Application of PSO-BP Neural Network Model in SCR system of tail Gas Post-treatment

• Published in: 2022 China Automation Congress (CAC) pp. 2404 - 2408
• Main Authors: Zhang, ZhengShan, Lv, XiaFu, Gong, WenBin, Wang, YePeng
• Format: Conference Proceeding
• Language: English
• Published: IEEE 25.11.2022
• Subjects: Ammonia; Data models; engine aftertreatment; Neural networks; Prediction algorithms; Predictive models; PSO-BP Neural Network Model; Selective catalytic reduction; Tail; Training
• ISSN: 2688-0938
• DOI: 10.1109/CAC57257.2022.10055283

With the release of the national standard GB17691- 2018, the exhaust emission index has entered a more stringent stage, SCR post-processing system as the most effective measure to reduce diesel engine NO x emissions [1]. This paper introduces the basic working principle of SCR and analyzes the problems faced by the traditional control mode. In view of the general control effect of open-loop pulse spectrum or closed-loop PID control, which is easy to cause ammonia leakage and other problems, this paper tries to use neural network algorithm to predict NO x emissions.Due to the excellent adaptability and fault tolerance of BP neural networks. In this paper, we decided to use BP neural network algorithm for further prediction. However, because the initial weights of the BP neural network are set according to experience, this will reduce the fitting effect. In order to solve this problem, the PSO-BP model is proposed to predict the emission of NO x and improve the prediction accuracy of BP neural network algorithm [2]. Considering dieselengine NO x emission to have lag between the working parameters,the history and current value of those parameters put into the model.Dealing with the inputting parameters by normalization and the output data by reversing normalization in the post-processing module of model have been done.The result of calibration shows the model has satisfying prediction accuracy.