Previous studies have revealed the significance of cortical hierarchy in simulation of functional connectome (FC) from structural connectome (SC) based on reduced Wong-Wang-Deco biophysical modelling approach. However, discrepancy has been noted as to the macroscopic gradient of fitted model parameters such as local recurrent strength, which warrants future research with advanced model training tech and functional constraints. Methods SC and FC from Human Connectome Project (HCP) dataset. Effective connectome (EC) from functional tractography (F-TRACT) dataset, summarized by single-pulse electrical stimulation and cortico-cortical evoked potentials (CCEPs) recording. Implemented by BrainPy platform, reduced Wong-Wang-Deco model was trained to fit FC and EC from SC, with its parameters optimized in a full degree of freedom fashion using back-propagation through time (BPTT). Results The simulated FC and EC were significantly correlated with empirical FC (r=0.72) and EC (r=0.73), respectively. The optimized model parameters were found within global balanced amplification (GBA) regime, i.e., strong inter-nodes excitation stabilized by intra-node recurrent inhibition. In addition, local recurrent strength and intrinsic noise level were found both positively correlated with the hierarchy (p<0.05), a pattern necessary to harmonize the divergence between the patterns of FC and EC node strength. Conclusion Biophysical modelling with GBA regime and specified cortical heterogeneity facilitates the simulation of both functional and effective brain connectomes.