### 基于深度学习的风电功率预测 Python 实现代码 以下是基于 LSTM 的多变量输入多步预测的 Python 示例代码，适用于风电功率预测场景。此代码结合了时间序列数据处理和深度学习框架 PyTorch 的实现方法。 #### 数据预处理部分 ```python import numpy as np import pandas as pd from sklearn.preprocessing import MinMaxScaler from torch.utils.data import Dataset, DataLoader class WindPowerDataset(Dataset): def __init__(self, data, seq_length=10): self.seq_length = seq_length self.data = data self.scaler = MinMaxScaler(feature_range=(-1, 1)) self.scaled_data = self.scaler.fit_transform(data) def __len__(self): return len(self.scaled_data) - self.seq_length def __getitem__(self, idx): x = self.scaled_data[idx:idx+self.seq_length, :-1] # 输入特征 y = self.scaled_data[idx+self.seq_length, -1] # 预测目标 (风电功率) return np.float32(x), np.float32(y) def prepare_dataset(file_path, seq_length=10, batch_size=32): df = pd.read_csv(file_path) dataset = WindPowerDataset(df.values, seq_length=seq_length) dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=False) return dataloader, dataset.scaler ``` #### 模型定义部分 ```python import torch import torch.nn as nn class LSTMPredictor(nn.Module): def __init__(self, input_dim, hidden_dim, num_layers, output_dim): super(LSTMPredictor, self).__init__() self.hidden_dim = hidden_dim self.num_layers = num_layers self.lstm = nn.LSTM(input_dim, hidden_dim, num_layers, batch_first=True) self.fc = nn.Linear(hidden_dim, output_dim) def forward(self, x): h0 = torch.zeros(self.num_layers, x.size(0), self.hidden_dim).requires_grad_() c0 = torch.zeros(self.num_layers, x.size(0), self.hidden_dim).requires_grad_() out, (hn, cn) = self.lstm(x, (h0.detach(), c0.detach())) out = self.fc(out[:, -1, :]) return out ``` #### 训练过程部分 ```python device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') def train_model(model, criterion, optimizer, dataloader, scaler, epochs=50): model.train() for epoch in range(epochs): total_loss = 0 for i, (inputs, targets) in enumerate(dataloader): inputs, targets = inputs.to(device), targets.to(device) outputs = model(inputs) loss = criterion(outputs.squeeze(), targets) optimizer.zero_grad() loss.backward() optimizer.step() total_loss += loss.item() avg_loss = total_loss / len(dataloader) print(f"Epoch [{epoch+1}/{epochs}], Loss: {avg_loss:.4f}") return model, scaler.inverse_transform(np.array([[outputs.cpu().detach().numpy()[i]] for i in range(len(outputs))])) ``` #### 测试与验证部分 ```python def test_model(model, dataloader, scaler): model.eval() predictions = [] actuals = [] with torch.no_grad(): for inputs, targets in dataloader: inputs, targets = inputs.to(device), targets.to(device) outputs = model(inputs) predictions.extend(scaler.inverse_transform(outputs.cpu().numpy()).tolist()) actuals.extend(scaler.inverse_transform(targets.view(-1, 1).cpu().numpy()).tolist()) return predictions, actuals ``` --- 上述代码实现了基于 LSTM 的风电功率预测模型[^4]。该模型支持多变量输入（如风速、温度、湿度等），并能够完成多步预测任务。为了提升预测性能，还可以考虑引入注意力机制或其他改进策略[^2]。 --- ###