基于神经网络的建筑能耗预测

由于目前只有很少一部分建筑师能掌握复杂的建筑能耗分析,因此本文利用MATLAB建立BP神经网络,将影响建筑能耗的18个因素作为网络的输入,进行学习训练,最后通过测试样本点数据预测建筑能耗,并与DeST-h模拟计算得到的结果比较,发现相对误差在3.5%以内,验证了该网络模型的可行性.该方法使建筑师在设计阶段能够简单且准确地获得设计建筑的能耗.

The forecast for construction dissipation of energy base on neural network

As there are very few architects can analysis the complicated building energy consumption. This paper sets up an BP ANN model in MATLAB. Taking eighteen variables that affect building energy consumption as inputs for training, predicting the building energy consumption based on the test samples, and comparing with the results from the analysis of DeST-h, We found this model demonstrates its effectiveness, This method helps architects predict building energy consumption of the building under design simply and accurately.