基于OSELM-SGPC的主蒸汽温度优化控制仿真研究  

Simulation Study of OSELM-SGPC-Based Optimal Control of Main Steam Temperature

作　　者：何华靖 HE Huajing(Guizhou Yaxi Power Generation Co.,Ltd.,Zunyi 563108,China)

机构地区：[1]贵州鸭溪发电有限公司,贵州遵义563108

出　　处：《自动化仪表》2023年第3期54-59,共6页Process Automation Instrumentation

摘　　要：为解决比例积分微分(PID)控制器难以在大迟延、大惯性的主蒸汽温度被控对象上取得理想控制效果的问题,通过改进极限学习机(ELM)网络构建预测模型控制器,在解决传统神经网络算法训练速度慢、模型参数选取复杂的问题的同时,对主蒸汽温度进行多步预测,从而更好地抑制扰动。将阶梯式广义预测控制器(SGPC)作为滚动优化控制器,进一步优化计算过程,从而搭建了在线贯序极限学习机-阶梯式广义预测控制器(OSELM-SGPC),充分兼顾预测效果与计算复杂度。通过主蒸汽温度设定值扰动、给煤量扰动以及给煤量和给水量叠加扰动等试验,并与传统串级PID控制作对比,充分验证了OSELM-SGPC的有效性和优越性。To solve the problem that the proportional integral differential(PID)controller is difficult to achieve the ideal control effect on the main steam temperature controlled object with large latency and inertia,a predictive model controller is constructed by improving the extreme learning machine(ELM)network,which can make multi-step prediction of the main steam temperature while solving the problems of slow training speed and complicated selection of model parameters of the traditional neural network algorithm,so as to better suppress disturbances.The stepped generalized predictive controller(SGPC)is used as a rolling optimization controller to further optimize the computational process,thus building an online sequential entreme learning machine-stepped generalized predictive controller(OSELM-SGPC),which fully balances the prediction effect and computational complexity.The effectiveness and superiority of the OSELM-SGPC controller is fully verified through the tests of main steam temperature setpoint perturbation,coal feed perturbation and superposition perturbation of coal and water feed,and comparison with the conventional cascade PID control.

关 键 词：超临界机组 主汽温 神经网络预测控制 阶梯式广义预测控制器 在线贯序极限学习机 多步预测 抗扰动 

分 类 号：TH81[机械工程—仪器科学与技术]