I woud guess that Rsh_array5pc is, at some point during the simulation, dropping to a value very near zero, such that the value of IRsh_5% approaches infinity. If you have an infinity term going into the sum, you'll have infinity leaving the sum, the input to your transfer function will be infinite, and you'll get that error.
The relevant concept here is that continuous state (such as the state in continuous transfer functions) is an integral, so the inputs to state can be viewed as derivatives. Thus, "Derivative of state '1'" can be read as "Input to internal state '1'", which is presumably also the input to the block. So you can read it as "Input to block <linked> is not finite" and then trace back.
You can probably solve the problem by putting a saturation block on the output of the divider, such that IRsh_5% is constrained to be between some appropriately large values, or by updating Rsh_array5pc so it never goes to 0. If Rsh_array5pc always carries the same sign you can do this with a Saturation block. Otherwise you can pull the sign out, saturate the magnitude, and put the sign back manually.
