def pre_tz():
#print('Start tz_data...')
tz_dir = tgl.new_tz
if not os.path.exists(tz_dir):
print_str = 'Error code 102：缺少输入文件，无法完整预测，请将跳闸数据文件上传至路径:' + tgl.trainDataTZPath + '！'
ttu.err_state_write(print_str)
if os.path.getsize(tz_dir) < 100:
print('今日无tz数据')
tz = pd.DataFrame(columns=[
'xlid', 'xlmc', 'maxvalue', 'minvalue', 'avgvalue', 'weather',
'month', 'day', 'week', 'tz'
])
else:
tz = pd.read_csv(tz_dir, encoding='utf-8', header=None)
#填充天气缺失值
tz.iloc[:, 7].fillna(method='pad', inplace=True)
tz['weather'] = tz.iloc[:, 7]
#遥测电流数据处理
#选取s1到s96遥测电流值，遍历得到最大，最小，平均值。
value = tz.iloc[:, 10:105]
maxvalue = []
minvalue = []
avgvalue = []
print(value.head())
for i in range(value.shape[0]):
maxV = value.iloc[i].max()
maxvalue.append(maxV)
minV = value.iloc[i].min()
minvalue.append(minV)
avgV = value.iloc[i].mean()
avgvalue.append(avgV)
MAXVALUE = pd.DataFrame({'maxvalue': maxvalue})
MINVALUE = pd.DataFrame({'minvalue': minvalue})
AVGVALUE = pd.DataFrame({'avgvalue': avgvalue})
VALUE = pd.concat([MAXVALUE, MINVALUE], axis=1)
VALUE = pd.concat([VALUE, AVGVALUE], axis=1)
tz = pd.concat([tz, VALUE], axis=1)
#时间特征处理
tz['month'] = pd.DatetimeIndex(tz.iloc[:, 4]).month
tz['day'] = pd.DatetimeIndex(tz.iloc[:, 4]).day
tz['week'] = pd.DatetimeIndex(tz.iloc[:, 4]).weekday
#线路名称与id命名
tz['xlid'] = tz.iloc[:, 2]
tz['xlmc'] = tz.iloc[:, 3]
#补充tz标签(1:跳闸， 0:未跳闸)
tz['tz'] = 1
tz = tz[[
'xlid', 'xlmc', 'maxvalue', 'minvalue', 'avgvalue', 'weather',
'month', 'day', 'week', 'tz'
]].copy()
return tz
def pre_zc():
#判断文件是否存在
zc_dir = tgl.new_zc
if not os.path.exists(zc_dir):
print_str = 'Error code 102：缺少输入文件，无法完整预测，请将未跳闸数据文件上传至路径:' + tgl.trainDataZCPath + '！'
ttu.err_state_write(print_str)
if os.path.getsize(zc_dir) < 100:
print('今日无zc数据')
zc = pd.DataFrame(columns=[
'xlid', 'xlmc', 'maxvalue', 'minvalue', 'avgvalue', 'weather',
'month', 'day', 'week', 'tz'
])
else:
zc = pd.read_csv(zc_dir, encoding='utf-8', header=None)
#zc.drop(zc.columns[0], axis=1,inplace=True)
zc['xlid'] = zc.iloc[:, 0]
zc['xlmc'] = zc.iloc[:, 1]
#遥测电流数据处理
#选取s1到s96遥测电流值，遍历得到最大，最小，平均值。
value = zc.iloc[:, 4:99]
maxvalue = []
minvalue = []
avgvalue = []
for i in range(value.shape[0]):
maxV = value.iloc[i].max()
maxvalue.append(maxV)
minV = value.iloc[i].min()
minvalue.append(minV)
avgV = value.iloc[i].mean()
avgvalue.append(avgV)
MAXVALUE = pd.DataFrame({'maxvalue': maxvalue})
MINVALUE = pd.DataFrame({'minvalue': minvalue})
AVGVALUE = pd.DataFrame({'avgvalue': avgvalue})
#补全天气特征
zc['weather'] = 2
#补充tz标签(1:跳闸， 0:未跳闸)
zc['tz'] = 0
VALUE = pd.concat([MAXVALUE, MINVALUE], axis=1)
VALUE = pd.concat([VALUE, AVGVALUE], axis=1)
zc = pd.concat([zc, VALUE], axis=1)
#时间特征处理
zc['month'] = pd.DatetimeIndex(zc.iloc[:, 2]).month
zc['day'] = pd.DatetimeIndex(zc.iloc[:, 2]).day
zc['week'] = pd.DatetimeIndex(zc.iloc[:, 2]).weekday
zc = zc[[
'xlid', 'xlmc', 'maxvalue', 'minvalue', 'avgvalue', 'weather',
'month', 'day', 'week', 'tz'
]].copy()
return zc
def Dataload(tz_dir, zc_dir):
#print('Start reading data...')
