多边界条件下热泵利用循环水余热的CPCS-RBF预测控制  

Heat Pump CPCS-RBF Predictive Control Based on Multiple Boundary Conditions in Circulating Water Waste Heat Recovery System

作　　者：周洪煜[1] 杜学森 张振华 黄耀珍 

机构地区：[1]重庆大学动力工程学院,重庆市沙坪坝区400030 [2]河南恩湃高科集团有限公司,河南省郑州市450016 [3]中国大唐集团科学技术研究院,北京市西城区100032

出　　处：《中国电机工程学报》2015年第3期645-651,共7页Proceedings of the CSEE

基　　金：中国大唐集团科学技术研究院重点科研项目(KYZ2013009)~~

摘　　要：循环水余热回收系统中,热泵热网水出口温度在跟踪供热负荷需求时,在受驱动蒸汽量的调节的同时,往往易受热网回水、循环水等工况变化的影响,传统PID控制方式超调量大、负荷跟踪能力差。提出一种混沌变异克隆选择-径向基函数(CPCS-RBF)直接多步预测控制策略,以热泵热网水出口温度预测值与设定值差值为目标函数,利用CPCS优化算法求取目标函数最小时的驱动蒸汽最佳值。预测模型由2个RBF神经网络结合热泵现场运行数据构建,以提高热泵系统适应工况变化的能力;实验结果表明,该控制策略能综合学习热网回水温度、循环水温度等参数的变化,使驱动蒸汽调门超前动作,及时跟踪供热负荷需求变化的同时,适应发电负荷变化下排气余热量的波动,具有更好的节能效果和变工况适应能力。In the circulating water waste heat recovery system, when heat pump heating net water outlet temperature trace heating load demand, that＇s not only adjusted by driven steam capacity, and is easily influenced by operating conditions variation of the heating net backwater and circulating water, the traditional PID control method has a large overshoot volume and a poor load tracking ability. So a chaotic particle clone selection（CPCS）- radial basis function（RBF） direct multi-step predictive control strategy was proposed, with difference between heat pump heat supply network water outlet temperature predicted value and the set values as the objective function, using CPCS optimization algorithm to calculate the optimal values of driven steam when the objective function is the minimum. The prediction model was constructed by two RBF neural networks according to the field operation data in order to improve the model variable condition adaptability. The experimental results show that the control strategy can comprehensively learn the change of the parameters such as the heating net backwater temperature and circulating water temperature, and make driven steam tone act in advance, trace heating load demand change in time, and adapt fluctuation of exhaust gas residual heat under power generation load change, so has better energy saving effect and variable condition adaptability.

关 键 词：循环水余热 直接多步预测控制 混沌变异克隆选择 驱动蒸汽 径向基函数(RBF)神经网络 

分 类 号：TK32[动力工程及工程热物理—热能工程]