应用人工神经网络预测室内全年动态采光

Application of Artificial Neural Network for Predicting Indoor Annual Dynamic Daylighting

在建筑设计早期阶段,了解建筑形态参数与室内采光之间的关系对设计优化至关重要。本文采用多层感知器(Multilayer Perceptron,MLP)神经网络,以四种主要特征(室外遮挡情况、建筑形态特征、开窗设置、测点位置信息)作为MLP的输入参数,通过计算机模拟收集的数据来构建神经网络,预测室内的全年自然采光质量(UDI<100 lx、UDI 100~2000 lx、UDI>2000 lx)。研究结果显示多层感知器神经网络模型在测试集中的回归决定系数R 2为0.984,均方误差MSE为11.624,准确性较高。对神经网络进行权重分析的结果表明,外部遮挡物的高度和建筑进深对输出结果影响最为显著。而窗台底部的标高和测点距窗户的距离对输出结果UDI的影响较小。神经网络模型为建筑设计预测日光提供了一种新的智能方法,有助于辅助建筑早期的设计决策。

In the early stage of architectural design,understanding the relationship between architectural form parameters and interior daylighting is crucial for design optimization.This study employs a Multilayer Perceptron(MLP)neural network,takes four main features(outdoor occlusion situation,architectural form characteristics,window opening settings,and measurement point location information)as the input parameters of the MLP,and builds the neural network through the data collected by computer simulation to predict the annual indoor natural daylighting quality(UDI<100 lx,UDI 100~2000 lx,UDI>2000 lx).The research results demonstrate that the MLP neural network model achieved a regression coefficient R 2 of 0.984 and a mean squared error(MSE)of 11.624 on the test dataset,indicating high accuracy.The weight analysis of the neural network reveals that the external shading height and building depth significantly influence the output.In contrast,the elevation of window sills and the distance of measurement points from windows have a minor impact on the results.The neural network model provides a new intelligent approach for predicting daylight in architectural design,assisting in early-stage design decision-making.