基于BP神经网络的PTC工质出口温度研究

RESEARCH ON THERMAL PERFORMANCE OF PTC BASED ON BP NEURAL NETWORK

在槽式抛物面太阳集热器的热性能研究中,数据往往具有随机性、非线性和不确定性等特点,采用传统建模方法经常做出大量假设,导致仿真精度不高且复杂。以槽式抛物面太阳集热器为研究对象,将传统理论模型与BP人工神经网络相互耦合,通过集热器热性能室外动态试验,建立工质出口温度的神经网络预测校正模型。引入Levenberg-Marquardt(LM)法对BP神经网络的权值及阈值进行优化。分析结果表明,预测校正模型可将绝对误差控制在3.8℃以内,相对误差保持在3.6%以内,可有效提高槽式抛物面太阳集热器热性能的仿真模型计算精度。

In the study of thermal performance of parabolic trough solar collector（PTC）,the data are often random,nonlinear and uncertain. Besides,since large amounts of assumption are involved in the traditional modeling methods,the outcomes are often inaccurate and intricate. Therefore,a predictive correction model were established to predict exitfluid temperature. The model was constructed by coupling the traditional model with BP neural network,and dynamicoutdoor thermal performance tests were performed. The Levenberg-Marquardt（LM）method was also applied to optimizethe weights and thresholds for the classic BP Newton algorithm. The results revealed that the absolute error of thepredictive correction model was under 3.8 ℃ and the relative error was lower than 3.6%,which suggested that thepredictive correction ANN model has a higher accuracy. It also indicated a lower complexity in calculation.