In order to overcome the difficulty of parameter tuning caused by the large lag and time-varying nonlinearity of the tank for lysozyme fermentation, a temperature control method based on LRWOA-LSTM-PID is proposed in this paper. Firstly, according to the intrinsic mechanism of the fermenter, a temperature mechanism model based on a dynamic equation is designed, which can better reflect the temperature changes in the fermenter. Secondly, a Proportional Integral Derivative (PID) parameter tuning method based on a Long-Short Term Memory Network (LSTM) is proposed, which takes advantage of the ability of LSTM to learn time sequence information and obtains the variation trend between error sequences under continuous time sampling, thereby adjusting network weights more reasonably and accelerating PID parameter tuning. Finally, a Whale Optimization Algorithm (WOA) based on the Lévy flight and random walk strategy (LRWOA) is proposed for the initialization of LSTM parameters; this algorithm has excellent optimization capabilities and overcomes the problem of LSTM falling into local optimal solutions prematurely during parameter randomization. The results show that the method proposed in this paper can achieve rapid tuning of PID parameters, thereby improving the convergence speed of the system and reducing system overshoot.

中文翻译：

---

基于LRWOA-LSTM-PID的溶菌酶发酵温度控制方法

为了克服溶菌酶发酵罐大滞后和时变非线性带来的参数整定困难，提出一种基于LRWOA-LSTM-PID的温度控制方法。首先，根据发酵罐的内在机理，设计了基于动力学方程的温度机理模型，能够更好地反映发酵罐内的温度变化。其次，提出了一种基于长短期记忆网络（LSTM）的比例积分微分（PID）参数整定方法，该方法利用LSTM学习时序信息的能力，获得连续情况下误差序列之间的变化趋势。时间采样，从而更合理地调整网络权值，加速PID参数整定。最后，提出了一种基于Lévy飞行和随机游走策略（LRWOA）的鲸鱼优化算法（WOA）来初始化LSTM参数；该算法具有优异的优化能力，克服了LSTM在参数随机化过程中过早陷入局部最优解的问题。结果表明，本文提出的方法能够实现PID参数的快速整定，从而提高系统的收敛速度，减少系统超调。