scaler = StandardScaler()，对三维数组(n_samples, n_time_steps, n_features)转换为2维数组进行标准化，还原为3维数组

可以通过以下方式将三维数组标准化为二维数组：

from sklearn.preprocessing import StandardScaler

scaler = StandardScaler()

# assume X is your 3D array with shape (n_samples, n_time_steps, n_features)
n_samples, n_time_steps, n_features = X.shape

# reshape X to 2D array
X_2d = X.reshape(n_samples * n_time_steps, n_features)

# standardize X_2d
X_2d_scaled = scaler.fit_transform(X_2d)

然后，可以通过以下方式将二维数组还原为三维数组：

# inverse transform to get the original 2D array
X_2d_unscaled = scaler.inverse_transform(X_2d_scaled)

# reshape to the original 3D array
X_unscaled = X_2d_unscaled.reshape(n_samples, n_time_steps, n_features)

注意，在还原过程中，一定要使用和标准化时一样的`scaler`进行逆变换。