soc估计:DESIGN AND DEVELOPMENT OF SoC ESTIMATION MODEL USING MACHINE LEARNING

这是一篇印度那边学生的毕业论文,唯一要记录的是里面提到了一个特征构造的思想,记录如下:
论文思想:
特征选用速度、电流、电压、温度、平均电压、平均电流、平均速度,模型用cnn+lstm+lr+lr
平均特征计算方式:近50个时刻的取值求平均。
模型参数:

# 建立网络结构
input_size = 7
hidden_size = 48
output_size = 1
num_layers = 1
rate = 5e-3
epochs = 1500


class LSTM_CONV(nn.Module):
    def __init__(self, input_size = input_size, hidden_size = hidden_size, output_size = output_size,
                 num_layers = num_layers):
        super(LSTM_CONV, self).__init__()

        self.conv = nn.Conv1d(in_channels = 1, out_channels = 6, kernel_size = 3, stride = 1)
        self.rnn = nn.LSTM(5, hidden_size, num_layers)
        self.reg_1 = nn.Linear(hidden_size, output_size)
        self.reg_2 = nn.Linear(6, output_size)

    def forward(self, x):
        x = self.conv(x)  
        x, _ = self.rnn(x)  
        s, b, h = x.shape
        x = x.view(s * b, h)  
        x = self.reg_1(x)  
        x = x.view(s, -1) 
        x = self.reg_2(x) 
        x = x.view(s, 1, 1)  
        return x


net = LSTM_CONV()
criterion = nn.MSELoss()
optimizer = torch.optim.Adam(net.parameters(), lr = rate, betas = (0.9, 0.999), eps = 1e-08)

你可能感兴趣的:(电池文献阅读,soc估计)