You can just do calculations with floats as much as possible, e.g., 15% = 0.15, binary search with epsilon = 0.0001.

Rounding with round(x, 2) are needed only when:
 1. K*L
 2. output of T(K)
-- "The calculation of the regional coefficient and all kinds the taxes is made with rounding off to two digits after a decimal point."

while K itself is float.