摘要详情

ID / 提交时间

73 / 2025-04-15 19:07:31

标题

Environmental Prediction Research for Layer Houses Based on Fusion of Physics and Data-Driven Models

关键字

temperature and humidity prediction, dynamic thermal balance model, LSTM neural network

主题及专题

3. 智慧牧场与动物福利 > （4）智慧牧场应用的案例实践

状态

摘要待审

作者

Xin Li / China Agricultural University；College of Water Resources and Civil Engineering

Hao LI / College of Water Resources and Civil Engineering； China Agricultural University;Key Laboratory of Agricultural Engineering in Structure and Environment； Ministry of Agriculture and Rural Affairs;Beiji

摘要

Temperature and humidity, as core environmental parameters in facility farming, are directly linked to the production performance and disease risks of laying hens. Current environmental control systems predominantly rely on threshold-triggered lagged regulation, which struggles to address environmental fluctuations caused by multi-source dynamic coupling effects such as metabolic heat production and ventilation disturbances. Traditional prediction methods are often constrained by either insufficient accuracy of single-mechanism models or the lack of interpretability in purely data-driven models. To resolve these challenges, this study proposes a hybrid prediction method integrating dynamic thermal balance mechanisms and Long Short-Term Memory (LSTM) neural networks to address the temperature and humidity prediction challenges in high-rise cage layer houses. The method quantifies core environmental parameters—including metabolic heat production, ventilation heat exchange, and envelope heat transfer—based on the law of energy conservation, while simultaneously establishing a humidity balance equation to characterize evaporative cooling processes. To address nonlinear errors in the model, an LSTM network is introduced to construct a residual correction module, leveraging its temporal memory characteristics and gating mechanisms to autonomously capture dynamic coupling relationships within multi-dimensional environmental data. Utilizing real-time monitoring data from a layer house in Henan Province, the method optimizes dynamic response capabilities to complex environmental parameters. By ensuring the interpretability of thermodynamic mechanisms, it provides an innovative data-physics dual-driven approach to support intelligent environmental regulation in layer houses.

重要日期

会议日期

10月20日

2025

至

10月23日

2025
04月15日 2025

摘要截稿日期
05月01日 2025

摘要录用通知日期
06月30日 2025

初稿截稿日期
08月01日 2025

终稿截稿日期
08月31日 2025

初稿录用通知日期
10月23日 2025

注册截止日期

主办单位

International Research Center for Animal Environment and Welfare (IRCAEW)
Chinese Society of Agricultural Engineering (CSAE)
China Agricultural University (CAU)
Rongchang District People’s Government
The National Center of Technology Innovation for Pigs

承办单位

Chongqing Academy of Animal Sciences (CAAS)
Key Lab of Agricultural Engineering in Structure and Environment, Chinese Ministry of Agriculture, Beijing, China