Monitoring and Early Warning of the Cage-Rearing Broiler Farming Environment Based on an Inspection Robot

In large-scale caged broiler houses, the combined effects of ventilation regulation, external meteorological disturbances, and changes in flock growth stages can easily cause the indoor thermal environment to deviate from the target settings, thereby increasing the difficulty of environmental control and production risks. Meanwhile, the broiler house environment exhibits pronounced spatial heterogeneity, while manual inspection is limited by low efficiency, high labor intensity, and delayed detection of anomalies. Therefore, this study proposes an inspection-robot-based method for predicting and providing early warnings for temperature deviation (TD) and temperature-humidity index (THI). In this method, TD is used to characterize the degree of deviation in environmental control, whereas THI is used to characterize the overall thermal environmental risk under the coupled effects of temperature and humidity. For the two prediction tasks, input variables were separately selected based on correlation analysis, and the same Wavelet-ECA-GRU model architecture was adopted for training to achieve short-term prediction over the next 30 min. The results show that the proposed model outperformed baseline models, including LSTM, TCN, and GRU. On the test set, the RMSE and R² for TD prediction were 0.2886 and 0.9238, respectively, while those for THI prediction were 0.4309 and 0.9287, respectively. Based on the TD classification results and THI risk thresholds, warning strategies for heat deviation, cold deviation, and humid-heat risk were established. The proposed method can identify potential thermal environmental anomalies in caged broiler houses in advance from the perspectives of environmental control deviation and comprehensive thermal environmental risk, thereby assisting farm managers in assessing whether timely adjustments to ventilation, heating, or cooling are needed.