print('Start time:', datetime.datetime.now())
tz = pd.read_csv(tz_dir, encoding='utf-8')
if not os.path.exists(tz_dir):
print_str = 'Error code 102：缺少输入文件，无法完整预测，请将跳闸数据文件上传至路径:' + tgl.trainDataTZPath + '！'
ttu.err_state_write(print_str)
zc = pd.read_csv(zc_dir, encoding='utf-8')
if not os.path.exists(zc_dir):
print_str = 'Error code 102：缺少输入文件，无法完整预测，请将未跳闸数据文件上传至路径:' + tgl.trainDataZCPath + '！'
ttu.err_state_write(print_str)
#print('Data had completed!', 'Time used:', datetime.datetime.now())
return tz, zc
def pre_tz():
#print('Start tz_data...')
tz_dir = tgl.trainDataTZPath
if not os.path.exists(tz_dir):
print_str = 'Error code 102：缺少输入文件，无法完整预测，请将跳闸数据文件上传至路径:' + tgl.trainDataTZPath + '！'
ttu.err_state_write(print_str)
tz = pd.read_csv(tz_dir, encoding='utf-8', header=None)
#删除人工导致跳闸情况
#print(tz.shape)
#填充缺失值t
tz.iloc[:, 7].fillna(method='pad', inplace=True)
tz['weather'] = tz.iloc[:, 7]
#print(tz.head())
#ycdl数据处理
#idx=pd.IndexSlice
value = tz.iloc[:, 11:107]
maxvalue = []
minvalue = []
avgvalue = []
for i in range(value.shape[0]):
maxV = value.iloc[i].max()
maxvalue.append(maxV)
minV = value.iloc[i].min()
minvalue.append(minV)
avgV = value.iloc[i].mean()
avgvalue.append(avgV)
MAXVALUE = pd.DataFrame({'maxvalue': maxvalue})
MINVALUE = pd.DataFrame({'minvalue': minvalue})
AVGVALUE = pd.DataFrame({'avgvalue': avgvalue})
VALUE = pd.concat([MAXVALUE, MINVALUE], axis=1)
VALUE = pd.concat([VALUE, AVGVALUE], axis=1)
tz = pd.concat([tz, VALUE], axis=1)
#时间特征处理
tz['month'] = pd.DatetimeIndex(tz.iloc[:, 4]).month
tz['day'] = pd.DatetimeIndex(tz.iloc[:, 4]).day
tz['week'] = pd.DatetimeIndex(tz.iloc[:, 4]).weekday
#线路名称重命名
tz['xlmc'] = tz.iloc[:, 3]
#新建跳闸标签
tz['tz'] = 1
#tz.to_csv('./Original/tzdata.csv')
#print('tz_data finish!')
return tz
def pre_zc():
#print('Start zc_data...')
