基于GRU神经网络的城市建筑垃圾产量预测  

Prediction for urban construction waste output based on GRU

作　　者：黄莺[1] 杨馥宇 HUANG Ying;YANG Fuyu(School of Civil Engineering,Xi'an University of Architecture and Technology,Xi'an Shaanxi 710055,China)

机构地区：[1]西安建筑科技大学土木工程学院,陕西西安710055 [2]西安建筑科技大学安德学院,陕西西安710055

出　　处：《工业安全与环保》2024年第10期93-98,共6页Industrial Safety and Environmental Protection

摘　　要：针对具有时间序列特征的城市建筑垃圾产量预测精度低的问题,提出了一种基于门控循环单元(GRU)神经网络的预测模型。首先,根据《河南统计年鉴》收集了2002-2022年郑州市建筑垃圾产量和相关影响因素数据并进行相关性分析。然后,构建GRU神经网络预测模型,划分训练集与验证集,确定隐藏层神经元个数,同时采用自适应矩估计(Adam)算法更新梯度。最后,引入循环神经网络(RNN)和长短期记忆网络(LSTM)模型进行比较,选择绝对误差(MAE)、百分比误差(MAPE)以及相关系数(R^(2))验证模型性能。结果表明,GRU模型的预测结果更接近真实值,且具有更强的线性相关性。此外,利用该模型对2023-2027年郑州市建筑垃圾产量进行预测,结果显示到2027年建筑垃圾将达到8675.84万t。To address the issue of low prediction accuracy for urban construction waste(C&W)generation with time series characteristics,a prediction model based on Gated Recurrent Unit(GRU)neural network was proposed.Initi ally,data on Zhengzhou's construction waste output and relevant influencing factors from 2002 to 2022 were collected and analysed for correlation using the"Henan Statistical Yearbook".Following this,a GRU neural network prediction model was constructed.The dataset was divided into training and validation sets,and the number of neurons in the hid den layer was determined,in addition,the Adaptive Moment Estimation(Adam)algorithm was employed for gradient updating.Lastly,the model's performance was compared with Recurrent Neural Network(RNN)and Long Short Term Memory(LSTM)network,using metrics such as mean absolute error(MAE),mean absolute percentage error(MAPE),and the R-squared(R^(2))coefficient.The results showed that the GRU model's predictions were closer to the actual values and exhibited stronger linear correlation.Furthermore,utilizing this model,a forecast was made for Zhengzhou's construction waste output during 2023 to 2027,predicting that by 2027,the construction waste will reach approximately 86.7584 million tons.

关 键 词：城市建筑垃圾 GRU网络 深度学习 时间序列 产量预测 

分 类 号：X799.1[环境科学与工程—环境工程] TP183[自动化与计算机技术—控制理论与控制工程]