Identification of hinging hyperplane autoregressive exogenous model using efficient mixed-integer programming.

Area of Science:

• Systems and Control Engineering
• Computational Science
• Nonlinear Dynamics

Background:

• Hinging hyperplane autoregressive exogenous (HHARX) models offer accurate approximation of nonlinear processes using piecewise affine forms.
• Traditional mixed-integer programming methods for HHARX identification are computationally intensive, limiting their application to small datasets.

Purpose of the Study:

• To develop a computationally efficient algorithm for HHARX model identification suitable for large-scale datasets.
• To enhance the practical applicability of piecewise model identification by addressing challenges like limited submodel switching and missing data.

Main Methods:

• A novel sequential optimization approach is proposed for HHARX model identification.
• The algorithm leverages mixed-integer programming techniques in a computationally efficient manner.
• The framework is designed to handle practical scenarios including limited submodel switching, missing output data, and specified steady states.

Main Results:

• The proposed algorithm demonstrates significant improvements in computational efficiency compared to traditional methods.
• Accurate HHARX models were successfully identified for large-scale experimental data.
• The method effectively handled complex piecewise model identification cases, including missing data and specified steady states.

Conclusions:

• The developed sequential optimization approach provides an efficient and accurate method for HHARX model identification.
• This advancement enables the application of HHARX modeling to larger and more complex real-world problems.
• The framework's ability to handle practical challenges enhances its utility in systems and control engineering.