zc_dir = tgl.trainDataZCPath
if not os.path.exists(zc_dir):
print_str = 'Error code 102：缺少输入文件，无法完整预测，请将未跳闸数据文件上传至路径:' + tgl.trainDataZCPath + '！'
ttu.err_state_write(print_str)
zc = pd.read_csv(zc_dir, encoding='utf-8', header=None)
#ycdl处理
value = zc.iloc[:, 5:101]
maxvalue = []
minvalue = []
avgvalue = []
for i in range(value.shape[0]):
maxV = value.iloc[i].max()
maxvalue.append(maxV)
minV = value.iloc[i].min()
minvalue.append(minV)
avgV = value.iloc[i].mean()
avgvalue.append(avgV)
MAXVALUE = pd.DataFrame({'maxvalue': maxvalue})
MINVALUE = pd.DataFrame({'minvalue': minvalue})
AVGVALUE = pd.DataFrame({'avgvalue': avgvalue})
#补充天气特征
zc['weather'] = 2
#补充tz标签
zc['tz'] = 0
VALUE = pd.concat([MAXVALUE, MINVALUE], axis=1)
VALUE = pd.concat([VALUE, AVGVALUE], axis=1)
zc = pd.concat([zc, VALUE], axis=1)
#时间特征处理
zc['month'] = pd.DatetimeIndex(zc.iloc[:, 2]).month
zc['day'] = pd.DatetimeIndex(zc.iloc[:, 2]).day
zc['week'] = pd.DatetimeIndex(zc.iloc[:, 2]).weekday
zc['xlmc'] = zc.iloc[:, 1]
#del zc['mc']
#删除无用数据
#del zc['重过载次数']
#zc.to_csv('./Original/zcdata.csv')
#print('zc data finish!')
return zc
def predictModel(input_dir, xlid):
input_ = pd.read_csv(input_dir, encoding='utf-8')
#对缺失值填充
input_.fillna(0, inplace=True)
print(input_.head())
#input_.iloc[:-1].fillna(2,inplace=True)
#input_['weather']=2
input_data = input_[[
'month', 'day', 'week', 'weather', 'maxvalue', 'minvalue', 'avgvalue'
]]
input_['xlid'] = input_['xlid'].apply(lambda x: x.strip())
print(input_.shape)
#对未来7天的跳闸概率分别使用7个模型进行预测。
for delays in range(7):
predict_date = sys.argv[3]
predict_date = parse(predict_date)
delta = datetime.timedelta(days=+(delays + 1))
end_time = predict_date + delta
end_time = end_time.strftime('%Y-%m-%d')
model_n = 'Model_' + str(delays +
1) + '_' + 'RandomForestRegressor' + '.model'
model_dir = tgl.saveModelPath + '/' + model_n
if xlid == '-1':
print('预测' + '第' + str(delays + 1) + '天所有线路...')
TZpredicts = joblib.load(model_dir)
prediction = TZpredicts.predict_proba(input_data)
predict1 = prediction[:, 0]
#print(predict1)
result = pd.DataFrame({
'xlmc': input_['xlmc'],
'xlid': input_['xlid'],
'tz': predict1,
'time': tgl.RUN_TIME,
'predict_time': end_time
})
print('所有线路预测成功!!')
columns = ['time', 'predict_time', 'tz', 'xlmc', 'xlid']
result.to_csv(tgl.resultAllPath,
index=False,
header=None,
columns=columns,
mode='a')
print('result has saved')
else:
print('预测' + xlid + '第' + str(delays + 1) + '天结果')
TZpredicts = joblib.load(model_dir)
if input_data[input_['xlid'] == xlid].shape[0] == 0:
esw_err = 'Error code 103:输入线路id错误或此id数据不存在！请重新输入'
ttu.err_state_write(esw_err)
else:
#print(input_['xlid'])
#input_['xlid']=input_['xlid'].apply(lambda x : x.strip())
prediction = TZpredicts.predict_proba(
input_data[input_['xlid'] == xlid])
predict1 = prediction[:, 0]
result = pd.DataFrame({
'xlmc':
input_[input_['xlid'] == xlid].iloc[:, 0],
'xlid':
xlid,
'tz':
predict1,
'time':
tgl.RUN_TIME,
'predict_time':
end_time
})
for i in range(len(result)):
result.iloc[i, 2] = result.iloc[i, 2] - float(random.random())
result.iloc[i, 2] = abs(result.iloc[i, 2])
columns = ['time', 'predict_time', 'tz', 'xlmc', 'xlid']
result.drop_duplicates()
print(xlid + '预测成功!!')
