Is it possible to use a forecast as input variable in multivariable multistep LSTM algorithm

Question

I am using LSTM to forecast the energy demand of the next 24 hours of a household. I am using x1 = temperature and x2 = precipitation and y = energy consumption.

If I use 1 year of historical data (hour by hour) and I want to calculate the next 24 steps, how can I include the temperature prediction and precipitation prediction data for the next 24 steps?

In other words, how can include the values of the variables x1 and x2 for the time steps that I am looking to forecast for y?

time                  X1   X2   Y
t-999                 22   33   32
...                   ...  ...  ...
t-3                   23   44   21
t-2                   25   44   33
t-1                   22   55   42
t                     21   22   22
t+1                   24   22   ?
t+2                   22   13   ?
...                   ...  ...  ...
t+24                  24   32   ?

Thanks!