基于神经网络模型的锅炉主蒸汽温度预测控制  

Predictive Control of Boiler Main Steam Temperature Based on Long Short-Term Memory Neural Network Model

作　　者：杨春来 殷喆 袁晓磊 金飞 YANG Chunai;YIN Zhe;YUAN Xiaoei;JIN Fei(State Grid Hebei Energy Technology Service Co.,Ltd.,Shijiazhuang 050021,China)

机构地区：[1]国网河北能源技术服务有限公司,河北石家庄050021

出　　处：《锅炉技术》2023年第5期30-36,共7页Boiler Technology

基　　金：河北省电力有限公司科技项目(TSS2020-19)。

摘　　要：为解决传统控制方法在火力发电机组蒸汽温度控制过程中存在的强非线性、大迟延的难题,提出了一种基于长短期记忆(LSTM)神经网络在线估计和粒子群(PSO)滚动优化的预测控制算法。该方法将常规串级控制系统的主回路控制器用预测控制器替代,采用LSTM神经网络建立主蒸汽温度控制系统的过程模型,通过多步预测实现了对复杂非线性系统模型的精确预测。利用PSO算法在线求解主蒸汽温度控制系统的最优预测控制律,避免了传统递推方法无法直接求解非线性优化问题。仿真结果表明:与传统主蒸汽温度串级控制策略相比,该控制算法明显改善了控制系统的快速性,抗扰能力较强,对主蒸汽温度这类具有非线性及模型不精确的被控对象有一定的参考价值。In order to solve the problem of strong nonlinearity and large delay in the traditional control method of boiler main steam temperature control system,a predictive control algorithm based on long short-term memory(LSTM)neural network and particle swarm optimization(PSO)is proposed.In this method,the main loop controller of conventional cascade control system is replaced by a predictive controller,the process model of main steam temperature control system is established by LSTM neural network,and the accurate prediction of complex nonlinear system model is realized by multi-step prediction.PSO algorithm is used to on-line solve the optimal predictive control law of main steam temperature control system,which avoids the inability of the traditional recursive method to directly solve the nonlinear optimization problem.The simulation results show that compared with the traditional main steam temperature cascade control,the control algorithm significantly improves the rapidity of the control system,and has strong anti-interference ability,has certain reference value for controlled objects with nonlinearity and model inaccuracy such as main steam temperature.

关 键 词：燃煤锅炉 主蒸汽温度 模型预测控制 长短期记忆神经网络 粒子群优化算法 

分 类 号：TK223.7[动力工程及工程热物理—动力机械及工程]