Recursive Battery Capacity Estimation via Total Least Squares

The patent transforms the capacity estimation problem from a complex nonlinear optimization problem into a simple linear regression problem by changing the parameter representation. Instead of directly estimating capacity C, the method estimates parameters α and β in the linear equation z(t) = α + βQ(t), where Q(t) is integrated current. This parameter transformation enables accurate capacity estimation using simple linear regression, dramatically reducing computational complexity while maintaining high precision.

The patent enables the battery system to perform self-diagnosis and capacity estimation using only its own operational data (state-of-charge measurements and current measurements during normal operation). The method does not require external testing equipment or imposed operational constraints - the battery system uses its own z(t) and Q(t) data from regular use to continuously estimate its capacity, making the process completely non-invasive and operationally flexible.

The patent replaces complex mechanical/invasive testing procedures with a mathematical modeling approach. Instead of using invasive tests to eliminate noise or obtain accurate measurements, the method uses linear regression that inherently accounts for noise in both z(t) and Q(t) measurements. The linear model z(t) = α + βQ(t) with least-squares fitting naturally handles measurement uncertainties without requiring additional sensors, testing equipment, or invasive battery procedures.