遗传神经网络在室内环境热舒适度融合评价中的应用研究

Research on Genetic Neural Network Fusion for Evaluation of Indoor Thermal Comfort Degree

针对室内环境热舒适度评价,为解决影响PMV(predicted mean vote)指标的各因素之间复杂的非线性关系,利用核主成分分析KPCA(kernel principal component analysis)的非线性映射方法,对输入变量进行特征提取,以消除各因素之间的非线性关系,然后利用遗传神经网络GNN(genetic neural network)进行融合评价。对比GNN和KPCA+GNN的仿真评价结果可知:对于该室内热环境舒适度融合评价问题,KPCA能提取影响PMV指标的主要因素成分,KPCA+GNN是有效的预测方法。

Aiming at the evaluation of indoor thermal comfort degree, and in order to solve the complex nonlinear relationship between the influencing factors of PMV （Predicted Mean Vote）index, the non-linear mapping approach of KPCA （ kernel principal component analysis） is introduced to extract characteristics of input variables and to eliminate the nonlinear relationship between variables. Then based on GNN（ genetic neural network）, the fusion evaluation of indoor thermal comfort degree is implemented. By the comparison of GNN and KPCA ＋ GNN, the simulative results show that： for the fusion evaluation of indoor thermal comfort degree, KPCA can extract the main influencing factors of PMV index, and KPCA ＋ GNN is an effective forecasting approach with high accuracy.