import torch import pandas as pd import numpy as np import matplotlib.pyplot as plt from torch import nn import os from torch.utils.data import TensorDataset, DataLoader import datetime torch.manual_seed(42) os.environ[ "KMP_DUPLICATE_LIB_OK"] = "TRUE" # 解决OMP: Error #15: Initializing libiomp5md.dll, but found libiomp5md.dll already initialized. pd.set_option('display.width', None) class LSTM(nn.Module): def __init__(self, input_size, hidden_size, output_size, num_layers=3): super().__init__() self.lstm = nn.LSTM(input_size, hidden_size, num_layers) self.fc1 = nn.Linear(hidden_size, 64) self.fc2 = nn.Linear(64, 128) self.fc3 = nn.Linear(128, output_size) self.ReLu = nn.ReLU() self.dropout = nn.Dropout() def forward(self, x): output, _ = self.lstm(x) s, b, h = output.shape output = output.reshape(-1, h) output = self.ReLu(self.fc1(output)) output = self.ReLu(self.fc2(output)) output = self.fc3(output) return output def normal(df): drop_col = [x for x in df.columns if len(df[df[x]==0])/len(df) >= 0.5] df.drop(columns=drop_col,inplace=True) for col in df.columns: try: high = df[col].describe()['75%'] + 1.5 * (df[col].describe()['75%'] - df[col].describe()['25%']) low = df[col].describe()['25%'] - 1.5 * (df[col].describe()['75%'] - df[col].describe()['25%']) df[col] = df[col].map(lambda x: np.nan if (x >= high) | (x <= low) else x) df[col] = df[col].fillna(method='ffill') df[col] = df[col].fillna(method='bfill') except: pass return df def create_data(df_industry, industry): dataset_x = [] dataset_y = [] # 按月份分组 grouped = df_industry.groupby(df_industry['stat_date'].dt.to_period('M')) # 遍历每个月的数据 for name, group in grouped: if len(group) == 31: dataset_x.append(list(group[industry].values[1:28])) dataset_y.append(list(group[industry].values[-3:])) if len(group) == 30: dataset_x.append(list(group[industry].values[:27])) dataset_y.append(list(group[industry].values[-3:])) if len(group) == 28: fst = group[industry].values[0] dataset_x.append([fst, fst, fst] + list(group[industry].values[1:25])) dataset_y.append(list(group[industry].values[-3:])) else: fst = group[industry].values[0] if len([fst, fst] + list(group[industry].values[1:26])) != 27: break dataset_x.append([fst, fst] + list(group[industry].values[1:26])) dataset_y.append(list(group[industry].values[-3:])) return np.array(dataset_x), np.array(dataset_y) # 创建数据集 file_dir = './浙江各地市行业电量数据' city1 = os.listdir(file_dir)[0] df_city = pd.read_excel(os.path.join(file_dir, city1)) df_city = normal(df_city) df_city = df_city.drop(columns='地市') df_city[df_city.columns[1:]] /= 10000 df_city['stat_date'] = df_city['stat_date'].map(lambda x: str(x).strip()[:10]) df_city.stat_date = pd.to_datetime(df_city.stat_date) industry = '全社会用电总计' df_industry = df_city[['stat_date', industry]] dataset_x, dataset_y = create_data(df_industry, industry) for industry in df_city.columns[2:]: df_level = df_city[['stat_date', industry]] x, y = create_data(df_level, industry) dataset_x = np.concatenate([dataset_x, x]) dataset_y = np.concatenate([dataset_y, y]) for excel in os.listdir(file_dir)[1:]: df_city = pd.read_excel(os.path.join(file_dir, excel)).drop(columns='地市') df_city = normal(df_city) df_city[df_city.columns[1:]] /= 10000 df_city['stat_date'] = df_city['stat_date'].map(lambda x: str(x).strip()[:10]) df_city.stat_date = pd.to_datetime(df_city.stat_date) for industry in df_city.columns[1:]: df_level = df_city[['stat_date', industry]] x, y = create_data(df_level, industry) dataset_x = np.concatenate([dataset_x, x]) dataset_y = np.concatenate([dataset_y, y]) print(dataset_x.shape, dataset_y.shape) 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, min_value) # 划分训练集和测试集 train_size = int(len(dataset_x) * 0.8) train_x = dataset_x[:train_size] train_y = dataset_y[:train_size] eval_x = dataset_x[train_size:] eval_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, 27) train_y = train_y.reshape(-1, 1, 3) eval_x = eval_x.reshape(-1, 1, 27) eval_y = eval_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) eval_x = torch.from_numpy(eval_x).to(device).type(torch.float32) eval_y = torch.from_numpy(eval_y).to(device).type(torch.float32) train_ds = TensorDataset(train_x, train_y) train_dl = DataLoader(train_ds, batch_size=32, shuffle=True, drop_last=True) eval_ds = TensorDataset(eval_x, eval_y) eval_dl = DataLoader(eval_ds, batch_size=64, drop_last=True) model = LSTM(27, 16, output_size=3, num_layers=3).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) min_loss = 1 for i in range(10): model.train() for j, (x, y) in enumerate(train_dl): x, y = x.to(device), y.to(device) out = model(x) loss = loss_function(out, y) loss.backward() optimizer.step() optimizer.zero_grad() train_loss.append(loss.item()) # if (i+1) % 100 == 0: # print(f'epoch {i+1}/1500 loss:{round(loss.item(),5)}') if (j + 1) % 10 == 0: print(f'epoch {i + 1}/200 step {j + 1}/{len(train_dl)} loss:{loss}') test_running_loss = 0 model.eval() with torch.no_grad(): for x, y in eval_dl: pred = model(eval_x) loss = loss_function(pred, y) test_running_loss += loss.item() test_loss = test_running_loss / len(eval_dl) if test_loss < min_loss: min_loss = test_loss best_model_weight = model.state_dict() print(f'epoch {i + 1} test_loss:{test_loss}') total_params = sum(p.numel() for p in model.parameters() if p.requires_grad) print(f"Total parameters in the LSTM model: {total_params}") # 保存模型 torch.save(best_model_weight, 'dy3.pth') # 读取模型 model = LSTM(27, 16, output_size=3, num_layers=3).to(device) model.load_state_dict(torch.load('dy3.pth')) # for test dataset_x = dataset_x.reshape(-1, 1, 27) # (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 # 模型输出 (seq_size, batch_size, output_size) pred_test = pred_test.view(-1).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--') # 分割线 左边是训练数据 右边是测试数据的输出 plt.legend(loc='best') plt.show()