Accurately predicting the photosynthetic rate of facility-grown grapes is often required under complex and variable environmental conditions. Cultivation practices can be optimized to enhance the efficiency of resource utilization. Conventional prediction models cannot frequently capture the inherent nonlinear interactions and temporal dynamic features in agricultural environments. The prediction can often depend on the isolated environmental factors or static features. This study aims to achieve the accurate and robust prediction of the photosynthetic rate in greenhouse grapes under these fluctuating conditions. A SPM prediction was also proposed using dynamic feature enhancement and phased Stacking fusion. This approach was systematically integrated the key environmental parameters, including air temperature, relative humidity, photosynthetically active radiation (PAR), and carbon dioxide concentration. A set of dynamic features were selected—such as the rates of change in temperature and humidity, time-lagged variables, and temporal period encoding. An enhanced dataset was constructed to better represent the transient environmental dynamics. A Stacking ensemble framework was employed to combine the predictions from multiple base learners: K-Nearest Neighbors (KNN), Gaussian Process Regression (GPR), Long Short-Term Memory neural networks (LSTM), and Support Vector Machine (SVM). In the varying physiological requirements of grapevines at different phenological stages—namely flowering, fruit expansion, and maturation—the weights were assigned to these base learners, and then dynamically adjusted for each stage using interpretations derived from Shapley Additive exPlanations (SHAP) values. The improved model was suitable for the stage-specific biological responses. The predictions from the base learners were then integrated using an eXtreme Gradient Boosting (XGBoost) model as the meta-learner, which was regularized to mitigate the overfitting for the model generalization. The performance and robustness at the three major growth stages were evaluated using a five-fold cross-validation protocol, coupled with the staged weighting strategy. Experimental results demonstrate that the SPM model consistently outperformed the conventional standalone models, such as KNN, GPR, LSTM, SVM, and XGBoost, particularly from the flowering to the expansion stage and finally to the maturation stage. The superior prediction accuracy was achieved to maintain the relatively low model complexity. In addition to comparisons with these single models, the SPM model was also evaluated against the advanced ensemble learning techniques, including Adaptive Boosting (AdaBoost), Bootstrap Aggregating (Bagging) and Blending. The best performance of the SPM model was verified in the synergistic combination of dynamic feature enhancement and growth phase-aware weighting over all stages. Particularly, the remarkable capabilities were obtained using the complex nonlinear relationships at the fruit expansion stage. A coefficient of determination (R²) was as high as 0.986. The error metrics—mean squared error (MSE), mean absolute error (MAE), and root mean squared error (RMSE)—were as low as 0.139 μmol/(m²·s), 0.179 μmol/(m²·s), and 0.368 μmol/(m²·s), respectively. Furthermore, the SPM model was registered the lowest Akaike Information Criterion (AIC) value among all models. The optimal balance between goodness-of-fit and model parsimony, in order to predict the photosynthetic rates at each growth stage, indicating the high prediction accuracy. The SPM framework can provide the clear interpretability of the contribution of various environmental drivers after SHAP analysis. Thereby the transparent prediction was offered to facilitate the physiological processes underlying photosynthesis. A reliable technical tool was also provided for the precision environmental control in the protected agriculture, enabling more efficient and sustainable greenhouse.

