import numpy as np import pandas as pd import torch from torch import nn import os import time import matplotlib.pyplot as plt t1 = time.time() os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE" DAYS_FOR_TRAIN = 10 torch.manual_seed(42) class LSTM_Regression(nn.Module): def __init__(self, input_size, hidden_size, output_size=1, num_layers=2): super().__init__() self.lstm = nn.LSTM(input_size, hidden_size, num_layers) self.fc = nn.Linear(hidden_size, output_size) def forward(self, _x): x, _ = self.lstm(_x) # _x is input, size (seq_len, batch, input_size) s, b, h = x.shape # x is output, size (seq_len, batch, hidden_size) x = x.view(s * b, h) x = self.fc(x) x = x.view(s, b, -1) # 把形状改回来 return x def create_dataset(data, days_for_train=5) -> (np.array, np.array): dataset_x, dataset_y = [], [] for i in range(len(data) - days_for_train-3): dataset_x.append(data[i:(i + days_for_train)]) dataset_y.append(data[i + days_for_train:i + days_for_train+3]) # print(dataset_x,dataset_y) return (np.array(dataset_x), np.array(dataset_y)) def normal(nd): high = nd.describe()['75%'] + 1.5*(nd.describe()['75%']-nd.describe()['25%']) low = nd.describe()['25%'] - 1.5*(nd.describe()['75%']-nd.describe()['25%']) return nd[(ndlow)] def data_preprocessing(data): data.columns = data.columns.map(lambda x: x.strip()) data.index = data.index.map(lambda x:str(x).strip()[:10]) data.index = pd.to_datetime(data.index,format='%Y-%m-%d') data.sort_index(inplace=True) data = data.loc['2021-01':'2023-08'] data.drop(columns=[i for i in data.columns if (data[i] == 0).sum() / len(data) >= 0.5], inplace=True) # 去除0值列 data = data.astype(float) for col in data.columns: data[col] = normal(data[col]) return data # 拼接数据集 # file_dir = './浙江各地市行业电量数据' # excel = os.listdir(file_dir)[0] # data = pd.read_excel(os.path.join(file_dir, excel), sheet_name=0, index_col='stat_date') # data.drop(columns='地市',inplace=True) # data = data_preprocessing(data) # # df = data[data.columns[0]] # df.dropna(inplace = True) # dataset_x, dataset_y = create_dataset(df, DAYS_FOR_TRAIN) # # for level in data.columns[1:]: # df = data[level] # df.dropna(inplace=True) # x, y = create_dataset(df, DAYS_FOR_TRAIN) # dataset_x = np.concatenate((dataset_x, x)) # dataset_y = np.concatenate((dataset_y, y)) # # # for excel in os.listdir(file_dir)[1:]: # # data = pd.read_excel(os.path.join(file_dir,excel), sheet_name=0,index_col='stat_date') # data.drop(columns='地市', inplace=True) # data = data_preprocessing(data) # # for level in data.columns: # df = data[level] # df.dropna(inplace=True) # x,y = create_dataset(df,DAYS_FOR_TRAIN) # dataset_x = np.concatenate((dataset_x,x)) # dataset_y = np.concatenate((dataset_y,y)) # # # df_x_10 = pd.DataFrame(dataset_x) # df_y_10 = pd.DataFrame(dataset_y) # df_x_10.to_csv('df_x_10.csv',index=False) # df_y_10.to_csv('df_y_10.csv',index=False) dataset_x = pd.read_csv('df_x_10.csv').values dataset_y = pd.read_csv('df_y_10.csv').values print(dataset_x.shape,dataset_y.shape) # # 训练 device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') # 标准化到0~1 max_value = np.max(dataset_x) min_value = np.min(dataset_x) dataset_x = (dataset_x - min_value) / (max_value - min_value) dataset_y = (dataset_y - min_value) / (max_value - min_value) print('max_value:',max_value,'min_value:',min_value) # 划分训练集和测试集 train_size = int(len(dataset_x)*0.7) train_x = dataset_x[:train_size] train_y = dataset_y[:train_size] # # 将数据改变形状，RNN 读入的数据维度是 (seq_size, batch_size, feature_size) train_x = train_x.reshape(-1, 1, DAYS_FOR_TRAIN) train_y = train_y.reshape(-1, 1, 3) # # 转为pytorch的tensor对象 train_x = torch.from_numpy(train_x).to(device).type(torch.float32) train_y = torch.from_numpy(train_y).to(device).type(torch.float32) model = LSTM_Regression(DAYS_FOR_TRAIN, 32, output_size=3, num_layers=2).to(device) # 导入模型并设置模型的参数输入输出层、隐藏层等 train_loss = [] loss_function = nn.MSELoss() optimizer = torch.optim.Adam(model.parameters(), lr=0.005, betas=(0.9, 0.999), eps=1e-08, weight_decay=0) min_loss = 1 # for i in range(500): # train_x,train_y = train_x.to(device),train_y.to(device) # out = model(train_x) # loss = loss_function(out, train_y) # loss.backward() # optimizer.step() # optimizer.zero_grad() # train_loss.append(loss.item()) # # if loss <= min_loss: # min_loss = loss # best_para = model.state_dict() # if i % 100 == 0: # print(f'epoch {i+1}: loss:{loss}') # # 保存/读取模型 # torch.save(best_para,'hy3.pth') model = LSTM_Regression(DAYS_FOR_TRAIN, 32, output_size=3, num_layers=2).to(device) model.load_state_dict(torch.load('hy3.pth',map_location=torch.device('cpu'))) # 测试 model = model.eval() # dataset_x = dataset_x.reshape(-1, 1, DAYS_FOR_TRAIN) # (seq_size, batch_size, feature_size) # dataset_x = torch.from_numpy(dataset_x).to(device).type(torch.float32) # # pred_test = model(dataset_x) # 全量训练集 # pred_test = pred_test.view(-1) # pred_test = np.concatenate((np.zeros(DAYS_FOR_TRAIN), pred_test.cpu().detach().numpy())) # # plt.plot(pred_test.reshape(-1), 'r', label='prediction') # plt.plot(dataset_y.reshape(-1), 'b', label='real') # plt.plot((train_size*3, train_size*3), (0, 1), 'g--') # plt.legend(loc='best') # plt.show() # model.load_state_dict(torch.load('hy3.pth',map_location=torch.device('cpu'))) # max_value = 354024930.8 # min_value = 0.0 # 测试 # file_dir = './浙江各地市行业电量数据' df = pd.read_excel(r'C:\Users\鸽子\Desktop\浙江电量20231129.xlsx',sheet_name=2) for city in df['city_name'].drop_duplicates(): df_city = df[df['city_name']==city].sort_values(by='stat_date').set_index('stat_date') # df_city.index = df_city.index.map(lambda x:str(x).strip()[:10]) # df_city.index = pd.to_datetime(df_city.index) # df_city = df_city.loc['2023-9'][:-3] result_dict = {} for industry in df_city.columns[1:]: df_city[industry] = df_city[industry].astype('float') x, y = create_dataset(df_city[industry], 10) x = (x - min_value) / (max_value - min_value) x = x.reshape(-1, 1, 10) x = torch.from_numpy(x).type(torch.float32).to(device) pred = model(x).view(-1) pred = pred * (max_value - min_value) + min_value result = pred.cpu().detach().numpy()[-3:] result_dict[industry] = list(result) df1 = pd.DataFrame(result_dict,index=['2023-11-28','2023-11-29','2023-11-30']) df1['city_name'] = city df1 = df1[df_city.columns] df1 = pd.concat((df_city.iloc[:27], df1)) print(df_city) print(df1) with pd.ExcelWriter(r'C:\Users\鸽子\Desktop\行业电量预测v1129.xlsx',mode='a',engine='openpyxl',if_sheet_exists='replace') as writer: df1.to_excel(writer,sheet_name=f'{city[4:6]}') print(time.time()-t1) print(result_dict)