【作者】 李永玲；

【导师】 王兵树； 马进；

【作者基本信息】 华北电力大学 ， 控制理论与控制工程， 2014， 博士

【摘要】 由于核电站运行十分复杂,且核能应用中可控裂变反应本身具有一定的风险性,一旦由于设备、系统失效、操作失误等原因导致泄漏,将可能会带来毁灭性的灾难,而运行人员核电站仿真机的培训对保证核电机组的安全经济运行具有重大的实际效果。随着高精度实时仿真机的纵深发展,大型核电机组的建模仿真与智能参数优化的研究显得尤为重要。本文利用机理分析、神经网络、流体网络等建模方法,分别从压水反应堆建模、冷却剂系统建模、核电一回路整体建模与仿真、模型参数优化等四方面进行了研究。主要内容包括：(1)建立了神经网络与机理分析相结合的反应堆混合计算模型。针对反应性计算模块,提出了一种应用神经网络实现控制棒负反应性、多普勒效应、慢化剂温度效应、溶解硼产生的负反应性的估算的方法,测试结果表明测试输出与目标值符合良好；建立了采用差分瞬跳法的中子通量计算模型、堆芯热功率及堆芯热传递的计算模型等。(2)建立了反应堆冷却剂系统各组成设备的机理模型。通过机理分析,建立了立式U型管自然循环蒸汽发生器两相动态数学模型；通过分析稳压器运行特性,建立了两相动态非平衡稳压器机理模型；通过流体网络的方法建立了冷却剂泵的数学模型。(3)以某900MW压水堆核电机组为例,编制了压水堆核电机组一回路系统各组成设备的模型算法并进行仿真分析,结果表明各模型参数变化趋势符合理论分析,证明了所建各个子模块本身合理性。通过模块搭接的方法建立了核电机组一回路整体模型,仿真结果表明所建模型稳态误差较小,动态特性符合理论分析,验证了所建核电机组一回路整体模型的正确性。(4)为提高模型精度,采用智能优化算法对模型中较难确定的参数进行优化。为了提高QPSO算法的收敛速度和寻优精度,提出了一种高速收敛QPSO算法(HSCQPSO)。应用国际标准函数进行测试,结果表明：在多数函数优化问题中,本文算法的收敛速度和寻优精度均取得了良好的效果。利用本文提出的算法对两相动态稳压器机理模型进行了参数优化,有效提高了模型精度,效果显著。结合本文建立的稳压器机理模型的基础上,设计了稳压器压力控制系统,同时应用本文提出算法优化控制器参数,仿真结果表明了本文提出算法具有较好的效果。针对上述内容研究的过程中,取得了一定的创新性成果：(1)提出了一种应用神经网络实现控制棒负反应性、多普勒效应、慢化剂温度效应、溶解硼产生的负反应性的估算的方法并建立了神经网络与机理分析结合的反应堆混合计算模型。(2)基于MATLAB通用平台,建立了900MW级压水堆核电站一回路系统的简化仿真模型,仿真验证表明稳态误差较小,动态特性合理。(3)提出了一种高速收敛量子粒子群优化(HSCQPSO)算法,并应用于两相动态稳压器机理模型进行了参数优化,有效提高了模型精度,效果显著。(4)设计了内模-PID稳压器压力控制系统,并应用本文提出HSCQPSO算法优化控制器参数,仿真结果表明本文设计的稳压器压力控制系统具有较强的鲁棒性和较好的调节品质。更多还原

【Abstract】 The devastating disaster will be happened because of leakage, which caused by the failure equipment or system and operating miss of nuclear power unit, because the nuclear power plant operation is very complex, and the application in fission reaction itself has some risk. However simulator training of operators has the serious practical effect of safety and economic operation of nuclear power plants. With the development of high precision real time simulation machine, it is very important for large-scale nuclear power generating units to research system modeling, simulation and intelligent parameters optimization. Pressure water reactor modeling, coolant system modeling, modeling and simulation of nuclear power plant primary loop, and model parameter optimization were researched in this paper, using mechanism analysis method, neural network method, fluid network modeling method and so on. The main innovative achievements are:(1) The reactor mixed calculation model based on the combination of neural network and mechanism analysis was set up. In the model of reactive module, we used neural network to realize the reactivity estimation of control rod negative reactivity, Doppler effect, moderator temperature effect, dissolved boron negative reaction. The test results indicated that the satisfactory effects are achieved. Though mechanism analysis, we built the reactor core thermal power calculation model, core heat transfer model and neutron flux by the method of divided instantaneous.(2) All the components mechanism models of the reactor coolant system were established. Through mechanism analysis, the dynamic mathematical model of two-phase natural circulation steam generator with vertical U-shaped tubes were built; by analyzing the characteristics of pressurizer, a two-phase dynamic non equilibrium pressurizer model was built; finally mathematical model of coolant pump was set up by using the method of fluid network.(3) All the component models of primary loop system of PWR nuclear power unit, such as reactor, steam generator, reactor coolant pump and pressurizer, were programmed into algorithms and simulated, taking a900MW PWR nuclear power unit as an example. The simulation results indicated that the trend of model parameters accords with the theoretical analysis, so the rationality of each sub module itself was proved. Depending on those models discussed above, the integral model of primary loop in nuclear power unit was finally established by module overlap method; the correctness of which was verified though analyzing the steady-state and dynamic characteristics. (4) To improve the accuracy of the model, the parameters which were difficult to determine were optimized with intelligent optimization algorithms. In order to improve the convergence speed and the optimization accuracy of QPSO algorithm, a new improved high speed convergence QPSO algorithm (HSCQPSO) was set up. The test results of international standard function indicated that:the convergence speed and the optimization accuracy of this algorithm were satisfied in most function optimization problems. The HSCQPSO algorithm was applied to optimize the parameters of two-phase dynamic pressurizer mechanism model, and improved the model accuracy effectively. Depending on the accurate model established in this paper, we designed pressure control system of pressurizer, and optimized the controller parameters using HSCQPSO algorithm in addition. The simulation results show that the proposed algorithm has better effect.Some innovative achievements are obtained, there are:(1) A method for realizing the reactivity estimation of control rod negative reactivity, Doppler effect, moderator temperature effect, dissolved boron negative reaction used neural network was realized, and calculation model combined mechanism analysis with reactor was proposed.(2) The integral model of primary loop in nuclear power unit was established based on Matlab, it has small steady-state error and reasonable dynamic characteristics.(3) A new improved high speed convergence QPSO algorithm (HSCQPSO) was set up, and was applied to parameters optimization of pressurizer.(4) An IMC-PID controller pressure was designed for pressure system of pressurizer, and optimized the controller parameters using HSCQPSO algorithm. The simulation results show that the proposed algorithm has better effect.更多还原