论文部分内容阅读
本论文将17集总反应动力学模型和线性回归模型用于在线预测工业级连续重整装置的芳烃收率。首先对17集总模型进行了简化处理从而提高模型计算速度,然后使用偏最小二乘法(PLS)对线性回归模型的参数进行估计,从而提高模型求解精度和稳定性。离线验证结果证明两个模型都是非常适合的。在此基础上,提出了有效的在线预测和在线校正策略,两个模型的在线预测精度分别是0.52wt%(重量百分数)和0.39wt%,与离线模拟精度几乎相当。两个模型可以同时用于在线预测,发挥各自的优势。
In this paper, 17 lumped kinetic models and linear regression models were used to predict the aromatic yields of industrial continuous reformer on-line. Firstly, the 17-lumped model is simplified to improve the speed of the model calculation. Then, the parameters of the linear regression model are estimated by partial least squares (PLS) to improve the accuracy and stability of the model. Off-line verification results show that both models are well suited. On this basis, an effective on-line prediction and online correction strategy are proposed. The online prediction accuracy of the two models is 0.52 wt% and 0.39 wt%, respectively, which is almost the same as the offline simulation accuracy. Both models can be used for online forecasting at the same time, giving full play to their respective advantages.