Modeling finite-rate chemistry in turbulent reacting flows is challenging because of the large span in length and time scales. Reynolds-averaged Navier-Stokes equations-based simulations do not resolve turbulent fluctuations and hence neglect their effect on the reaction rates. Turbulence-chemistry interaction was accounted for in RANS simulations via quadrature-based integration of the reaction rates, calculated using a temperature probability density function with a presumed Gaussian shape and transported mean and variance. The effect on light olefin yield was 0.1-0.2 wt % absolute. A dynamic zoning method was implemented to reduce the computational cost by performing chemical rate calculations only once for thermodynamically similar cells. Speedups of 50-190 were observed while the relative error on conversion remained below 0.05%. The advantages of the presented methodology were illustrated for a large-scale butane-cracking U-coil reactor.