Operator-orthoregressive methods for identifying coefficients of linear difference equations

We propose a new family of operator-orthoregressive methods for identifying the coefficients of linear difference equations from measurements of noisy solution at short time intervals. This family includes special cases of orthogonal regression (TLS) and variational identification (STLS) methods. The conditions of identifiability, as well as quantitative indicators of local identifiability, based on the numerical characteristics of the ellipsoids of deviations of the identified coefficients at small disturbances in measurements, are obtained. Computational algorithms are mentioned.