simple 3 variable regression gives a very unstable model as correlation coefficient between the two independent variables(charging time and temperature) has very high absolute value

If you use a model Y = B0 + B1X1 + B2X2, the VIF for your terms is small. So, you don't need to worry about the correlation. 

Using the model:

Battery discharging = 11.90 + 0.3291 Battery charging - 0.3188 Out door temperature

From this model, I got: 

    Fit                SE Fit             95% CI                         95% PI

5.88291      0.360004 (5.16840, 6.59741) (3.73503, 8.03078)

So, I would expect a computer with 4hrs of charge and an external temp of 23 degrees is 5.88hrs. 

 Thank you for your answer

But consider the coefficient of determination ,it's value is 0.86 for your model for the entire dataset.

please look at the excel sheet attached

i have sorted the dataset on the values of outside temperature,the coefficient f determination comes out to be 0.434 for the first 47 records.

this indicates that the relationship between charging time and discharging time is not likely to be linear at low temperatures.

that implies that this model works poorly for low temperatures.