The charging cut-off voltage is the voltage point that most electric vehicles (EVs) will experience during charging. There is a series of problems to be optimized such as ignoring the initial capacity error and battery aging by the traditional ampere-hour integration method. A two-layer ensemble extreme learning machine (Ensemble ELM) algorithm is proposed to realize the joint estimation of SOC and SOH under the charging cut-off voltage of lithium-ion batteries. First, this study extracts the health indicator (HI), which is easily measured, and establishes the model between HI, charging time and SOH based on an Ensemble ELM. Second, the easily measured HI is used to estimate the charging time, something that is difficult to measure online. The online SOC correction based on the ampere-hour integration method is realized at the charging cut-off voltage. This method fully considers the uncertainty of the initial charging state of EVs, and can guide EV users to charge reasonably. In addition, the problem of output instability of a single ELM model is solved by selecting appropriate integration degree of the ensemble ELM model. Finally, the proposed method is tested on NASA and CALCE datasets. The results show that the root-mean-square-error (RMSE) of SOC estimation is less than 1.5% for a lithium-ion battery at charging cut-off voltage. Compared with other common algorithms, the ensemble ELM shows a higher training and test accuracy with short estimation time.