Enhancing Air Pre-Heater Temperature Control Using Hybrid Machine Learning and Optimization Techniques

Controlling the temperature in Air Pre-Heater (APH) systems is key to energy efficiency in the industry. Traditional controllers like Proportional Integral Derivative (PID) and Model Predictive Controllers (MPC) struggle to adapt to APH systems' dynamic nature. The purpose of this study is to examine machine learning regression models such as Support Vector Regression (SVR), Decision Trees, and Random Forests in order to predict the temperature of the APH accurately. The model's performance was improved using advanced tuning methods such as Particle Swarm Optimization (PSO), Bayesian Optimization, and a hybrid PSO-Bayesian approach. It is found that the Random Forest model optimized with the hybrid PSO-Bayesian method performs best, resulting in a Root Mean Square Error (RMSE) of 0.450, a Mean Square Error (MSE) of 0.243, and an R2 score of 1.094. Comparatively, the SVR model (with RBF kernel) has higher errors: RMSE = 4.198, MSE = 17.624, R2 = 0.845. With RMSE = 1.696, MSE = 2.877, and R2 = 0.975, the Decision Tree model is effective; however, it overfits. Combining machine learning with hybrid optimization techniques can greatly enhance industrial automation, according to these results. In this way, APH systems become smarter, more flexible, and more energy-efficient.