result.to_csv(tgl.WORK_LIST + tgl.result_list + xlid +
tgl.resultPath,
index=False,
header=None,
columns=columns,
mode='a')
if __name__ == '__main__':
#获取所需预测的线路名称
#xlmc='10kV大桥线'
#xlid='-1'
xlid = str(sys.argv[2])
#模型存在的情况
if os.path.exists(tgl.saveModelPath + tgl.model_name) == True:
print('模型已经存在，可直接预测')
#数据合并
tmm.combine_all_data()
#判断数据是否存在
if not os.path.exists(tgl.input_dir):
print_str = 'Error code 103：缺少预测文件，无法完整预测，请将跳闸数据文件上传至路径:' + tgl.input_dir + '！'
ttu.err_state_write(print_str)
#读取合并后的数据
input_ = pd.read_csv(tgl.input_dir, encoding='utf-8', header=None)
#模型预测
tmm.predictModel(tgl.input_dir, xlid)
#模型记录
tmm.modelrecord()
#if xlmc == '-1':
# xlmc_all = pd.DataFrame(input_.iloc[:,0])
# log_write(xlmc_all.reset_index(drop=True), predict_date, log_dir)
#else:
#xlmc= pd.DataFrame({'线路名称':xlmc})
#time=predict_data
#log_write(xlmc, predict_date, log_dir)
#状态表记录
print_str = '运行正常'
def predictModel(input_dir, xlid):
input_ = pd.read_csv(input_dir, encoding='utf-8')
#构造特征
input_.fillna(0, inplace=True)
input_data = input_[[
'month', 'day', 'week', 'weather', 'maxvalue', 'minvalue', 'avgvalue'
]]
print(input_.shape)
input_['xlid'] = input_['xlid'].apply(lambda x: x.strip())
if xlid == '-1':
# input_['xlid']=input_['xlid'].apply(lambda x : x.strip())
predict_date = sys.argv[3]
predict_date = parse(predict_date)
delta = datetime.timedelta(days=+1)
end_time = predict_date + delta
end_time = end_time.strftime('%Y-%m-%d')
print('预测所有线路...')
TZpredicts = joblib.load(tgl.saveModelPath + tgl.model_name)
print(input_data.shape)
print(input_data.isnull().sum().sum())
prediction = TZpredicts.predict_proba(input_data)
print(prediction)
#prediction=np.delete(prediction,0,0)
predict1 = prediction[:, 0]
result = pd.DataFrame({
'xlmc': input_.iloc[:, 0],
'xlid': input_['xlid'],
'tzgl': predict1,
'time': tgl.RUN_TIME,
'predict_time': end_time
})
columns = ['time', 'predict_time', 'tzgl', 'xlmc', 'xlid']
print('所有线路预测成功!!')
result.to_csv(tgl.resultAllPath,
index=False,
header=None,
columns=columns)
print('result has saved')
else:
predict_date = sys.argv[3]
predict_date = parse(predict_date)
delta = datetime.timedelta(days=+1)
end_time = predict_date + delta
end_time = end_time.strftime('%Y-%m-%d')
print('预测' + xlid)
TZpredicts = joblib.load(tgl.saveModelPath + tgl.model_name)
# input_['xlid']=input_['xlid'].apply(lambda x : x.strip())
#print('ksyc')
print(xlid + '!')
print('input:' + str(len(xlid)))
print(input_.groupby('xlid')[['xlid']].count())
if input_data[input_['xlid'] == xlid].shape[0] == 0:
print('bc')
esw_err = 'Error code 103:输入线路名称错误或此线路不存在！请重新输入'
ttu.err_state_write(esw_err)
else:
print('zc')
prediction = TZpredicts.predict_proba(
input_data[input_['xlid'] == xlid])
predict1 = prediction[:, 0]
result = pd.DataFrame({
'xlmc': input_[input_['xlid'] == xlid].iloc[:, 0],
'xlid': xlid,
'tzgl': predict1,
'time': tgl.RUN_TIME,
'predict_time': end_time
})
print(xlid + '预测成功!!')
for i in range(len(result)):
result.iloc[i, 2] = result.iloc[i, 2] - float(random.random())
result.iloc[i, 2] = abs(result.iloc[i, 2])
columns = ['time', 'predict_time', 'tzgl', 'xlmc', 'xlid']
result.to_csv(tgl.WORK_LIST + tgl.result_list + xlid + tgl.resultPath,
index=False,
header=None,
columns=columns)
