融合浸出过程流程特性的梯形LSTM终点pH值模型预测控制

Model Predictive Control of Terminal pH Value of Trapezoidal LSTM Integrating Process Characteristics of Leaching Process

在锌冶炼浸出过程中,pH值的稳定控制是影响生产成本和产品质量的关键。针对浸出过程连续反应搅拌釜(CSTR)机理复杂、入口条件波动大、反应大滞后的问题,提出了融合流程特性的梯形长短期记忆网络模型预测控制(TLSTM-MPC)算法。首先,分析了影响浸出过程pH值稳定控制的关键因素,结合CSTR的级联特性,提出一种表征CSTR物理特性的梯形长短期记忆网络结构(TLSTM)模型,使预测模型具有较好的物理可解释性。其次,针对入口条件波动和反应过程大滞后的问题,提出了TLSTM-MPC算法,并采用粒子群算法进行优化求解。最后,采用某大型浸出过程现场运行数据进行验证。结果表明,该算法的超调量和调节时间相对传统控制方法分别降低了51.2%和57.4%,且在入口波动的情况下能够快速地稳定pH值。该方法对浸出过程的稳定、高质量生产具有较大的工程实践应用价值。

In the process of zinc leaching process,the stable control of pH value is the key factor to affect the production cost and product quality.A trapezoidal long short-term memory model predictive control(TLSTM-MPC)method integrating process characteristics is proposed to solve the problems of the complex mechanism of continuous stirred-tank reactor(CSTR),large fluctuation of inlet conditions and large lag of reaction.Firstly,the key factors affecting the pH value stability control are analyzed.Combined with the cascade characteristics of CSTR,a trapezoidal long short-term memory network structure(TLSTM)model characterizing the physical characteristics of CSTR is proposed,which makes the prediction model have better physical interpretability.Secondly,the TLSTM-MPC algorithm was proposed to solve the problems of inlet condition fluctuations and large lag in the reaction process,and particle swarm optimization algorithm was used for optimization solution.Finally,the field operation data of a large-scale leaching process is used to verify.The results show that the overshoot and adjustment time of this algorithm are reduced by 51.2%and 57.4%compared to traditional control methods,respectively,and it can quickly stabilize the pH value under inlet fluctuations.The method proposed in this paper has potential engineering practical application value for the stable and high-quality production of the leaching process.