Abstract
Nonlinear system prediction plays an important role in the practical thermal process, and deep learning algorithm is now popular in nonlinear dynamic system modeling because of its powerful learning ability. In this paper, the dynamic artificial neural networks (DANNs), which can be divided into two different types with external dynamic characteristics and internal dynamic characteristics, are analyzed. The mathematical formulations of feedforward deep neural network (DNN), traditional recurrent neural network (RNN) and Long-Short Term Memory network (LSTM) models are given. Furthermore, the structure of deep Hybrid Neural Network (DHNN) is described. Finally, the applicability of the above models in the thermal nonlinear process with different structural features is discussed. Simulation experiments reveal that DANNs with internal dynamic characteristics more suitable for solving thermal nonlinear system modeling problems with unknown order, and DHNN based on LSTM model has performed much better in approximating the dynamics of the thermal process with state parameters.
Highlights
Recent years, system modeling has made great progress due to the huge demand for controller design, and process analysis [1, 2]
System prediction for nonlinear system usually has developed by focusing on specific classes of system and can be broadly categorized into five basic description approaches, each defined by a model class: Volterra series models [5], block structured models, neural network models [6], NARMAX models [7], State-space models [8]
We apply deep neural network (DNN), recurrent neural network (RNN), Long-Short Term Memory network (LSTM) and deep Hybrid Neural Network (DHNN) based on LSTM models to thermal nonlinear dynamic prediction and compare the prediction ability of the above models in two situations: one is nonlinear dynamic objects with known order, the other is unknown input and output parameter order
Summary
System modeling has made great progress due to the huge demand for controller design, and process analysis [1, 2]. Compared with FFNN's difficulty, the neural network which include the dynamics directly into its structure can learn the dynamics of the system without requiring any apriori knowledge regarding the system. These neural networks with self-feedback loops are called RNN. The deep hybrid neural network based on DNN and RNNs is rarely applied in thermal nonlinear dynamic processes modeling/ identification. We apply DNN, RNN, LSTM and DHNN based on LSTM models to thermal nonlinear dynamic prediction and compare the prediction ability of the above models in two situations: one is nonlinear dynamic objects with known order, the other is unknown input and output parameter order. A brief mathematical theory of the deep neural networks which can represent the above types is introduced
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
