内部热耦合精馏塔基于混合优化算法的RBF神经网络内模控制策略

Internal thermally coupled distillation column based on RBF neural network of internal model control by hybrid optimization algorithm

内部热耦合精馏(ITCDIC)是迄今为止所提出的四大精馏节能技术中节能效果最高,但唯一没有商业化的节能技术,比常规精馏要节能40%以上,没有商业化的主要种类基于减聚类、K-means原因之一在于该过程具有较强的非线性、复杂动态特性以及耦合性,给控制方案的设计带来了诸多困难。由于径向基(RBF)神经网络具有快速学习并能逼近任意非线性函数的优点,本文提出了一种基于RBF神经网络内模控制的混合优化算法,是一种粒子群优化的混合优化算法,以苯-甲苯物系作为研究实例,并与国际公开报道的结果进行了详细比较,研究结果表明基于混合优化算法的RBF神经网络内模控制相比于传统的PID、常规RBF算法和国际公开报道有着更好的控制效果。

Internal thermally coupled distillation is the most promising of four major distillation energy-saving technologies, which can save more than 40 % energy compared with traditional distillation process, but it has not been widely used. The bottleneck that prevents the process from being commercialized is the operational difficulties due to the nonlinearity, complex dynamics and interactive nature Of the process. The radial basis（RBF） neural network has fast learning and can identify any nonlinear function. We presented a hybrid optimization algorithm for RBF neural network, which is based on particle swarm optimization, gradient descent method, K-means clustering and subtractive clustering algorithm. Take the benzene-toluene system as a research case and compared with the international public reporting results. The hybrid optimization algorithm for RBF neural network is more reliable than PID algorithm, conventional RBF and the international public reporting results.