<p>In this work, a 3D transient thermoelectric model of an aluminum electrolysis cell (AEC) is introduced, which considers variations in bath and ledge composition. To this end, a submodel accounting for key phenomena affecting bath chemistry and considering ledge composition variations is developed in Python and coupled with a transient thermoelectric finite-element slice model of an aluminum electrolysis cell. The submodel introduces a layered-structure approach to track ledge layers solidified under different cooling rates, each with a distinct composition. Accordingly, a new function is proposed to relate ledge composition to bath solidification rate. Considering a layered composition for the ledge affects mass balance and bath liquidus temperature, while the ledge thermophysical properties are kept constant. The full model is calibrated and validated using experimental data from an industrial power modulation trial. A ~ 14 pct decrease in input current from the nominal steady-state value leads to a peak solidification rate 90 minutes after the current decrease. During the 4 hours of power decrease, the bath liquidus temperature exhibits a significant decrease of 8.1 °C. Furthermore, the new model is challenged by simulating a power modulation case with several cycles of melting/freezing. The results show that incorporating bath solidification rate in bath chemistry calculations significantly influences bath/ledge phase change. After recovering the steady state at nominal current, the predicted ledge thickness is higher than its initial value, aligning with prior experimental findings. The proposed model advances the development of more precise transient thermoelectric models of AECs and paves the way for future progress in modeling ledge composition variations.</p>