Variations of multiple environmental factors and multi-objective optimisation control in a pig house

With the increasing scale and intensification of pig farming, the environmental quality of pig houses has become crucial for pigs' health and reproductive ability. Numerous environmental factors in the pig house have formed a complex nonlinear microclimate environment that is dynamic, time-varying, and coupled, making it difficult to achieve precise control of the environment. In this study, sixty days of indoor environmental data were collected by an Internet of Things platform to analyse the diurnal changes in indoor temperature, humidity, and concentrations of CO₂ and NH₃, as well as the seasonal changes in winter and summer. A multi-objective control strategy for optimising the pig house environment using Double Deep Q-Network (DDQN) was established. The results showed that 1) there were significantly higher correlations between multiple environmental factors in summer than in winter; 2) the determination coefficients R² of the multiple linear regression models constructed with indoor temperature and CO₂ concentration as dependent variables reached 0.915 and 0.778, respectively; 3) the DDQN control strategy kept the indoor temperature variation within ±1.7 °C, compared to ±2.1 °C with the traditional temperature threshold control strategy (TTCS); 4) the total running time of the three fans in a day under the DDQN control strategy was 28.01 h, with the total power consumption of 11.4 kWh, and an energy-saving rate of 7.39 % and the indoor various environmental factors are closer to the setting values. This research offers feasible reference for the intelligent and energy-saving environmental regulation in large-scale pig production for precise environmental control.