Parallel-connected modules have been widely used in battery packs for electric vehicles nowadays. Unlike series-connected modules, the direct state inconsistency caused by parameter inconsistency in parallel modules is current and temperature non-uniformity, thus resulting in the inconsistency in the speed of aging among cells. Consequently, the evolution pattern of parameter inconsistency is different from that of series-connected modules. Since it’s practically impossible to monitor each cell’s current and temperature information in parallel modules on EVs, considering cost and energy efficiency, it’s necessary to study how the parameter inconsistency evolves in parallel modules considering the initial parameter distribution, topology design and working condition. In this study, we assigned cells of 18650 format into several groups regarding the degree of capacity and resistance inconsistency. Then all groups are cycled under different environmental temperature and current profile. The parameter of each cell and the current and temperature distribution is calibrated regularly. The result shows that, under the same working condition, a relatively larger initial parameter inconsistency would result in a faster deteriorate speed for parallel modules. However, a self-balancing mechanism in parallel modules is observed in the experiments, which could make the degree of parameter inconsistency smaller as the module ages, especially under low temperatures. An electro-thermal model for parallel modules is analyzed and built to simulate the aging phenomenon for parallel modules. The simulation result fits well with the experiments. The evolution pattern of parameter inconsistency is then analyzed and concluded using this model. Though under the limitations of time and conditions for more validations, this study may still be helpful for the design and usage of battery packs.