def TestOut(self, batchtest=False):
if not batchtest:
print('Network Out Test Real-time')
Batch = 1
else:
print('Network Out Test Batch')
Batch = 3
PhyTemp = tensor([[list(random(size=self.NubTimeSeq)) for _ in range(self.NubPhyPara)] for _ in range(Batch)],
dtype=torch.float)
TOOL.ALLP(PhyTemp)
CompTemp = tensor([[list(random(size=self.NubTimeSeq)) for _ in range(self.NubComPara)] for _ in range(Batch)],
dtype=torch.float)
TOOL.ALLP(CompTemp)
mu_v = self.GetPredictActorOut(x_py=PhyTemp, x_comp=CompTemp) # [3, 1]
TOOL.ALLP(mu_v)
mu = mu_v.data.numpy()
logstd = self.logstd.detach().numpy()
TOOL.ALLP(mu, 'mu')
TOOL.ALLP(logstd, "logstd")
action = mu + np.exp(logstd) * np.random.normal(size=logstd.shape)
TOOL.ALLP(action, "act")
action = mu + np.exp(logstd) * np.random.normal(size=logstd.shape)
TOOL.ALLP(action, "act")
action = np.clip(action, -1, 1)
TOOL.ALLP(action, "action clip")
def TestOut(self, batchtest=False):
if not batchtest:
print('Network Out Test Real-time')
Batch = 1
else:
print('Network Out Test Batch')
Batch = 3
PhyTemp = tensor([[list(random(size=self.NubTimeSeq)) for _ in range(self.NubPhyPara)] for _ in range(Batch)],
dtype=torch.float)
TOOL.ALLP(PhyTemp)
CompTemp = tensor([[list(random(size=self.NubTimeSeq)) for _ in range(self.NubComPara)] for _ in range(Batch)],
dtype=torch.float)
TOOL.ALLP(CompTemp)
self.GetPredictActorOut(x_py=PhyTemp, x_comp=CompTemp)
def TestOut(self, batchtest=False):
if not batchtest:
print('Network Out Test Real-time')
Batch = 1
else:
print('Network Out Test Batch')
Batch = 3
PhyTemp = tensor([[list(random(size=self.NubTimeSeq)) for _ in range(self.NubPhyPara)] for _ in range(Batch)],
dtype=torch.float)
TOOL.ALLP(PhyTemp)
CompTemp = tensor([[list(random(size=self.NubTimeSeq)) for _ in range(self.NubComPara)] for _ in range(Batch)],
dtype=torch.float)
TOOL.ALLP(CompTemp)
val = self.GetPredictCrticOut(x_py=PhyTemp, x_comp=CompTemp) # [3, 1]
TOOL.ALLP(val, "val")
val = np.clip(val.data.numpy(), -1, 1)
TOOL.ALLP(val, "val")
def get_reward(self, AMod):
"""
R => _
:return:
"""
V = {
'CNSTime': self.mem['KCNTOMS']['Val'], # CNS Time : KCNTOMS
'MalTime': self.mem['cMALT']['Val'], # Mal time
'PVCT': self.mem['PVCT']['Val'], # VCT_pressure : PVCT
'VCT_level': self.mem['ZVCT']['Val'], # VCT_level : ZVCT
'BLV616': self.mem['BLV616']['Val'], # LV616_pos : BLV616
'KLAMPO71': self.mem['KLAMPO71']['Val'], # CHP1 : KLAMPO71
'KLAMPO72': self.mem['KLAMPO72']['Val'], # CHP2 : KLAMPO72
'KLAMPO73': self.mem['KLAMPO73']['Val'], # CHP2 : KLAMPO73
'WCMINI': self.mem['WCMINI']['Val'], # Mini Flow : WCMINI
'BHV30': self.mem['BHV30']['Val'], # HV30_pos : BHV30
'BHV50': self.mem['BHV50']['Val'], # HV50_pos : BHV50
'BFV122': self.mem['BFV122']['Val'], # FV122_pos : BFV122
'WAUXSP': self.mem['WAUXSP']['Val'], # AuxSpray_Flow : WAUXSP
'BHV40': self.mem['BHV40']['Val'], # HV40 : BHV40
'WNETCH': self.mem['WNETCH']['Val'], # Total_Charging_Flow: WNETCH
'URHXUT': self.mem['URHXUT']['Val'], # Letdown_temp : URHXUT
'UCHGUT': self.mem['UCHGUT']['Val'], # Charging_temp : UCHGUT
'ZINST58': self.mem['ZINST58']['Val'], # PZR_press : ZINST58
'PZR_level': self.mem['ZINST63']['Val'], # PZR_level : ZINST63
'BLV459': self.mem['BLV459']['Val'], # Letdown_pos : BLV459
'BHV1': self.mem['BHV1']['Val'], # Orifice_1 : BHV1
'BHV2': self.mem['BHV2']['Val'], # Orifice_2 : BHV2
'BHV3': self.mem['BHV3']['Val'], # Orifice_3 : BHV3
'BPV145': self.mem['BPV145']['Val'], # Letdown_HX_pos : BPV145
'ZINST36':
self.mem['ZINST36']['Val'], # Letdown HX Press : ZINST36
'BHV41': self.mem['BHV41']['Val'], # HV41_pos : BHV41
'KHV43': self.mem['KHV43']['Val'], # HV43_pos : KHV43
'WEXLD': self.mem['WEXLD']['Val'], # VCT_flow : WEXLD
'WDEMI': self.mem['WDEMI']['Val'], # Total_in_VCT : WDEMI
}
PZR_level_set = 57
VCT_level_set = 74
r = [0, 0]
r[0] = TOOL.generate_r(curr=V['PZR_level'],
setpoint=PZR_level_set,
distance=0.5,
max_r=0.5,
min_r=-5)
r[1] = V['CNSTime'] / 50 # [1~500] 값 생산
r[1] = r[1] / 250 #[1/250 ~ 2]
self.Loger_txt += f'R:,{r},\t'
r = sum(r) / 100
# r = self.normalize(sum(r), 0, -5, 0.5) / 10
self.AcumulatedReward += r
return r