A Smart CSP (S-CSP) method is developed and integrated with our previous correlated dynamic adaptive chemistry and transport (CO-DACT) method to further improve the computational efficiency of the chemical integration. The S-CSP method resolved a few challenges in the original simple CSP method including the failure of matrix inversion of the fast reaction subspace and the inefficiency in computation of reaction timescales. A pre-generated library is proposed to eliminate the linearly dependent reactions in the matrix associated with the fast reaction subspace. The characteristic time of elementary reactions are obtained analytically. The reaction rates of the selected fast reactions are represented by the combinations of slow reactions. Therefore, the proposed Smart CSP method can automatically remove the stiffness of the ODE system on the fly without human experience. The correlated reduced mechanisms and correlated transport coefficients are also updated dynamically by using the CO-DACT method. The homogenous auto-ignition in the H2/air mixture is numerically modeled as the validation of the present S-CSP method. The results not only demonstrate the accuracy of the proposed S-CSP method but also show its potential to accelerate the chemical integration with large detailed chemical kinetics.