LSTM-CNN Network-Based State-Dependent ARX Modeling and Predictive Control with Application to Water Tank System

Industrial process control systems commonly exhibit features of time-varying behavior, strong coupling, and nonlinearity. Obtaining accurate mathematical models these nonlinear achieving satisfactory performance is still a challenging task. In this paper, data-driven modeling techniques deep learning methods are used to accurately capture category smooth system’s spatiotemporal features. The operating point may change over time, their characteristics can be locally linearized. We use fusion the long short-term memory (LSTM) network convolutional neural (CNN) fit coefficients state-dependent AutoRegressive with eXogenous variable (ARX) model establish LSTM-CNN-ARX model. Compared other models, hybrid more effective in capturing due its incorporation strengths LSTM for temporal CNN spatial characteristics. model-based predictive (MPC) strategy, namely LSTM-CNN-ARX-MPC, developed by utilizing model’s local linear global comparison experiments conducted on water tank system show effectiveness MPC methods.