CFD simulations of reacting flows are fundamental investigation tools used to predict combustion behaviour and pollutants formation in modern spark-ignition internal combustion engines. Most of the flamelet-based combustion models adopted in current simulations use the fuel/air/residual laminar flame speed as a background to predict the turbulent flame speed. This in turn is a fundamental requirement to model the effective burn rate. The consolidated approach in engine combustion simulations relies on the adoption of empirical correlations for laminar flame speed, which are derived from fitting activity of combustion experiments. However, these last are conducted at largely different pressure and temperature ranges from those encountered in engines: for this reason, correlation extrapolation at engine conditions is inevitably accepted and relevant differences between proposed correlations emerge even for the same fuel and conditions. The lack of predictive chemistry-grounded correlations leads to a wide modelling uncertainty, requiring extensive model tuning when validating combustion simulations against engine experiments. In this paper a fitting form based on fifth order logarithmic polynomials is applied to reconstruct correlations for a set of Primary Reference Fuels (PRFs), namely iso-octane, n-heptane, toluene and for a gasoline surrogate. Experimental data from literature are collected as well as existing computations for laminar flame speed. These last extend up to engine-relevant conditions where experiments are not available, and they constitute a model-based prediction of flame behaviour at such states. Such literature and calculated data set constitute the target values for the fitting polynomials, which are shown to be representative of a wide range of engine-typical operating points.