基于自适应RBF神经网络预测堆芯热工水力参数的方法研究

Research on adaptive RBF neural network prediction method for core thermal-hydraulic parameters of fast reactor

反应堆堆芯热工参数的变化直接影响反应堆的安全,准确预测反应堆堆芯在各种工况下的关键热工参数变化趋势,能够大幅度提高反应堆的安全性,有效防止核电厂事故的发生。堆芯内热工水力特性参数受诸多因素的影响,为对其预测方法进行初步研究,确定神经网络预测的可行性,本文选用中国实验快堆(China Experimental Fast Reactor,CEFR)为研究对象,以燃料包壳表面最高温度、质量流量为预测量,通过子通道程序Subchanflow生成数据样本后,使用目前应用较为广泛的两种自适应神经网络方法自行开发预测程序,开展CEFR燃料组件稳态工况下热工参数预测分析,以及选用1/2的CEFR堆芯为研究主体,开展瞬态工况下热工参数的单步与连续预测分析。结果表明:相较于自适应反向传播(Backpropagation,BP)神经网络,自适应径向基(Radial Basis Function,RBF)神经网络具有更强的拟合能力和更高的预测精度,其在稳态工况下最大误差为0.5%;在瞬态工况下,存在个别局部点预测精度较差,但总体上自适应RBF神经网络在温度和质量流量预测良好,温度平均相对误差不超过1%,而流量平均相对误差不超过6%。自适应RBF神经网络模型能够在流动不稳定情况下提供较短时间内的实时预测,其预测结果具有一定参考价值。

[Background]Alterations in thermal parameters directly affect the safety of reactors.Accurately predicting the trends of key thermal parameters under various working conditions can greatly improve reactor safety,thereby effectively preventing the occurrence of nuclear power plant accidents.The thermal and hydraulic characteristic parameters in the reactor are affected by many factors.[Purpose]This study aims to explore the prediction method of core thermal-hydraulic parameters of fast reactor and determine the feasibility of neural network prediction.[Methods]The China experimental fast reactor(CEFR)was selected as the research object,and the maximum temperature of the fuel cladding surface and mass flow rate were used as predictor variables.After data samples were generated through the Subchannel code(named Subchanflow),two widely used adaptive neural networks were employed to perform the thermal parameter forecast analysis of CEFR fuel assembly under steady state conditions,and CEFR 1/2 core was taken as subject to carry out single-step and continuous predictive analysis of thermal parameters under transient conditions.[Results]The results show that,compared with adaptive BP neural network,the adaptive RBF neural network exhibits a better fitting ability and higher forecasting accuracy,and its maximum error under steady-state conditions is 0.5%.Under transient conditions,some local points have poor forecasting accuracy,however,the adaptive RBF neural network is generally excellent at predicting temperature and mass flow.The average relative error of temperature does not exceed 1%,and the average relative error of flow does not exceed 6%.[Conclusions]The adaptive RBF neural network model can provide real-time forecasting in a short time under unstable flow conditions,and its forecasting results have certain reference value.