Optimization and analysis strategy for phenol and ammonia recovery process with feeding variability: Introducing clustering and nonlinear regression

Abstract

Feeding variability appears extensively in phenol and ammonia recovery process of coal chemical plant, which has caused severe problems in effluent quality. This study is aim to take the feeding variability into consideration for process optimization and analysis of the phenol and ammonia recovery process, have been not studied in previous efforts. For this purpose, this study proposed an optimization and analysis strategy for phenol and ammonia recovery process with feeding variability which introduced clustering and nonlinear regression. Firstly, 56,245 datasets were collected and preprocessed. 27 scenarios were generated by k-means clustering based method to represent the feeding variability and from which 7 typical scenarios were identified. Then, process model and costing model was established through equilibrium-stage model of extractor and Aspen Plus. Steam consumption of solvent recovery tower was estimated more accurately by nonlinear regression via BAS-BPNN. The mean relative error with the industrial value was decreased to 3.9% from 47% of the above-mentioned process model. Finally, sensitivity analysis was conducted via Sobol’ method to pick decision operating variables for lower optimization computing cost. Based on these, deterministic optimization models were individually solved at seven typical scenarios via genetic algorithm, and scenarios-between contrastive analysis suggested that the feeding variability directly affected ideal process operational policy with the maximum cost gap of 571,648 US$/a between typical scenarios. Additionally, stochastic optimization was solved and the results showed that the expected costs were 2.31%, 3.17% and 9.04% larger than the deterministic model. Obviously, the increase in the process's robustness and decrease in the likelihood of process performance deterioration under feeding variability were at the expense of cost.