In the form you have phrased the question, it must be assumed that a random location inside the data might be updated (as opposed to the case where more data is added at the end.)
As such, we need to understand the excel data as being a data matrix, as opposed to it being timeseries data. Timeseries data would involve processing rows of input data at the time associated with the row, with it not being possible to go back and re-do steps if the data changes for an already-processed time.
You cannot use From Spreadsheet for this purpose. From Spreadsheet is strictly for reading matrix data in timeseries form (that is, there is a time column that indicates when each row is to come into effect.)
I am not sure that there are methods to periodically re-read excel data matrices as pure blocks of data; I cannot think of any at the moment, but I might be overlooking something.
The situation would be different if you are willing to run the simulation under the control of MATLAB, and if all of the simulation is to be done each time. In that situation, you would have the MATLAB level re-read from the excel file, and have the MATLAB level perform set_param() calls to set up block parameters, and then call sim() to run all of the simulation, and then loop back.