps))
    L_dS = torch.sum(loss_fn(-clf(x_dS) + eps))

    dVf = []
    L_SnG = 0
    for i in range(x_SnG.size()[0]):
        dV = torch.autograd.grad(clf(x_SnG[i]),x_SnG)[0][i]
        f = torch.tensor(f_SPM(x_SnG.detach().numpy()[i],u(x_SnG[i]).detach().numpy(),p)[0],dtype=torch.float32)
        dVf.append(torch.matmul(dV,f).detach().numpy())
        L_SnG += loss_fn(torch.matmul(dV, f) + eps)

    loss_V = L_I + L_dS + L_SnG
    dVf_max_k.append(max(dVf)[0])

    optimizerV.zero_grad()
    loss_V.backward()
    optimizerV.step()

    L_I = torch.sum(loss_fn(clf(x_I) + eps))
    L_dS = torch.sum(loss_fn(-clf(x_dS) + eps))

    dVf = []
    L_SnG = 0
    for i in range(x_SnG.size()[0]):
        dV = torch.autograd.grad(clf(x_SnG[i]),x_SnG)[0][i]
        f = torch.tensor(f_SPM(x_SnG[i].detach().numpy(),u(x_SnG[i]).detach().numpy(),p)[0],dtype=torch.float32)
        dVf.append(torch.matmul(dV,f).detach().numpy())
        L_SnG += loss_fn(torch.matmul(dV, f) + eps)

    loss_u = L_I + L_dS + L_SnG

    optimizerU.zero_grad()
    loss_u.backward()
    optimizerU.step()

    loss_k.append(loss_u.detach())


plt.plot(loss_k)
plt.ylabel('Loss')
plt.xlabel('Iteration')
plt.grid()
plt.show()

plt.plot(dVf_max_k)
plt.ylabel('max(dV/dx * f(x,u))')
plt.xlabel('Iteration')
plt.grid()
plt.show()

print("x_I: ", np.max(clf(x_I).detach().numpy()))
print("x_dS: ", np.min(clf(x_dS).detach().numpy()))
print("x_S\G: ", dVf_max_k[-1])


# def dVf_u(x_SnG,i,u):
#     dV = torch.autograd.grad(clf(x_SnG[i]),x_SnG)[0][i].detach().numpy()
#     f = torch.tensor(f_SPM(x_SnG[i].detach().numpy(),u,p)[0],dtype=torch.float32)
#     f = f.detach().numpy()
#     return np.matmul(dV,f)

# for i in range(x_SnG.size()[0]):
#     dVf_u_partial = partial(dVf_u,x_SnG,i)
#     U = np.linspace(-5,0,100)
#     dVfs = []
#     for u in U:
#         dVfs.append(dVf_u_partial(u))
#     plt.plot(U,dVfs)
#     plt.title(str(i))
#     plt.xlabel('Input Current, I (A)')
#     plt.ylabel('dV*f(x,u)')
#     plt.grid()
#     plt.show()