基于BP神经网络的热舒适性指标计算

人的舒适性与其影响因子之间存在着复杂的非线性关系，由于人工神经网络能够反映非线性的映射关系，因此利用神经网络可实现对舒适性指标的计算．首先对BP神经网络做了一个简要介绍，并建立了一个用于训练计算舒适性PMV指标的3层网络．以Fanger公式计算了输入样本的期望输出值，利用输入样本与期望输出值对网络进行训练．神经网络达到误差要求后计算PMV指标，其中未经网络训练的样本计算结果与Fanger公式计算结果平均偏差为0．026，最大偏差为0．113，计算误差均在工程应用的允许范围之内．

Thermal Comfort Index Computing Based on BP Neural Networks

Human's comfort has a complicated non-linear relation with its influencing factors.Since the artificial neural networks can reflect the non-linear mapping relation,the neural networks can be used to compute the comfort index.The BP neural networks are simply introduced in this paper and a 3-layers BP neural network is established to compute the comfort PMV index.The Fanger formula is used to calculate the expected output values of the input samples and the network is trained by the input samples and the exp...