• Effect of expanded natural graphite (ENG) content gradient on reactor was studied. • An optimization method for ENG distribution based on entransy theory was proposed. • A larger ENG content gradient led to a more uniform temperature distribution. • The number of optimization iterations had little effect on optimal results. • A simplified configuration based on the optimized results was discussed. The poor thermal conductivity of MH beds limits hydrogen absorption in metal hydride (MH) reactors for thermochemical heat storage. The compaction of MH powder with expanded natural graphite (ENG) is an efficient and cost-effective approach for the heat transfer enhancement of MH beds. However, the addition of ENG reduces the hydrogen storage density of the compacted composites. How to optimize the allocation of a limited amount of ENG in the MH bed has not been studied. In this study, a multilayer MH bed configuration with an ENG grade content was adopted to improve the heat transfer performance of an MH reactor for thermochemical heat storage. The effect of the ENG content gradient on the performance of the MH reactor was investigated using a mathematical model of the MH reactor. Furthermore, an optimization method based on the entransy dissipation extremum principle (EDEP) was proposed to optimize the ENG distribution in magnesium hydride (MgH 2 )-ENG compacts. The results showed that a larger ENG content gradient resulted in faster hydrogen absorption and better heat-discharge performance. Compared with a configuration with a uniform ENG content, the optimized configuration based on the EDEP exhibited a shorter reaction time, more uniform bed temperature distribution, and 15.33% higher gravimetric exergy-output rate. In addition, the number of optimization iterations had little effect on the optimal results, which could be obtained by implementing only one optimization iteration. The optimization results showed that more ENG should be allocated to the internal layer of the MgH 2 -ENG compact. The compacts were divided into three layers, and the ENG content of each layer was calculated according to the optimized configuration based on the EDEP. The optimization method proposed in this study can be applied to optimize the high-conductivity material allocation in the volume-to-point problem for transient heat conduction.