Proceedings of TUSUR University / Archive / Issue of Journal № 3, t. 27, 2024 / Methodology to improve the quality of neural network modeling of dynamic objects

Methodology to improve the quality of neural network modeling of dynamic objects

DOI: 10.21293/1818-0442-2024-27-3-92-99

Download article in PDF format

JATS xml

Abstract: The problem of neural network modeling of nonlinear dynamic objects using recurrent neural networks is considered. An approach to improve the accuracy of modeling using a static neural network of the «multilayer perceptron» type, that processes correlation dependencies of a dynamic process and approximates the modeling error, is proposed. A technique for synthesis and application of the correlation neural network model CCF-MLP improving the quality of modeling of a conventional recurrent neural network, is formulated. Simulation experiments are carried out with a neural network recurrent network of the GRU type, that models the behavior of a nonlinear dynamic object, as well as GRU with the proposed CCF-MLP model. The improvement in the quality of modeling (RMSE, MAPE) is confirmed in the case of using CCFMLP both in the presence and absence of noise in the observed data. The practical applicability of the proposed method was tested on a real liquid level control system.

Keywords: cross-correlation function, multilayer perceptron, recurrent neural network, dynamic object modeling, nonlinear dynamic object

For citation:
Van S., Eliseev V. L. Methodology to improve the quality of neural network modeling of dynamic objects. Doklady Tomskogo gosudarstvennogo universiteta sistem upravleniya i radioelektroniki, 2024, vol. 27, no. 3, pp. 92–99. DOI: 10.21293/1818-0442-2024-27-3-92-99

Authors and copyright holders:

Van S. , National Research University «Moscow Power Engineering Institute» (Moscow, Russia)
Eliseev V. L. , National Research University «Moscow Power Engineering Institute» (Moscow, Russia)

1. Shumikhin A.G., Boyarshinova A.S. [Controlled Technological Objects Parametric Identification with Dynamic Neural Network Models]. Bulletin of Perm National Research Polytechnic University, 2015, no. 11, pp. 21–38 (in Russ.).
2. Schoukens J., Ljung L. Nonlinear system identification: A user-oriented road map. IEEE Control Systems Magazine, 2019, vol. 5, no. 4, pp. 28–99.
3. Yamada T., Yabuta T. Dynamic system identification using neural networks. IEEE Transactions on Systems, Man, and Cybernetics, 1993, vol. 23, no. 1, pp. 204–211.
4. Wang J., Chen Y. A fully automated recurrent neural network for unknown dynamic system identification and control. IEEE Transactions on Circuits and Systems I: Regular Papers, 2006, vol. 53, no. 6, pp. 1363–1372.
5. Legaard C., Schranz T., Schweiger G., Drgoňa J. Constructing neural network based models for simulating dynamical systems. ACM Computing Surveys, 2023, vol. 55, no. 11, pp. 1–34.
6. Gonzalez J., Yu W. Non-linear system modeling using LSTM neural networks. IFAC–PapersOnLine, 2018, vol. 51, no. 13, pp. 485–489.
7. Bonassi F., da Silva C.F., Scattolini R. Nonlinear MPC for offset-free tracking of systems learned by GRU neural networks. IFAC-PapersOnLine, 2021, vol. 54, no. 14, pp. 54–59.
8. Al Seyab R.K., Cao Y. Nonlinear system identification for predictive control using continuous time recurrent neural networks and automatic differentiation. Journal of Process Control, 2021, vol. 18, no. 6, pp. 568–581.
9. Vaswani A., Shazeer N.M., Parmar N., Uszkoreit J., Jones L., Gomez, A.N., Kaiser Ł., Polosukhin I. Attention is all you need. Advances in Neural Information Processing Systems, 2017.
10. Lin T., Goyal P., Girshick R., He K., Dollár P. Focal loss for dense object detection. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, vol. 42, no. 2, pp. 318–327.
11. Madry A. Towards deep learning models resistant to adversarial attacks. arXiv preprint arXiv:1706.06083, 2017.
12. Gonzalez J., Yu W. Non-linear system modeling using LSTM neural networks. IFAC-PapersOnLine, 2018, vol. 51, no. 13, pp. 485–489.
13. Eliseev V.L., Wang X. [Application of the cross-correlation function as a characteristic of the dynamics of non-stationary, non-linear objects and neural network models]. Materials of the 28th Baikal All-Russian Conference with international participation «Information and Mathematical Technologies in Science and Management». Irkutsk, 2023, 57 p. (in Russ.).
14. Wang X., Eliseev V.L. [Detecting Dynamic Object Model Disorders Using Correlation Function]. Materials of the Technical Conference of Students and Postgraduates «Radio Electronics, Electrical Engineering and Power Engineering». M.: Raduga, 2024, 301 p. (in Russ.).
15. Wang X., Eliseev V.L. [Neural Network Method for Detecting Changes in the Operating Mode of a Nonlinear Dynamic Object]. News of the Tula State University. Technical Sciences, 2024, no. 3, pp. 48–56 (in Russ.).
16. Wigren T., Schoukens J. Three free data sets for development and benchmarking in nonlinear system identification. 2013 European Control Conference (ECC). Zurich, 2013, pp. 2933–2938.