Adaptive Modeling Strategy Integrating Feature Selection and Random Forest for Fluid Catalytic Cracking Processes

Abstract

This study proposes a hybrid approach for the modeling of the fluid catalytic cracking (FCC) process, with the aim to establish an adaptive and accurate product yield prediction model. Because of the uncertainties in crude oil quality and the complexity of the FCC process, which, for example, has highly coupled process variables with high dimensionality and strong interference, it is difficult for existing first-principles-based methodologies to deliver accurate results. To tackle this, this study proposes a machine-learning-based modeling approach that integrates an intelligent feature selection strategy with random forest for the process modeling. First, the adaptive immune genetic algorithm (AIGA) is applied to screen for the most relevant process indicators from the collected process data, including the operation parameters for the relevant process devices and the property data of the feed stream. Second, random forest (RF) is employed to establish the FCC process models based on the selected process indicators in the first step. The approach is illustrated by its application in a real FCC production process, for which 10 months of historical production data were used to train and test the proposed AIGA-RF model to determine the product yield predictions for four products. Comparisons between the proposed method and other methods were also conducted. The result indicates that the proposed method is able to remove the disturbance variables and is found to be adaptable to different product yield prediction scenarios. It could be a good reference for online process optimization and control of FCC processes.

Adaptive Data Dimensionality Reduction for Chemical Process Modeling Based on the Information Criterion Related to Data Association and Redundancy

“…Moreover, good adaptabilities and accuracies for chemical process modeling applications have also been reported. 26,55,56 Hence, this paper will continue choosing RF for the subsequent modeling process. A grid search strategy is then performed for tuning the parameters in RF to improve its The best metric values are bolded and underlined.…”

“…• Single-factor methods: Pearson correlation (Pearson-Corr), 20 Spearman correlation (SpearmanCorr), 21 dis-tance correlation (DistCorr), 22 mutual information (MI), 23,24 and maximal information coefficient (MIC) 25 • Optimization-based methods: genetic algorithm (GA) 26 and particle swarm optimization (PSO) 27 • Recursive feature elimination (RFE) 28 • Information entropy-based methods: joint mutual information (JMI), 29 joint mutual information maximization (JMIM), 30 minimum-redundancy maximum-relevance (MRMR), 31,32 and conditional mutual information maximization (CMIM) 33 • Random forest-based methods: mean decrease impurity (MDI) 34 and mean decrease accuracy (MDA) 35 • Regularization-based methods: Lasso, 36 Ridge, 37 and Elastic Net Feature extraction methods: PCA, KPCA, PLS, LLE, and LDA. Some of the above methods, including PearsonCorr, SpearmanCorr, DistCorr, MI, MIC, JMI, JMIM, MRMR, CMIM, MDI, and MDA, can assign importance scores to features and then sort them.…”

“…The data dimensionality reduction methods frequently used in machine learning and chemical process modeling are listed below: Single-factor methods: Pearson correlation (PearsonCorr), Spearman correlation (SpearmanCorr), distance correlation (DistCorr), mutual information (MI), , and maximal information coefficient (MIC) Optimization-based methods: genetic algorithm (GA) and particle swarm optimization (PSO) Recursive feature elimination (RFE) Information entropy-based methods: joint mutual information (JMI), joint mutual information maximization (JMIM), minimum-redundancy maximum-relevance (MRMR), , and conditional mutual information maximization (CMIM) Random forest-based methods: mean decrease impurity (MDI) and mean decrease accuracy (MDA) Regularization-based methods: Lasso, Ridge, and Elastic Net Feature extraction methods: PCA, KPCA, PLS, LLE, and LDA. …”

Comprehensive Machine Learning-Based Model for Predicting Compressive Strength of Ready-Mix Concrete

“…Principal component analysis is a typical filter method that performs feature selection based on the statistical performance of the original dataset and is independent of the subsequent learning algorithm. Different from the filter methods, the wrapper methods tightly couple the subsequent prediction algorithm with the feature selection process [ 32 ]. In other words, the feature selection process is optimized based on the feedback from the subsequent algorithm performance.…”

A Novel Method to Investigate the Activity Tests of Fresh FCC Catalysts: An Experimental and Prediction Process from Lab Scale to Commercial Scale

“…Fluid catalytic cracking (FCC) technology has been (and is still) one of the most important conversion processes in petroleum refinery for converting heavy fractions to more valuable fuels, such as gasoline, diesel, liquefied petroleum gas (LPG), olefinic gases, and some other products [1][2][3]. Due to the high flexibility of operation for different types of feedstocks, such as biomass-derived feedstocks, FCC technology has been long-lasting, and witnessed several stages of developments and revolutions for catalyst, feedstock, process technology, and reactor design [4][5][6].…”