Serious lag and nonlinearity have been found in the temperature control of the livestock bioliquefaction device. Existing landfill, incineration, and chemical treatments cannot fully meet the large-scale production in recent years, such as secondary pollution, high costs, and long cycles. Fortunately, the bioliquefaction technology can also degrade them into high-value liquid amino acid fertilizers. Among them, temperature is one of the key parameters- emulsification sterilization is required for a constant temperature of 100°C for 30 min, while fermentation requires a constant temperature of 35°C for 24 h. However, the conventional PID control is difficult to cope with the insufficient dynamic and steady-state accuracy of the system. Therefore, this study aims to construct an efficient temperature control scheme for the harmless and resourceful processing of livestock wastes. An intelligent control strategy was also proposed using the wild horse optimization (WHO) and BP neural network PID (BPNN-PID). Specifically, the reactor contained a reaction vessel, a stirrer, and a jacket layer. The power of the heating rod was controlled by a controllable silicon voltage regulator, and the cooling control was adjusted by the cold-water valve. In hardware, the ST20 CPU module of the Siemens S7-200 Smart PLC was combined with the EM AM06 analog input/output expansion. The OPC UA protocol was used to realize the real-time data interaction between the industrial PC and the PLC. The monitoring interface was developed to record the temperature and pH values using KinSealStudio. An "offline optimization + online adjustment" architecture was adopted: Firstly, the offline optimization of BPNN initial weight thresholds was used by the Wild Horse Optimization, in order to avoid the local optimization and slow convergence; Then, a three-layer BPNN (input layer with 3 nodes: temperature deviation, total deviation, deviation change; hidden layer with 5 nodes; output layer with 3 nodes: PID parameters K_p, K_i, and K_d) was constructed. Online adjustment of the dynamic parameter was carried out to verify the effectiveness of this strategy. An anti-integral saturation module was also added to prevent system instability. A simulation platform was established in the Matlab/Simulink platform, with the conventional PID, BPNN-PID, PSO-BPNN-PID, and GA-BPNN-PID as the controls. There was a better control performance of the BP neural network PID control algorithm using Wild Horse Optimization, when a constant step input was applied in the first 1500 s. No overshoot was observed during temperature rise, indicating the faster convergence to the target. A regulation time of 484.9 s was shorter than the conventional PID and BPNN-PID by 86.7 and 32.8 s, respectively. The overshoot was reduced by 0.11 percentage points, compared with the conventional PID. It was also shorter than the PSO and GA optimized BPNN-PID by 11.2 and 25.5 s, respectively. The fast response and small overshoot were still maintained during the cooling stage from 100 °C to 35 °C. In the temperature control experiment of the livestock bioliquefaction device, the better control performance was achieved in the BP neural network PID control algorithm using Wild Horse Optimization, with the rising time similar to PSO and GA-BPNN-PID at 58 min, the lowest overshoot of 0.66%, and a regulation time of 69 min, which was shorter than conventional PID by 45 min, and BPNN-PID by 13 min. The steady-state error was 1.56%, which was reduced by 3.12, 1.57, 1.37, and 0.31 percentage points, respectively, compared with the rest four controllers. The output was more accurately approaching the set temperature. The WHO-BPNN-PID control algorithm performed better in regulating the speed and overshoot. The large hysteresis and nonlinear of the system were effectively reduced to fully meet the temperature control requirements of the biological liquefaction of diseased and dead livestock and poultry.

Environmental quality in layer houses has been one of the most important influencing factors on the health level and production performance of laying hens in the large-scale poultry industry. This study aims to explore the impact of environmental quality of layer houses on the production performance of laying hens in summer. An analytical method was also proposed using multivariate data fusion. Firstly, seven environmental factors were detected, including temperature, relative humidity, wind speed, light, and concentration of CO₂, NH₃, and PM_2.5, according to the thermal, light, and gas environment. Then, the membership function was used to determine the basic probability distribution function of each factor. The correlation coefficient matrix was also utilized to optimize the support and correlation matrix of environmental factors. After that, the environmental factors were weighted and normalized to obtain the credibility matrix and evidence weights. The basic probability distribution functions were achieved in the thermal, light, and gas environment groups. Finally, the improved D-S evidence theory was used to fuse and iterate the basic probability allocation functions of the three groups, in order to evaluate the environmental quality at each detection point of layer house. A comparison was then made to reveal the impact of environmental quality of layer houses on the production performance of laying hens. A validation experiment was conducted in a layer house with eight tiers of battery cages in summer. The experimental results indicated that the best location was achieved in the front on the lower four tiers of the layer house, in terms of the environmental quality and average laying rate. The worst location for average laying rate was in the middle of the layer house with the comprehensive evaluation of environmental quality normal. The best location for the average laying rate on the upper four tiers was in the middle with the comprehensive evaluation of environmental quality suitable, whereas, the location with the worst average laying rate and environmental quality was at the back of layer house. The most suitable one was found in the comprehensive evaluation of environmental quality at detection points with an average laying rate higher than 86%. While the normal was observed lower than 86%. Once the comprehensive evaluation of the environmental quality was suitable, the average laying rate was relatively higher. On the contrary, the average laying rate was lower with the normal comprehensive evaluation of environmental quality. Furthermore, the improved D-S evidence theory can be expected to accurately evaluate the environmental quality, whereas, the D-S evidence theory cannot, particularly when environmental evidence conflicts with each other. The findings can provide an effective way to accurately evaluate the environmental quality of layer houses in summer, in order to clarify the impact of environmental quality of layer houses on the production performance of laying hens.