Towards intelligent microclimate control in cattle housing: A dynamic-balance framework via multi-constraint optimization

• Multi-constraint dynamic framework optimizes cattle barn temperature and air quality. • Hybrid model fuses data trends with thermophysics for robust microclimate prediction. • EWMA-smoothed evolution with balance constraints ensures stable convergence. • Multi-curtain balance enforcement maintains self-adjustment capability and stability. • MBGMOPSO excels existing methods in balance, smoothness and dual-objective precision. Natural ventilation in cold-region cattle barn presents a fundamental conflict between air exchange and heat preservation. Traditional control struggles to reconcile the competing objectives of temperature maintenance and air quality management. To address this challenge and achieve coordinated optimization of temperature and CO₂ levels, this study proposes a Multi-Constraint Balanced-Guide Multi-Objective Particle Swarm Optimization (MBGMOPSO) algorithm. This approach integrates a data-driven Support Vector Machine (SVM) with a simplified physical model, incorporating spatial equilibrium and temporal smoothing constraints to regulate six curtain actuators. Comparative evaluation against MOPSO, NSGA-II, PID, and fuzzy logic controllers demonstrates regime-dependent performance characteristics. Relative to MOPSO and NSGA-II, MBGMOPSO attains the best curtain-coordination performance, cutting the north–south imbalance to 6.36% and outperforming PID and FLC by 74.2 % and 78.9 %, respectively, thereby eliminating airflow short-circuiting and ventilation dead zones. Among the multi-objective algorithms, it also achieves the highest smoothness index, surpassing PID and FLC by 16.2% and 41.9 % while reducing actuator wear. Under summer high-temperature saturated conditions, MBGMOPSO keeps temperature RMSE at 0.64°C and CO₂ RMSE at 57.14 ppm with the smallest variance, delivering a low-wear, season-robust precision ventilation strategy for cold-region barns.