基于神经网络预测ORC系统的最佳蒸发温度

有机朗肯循环系统(ORC)的蒸发温度是决定系统净发电量的关键参数。采用热力学的方法建立数值模型,计算了不同热源温度、冷凝温度及蒸发器夹点温差下的最佳蒸发温度。以此为样本,对神经网络模型进行训练,研究神经网络对ORC系统最佳蒸发温度的预测效果。结果表明,训练速率为0.4、隐层神经元数目为5、训练函数为“trainlm”时,神经网络的预测精度最高。采用两种方式对神经网络的预测结果进行验证,具体为:(1)以9:1比例划分训练集和验证集,(2)系统输入参数取值范围内随机生成100组数据作为验证集。两种验证方式的结果均显示,神经网络对ORC蒸发温度的预测值与数值模拟值较为接近,误差均在2%范围内,表明神经网络模型可以较好的预测ORC最佳蒸发温度,可以为ORC系统的运行参数优化提供参考。

Prediction of Optimal Temperature for Organic Rankine Cycle based on Artificial Neural Network

Optimal temperature of Organic Rankine Cycle(ORC)has been an inevitable parameter in generating maximum net power.Theoretical thermal mathematicalmodel is constructed to obtain optimal temperatures used as training samplesbyaltering heat source temperatures,condensing temperatures and pinch point temperatures.In this paper,artificial neural network(ANN)is installed to predict the optimal temperature to generate maximum power.The result shows that different parameters show great effects on prediction accuracy.Eventually,training rate,nodes number of hidden layer and training function are determined as 0.4,5,“trainlm”,respectively.After being trained repeatedly,the flowing two different methods are adapted to test accuracy of ANN model:(1)Samples are divided by 9(training samples):1(testing samples).(2)Generate 100 samples within the ranges of inputs randomly as testing samples.Compared the optimal temperatures obtained bytheoretical thermal mathematical model,the maximum relative errors tested with two different approaches are both less than 2%,which indicates that the proposed ANN model shows a strong ability to parametric optimization and it can be used in practical background.