pH NEUTRALIZATION PROCESS MODELLING AND CONTROL USING LONG SHORT-TERM MEMORY NETWORK-BASED NONLINEAR MODEL PREDICTIVE CONTROL

pH modelling and control are of great importance in many chemical and biological processes. However, achieving accurate pH control can be challenging using linear control methodologies due to the highly nonlinear nature of pH dynamics. This paper presents an innovative approach using Long Short-Term Memory (LSTM) neural networks for modelling and control of the pH neutralization process. In this approach, an LSTM model was trained using input-output data collected from the pH process. The trained model was then employed as a prediction model in a nonlinear model predictive control (NMPC) algorithm. Finally, the designed LSTM-based NMPC was implemented on a simulation model of the pH process. Validation results demonstrated that the trained LSTM model achieved an excellent fit. The closed-loop results showed that LSTM-based NMPC exhibited superior setpoint tracking and disturbance rejection, comparable to the NMPC based on first principle (FP) model. However, the LSTM-based NMPC significantly outperformed the linear model predictive control (LMPC).