An electric automobile incorporating a secondary battery has a disadvantage such as a difficulty in knowing the remaining capacity accurately and in predicting the time when the remaining capacity becomes zero because of deterioration of the secondary battery. The internal resistance is estimated with high accuracy even when the secondary battery deteriorates. Data used for learning or estimation is a data group (also referred to as data with regenerative charging) that is limited to data acquired within a certain time range around the end of regenerative charging. Such data within the limited range is extracted, used for learning, and subjected to the estimation. Thus, a value of the internal resistance can be output with high accuracy, specifically, with a mean error rate of 1% or less.