Abstract:With the development of scaled pig farm, the environmentalcontrolled breeding production with closed house has got a rapid progress in recent years. However, in order to maximize the commercial interests, there are always limited living space designed for pigs in the closed pig house. Indoor environmental quality, especially the thermal environment quality, is particularly important in the limited living space of the closed pig house, which has significant effect on pigs health, welfare and reproductivity. The indoor environment mainly includes thermal environment, harmful gas, dust, bacteria, light, etc. The thermal environment mainly refers to the indoor air temperature and humidity. The indoor air temperature is one of the most important environmental factors that directly affects the heat balance of pigs. Because pigs maintain a constant body temperature and carry out normal life activities through the balance of heat production and dissipation. So, indoor air temperature takes a critical role on keeping a constant pig body temperature and affect the health level and reproductive capacity of pigs. The humidity affects the evaporation and the body heat regulation of pigs. The high temperature and high humidity environment will seriously affect the pigs’ daily weight gain, at the same time, it will cause bacteria growth and disease. So, the indoor air temperature and humidity were payed much attention by many researchers in the past decades in order to maintain a suitable indoor environment for pigs. An optimized control strategy, an accuracy and timeliness environmental control was the first important task for pig house environmental control system. At present, the operation of environmental control devices in pig house mainly relies on data that collected by sensors. However, due to the data collected by sensors can only reflect the current indoor environmental conditions, it can not predict the trend of environmental variation in pig house, thus can not adjust the operation status of environmental control device in advance, to some extent, which leads to some time lag of environmental control system. Predictions of indoor environment is an effective way to provide a precision and optimal control strategy with forecasting for the indoor temperature and humidity variations to avoid some control lags. Combined with the actual historical temperature and humidity data and external influence data that collected by sensors, and based on the deep learning method, the long short-term memory (LSTM) prediction model was developed to achieve an accurate prediction and verification of temperature and humidity variation in pig house. The results showed that the predictions of temperature and humidity were consistent well with the observations whatever in winter or in summer. The maximum error of temperature was 1.9℃, and the mean error was 0.5℃; the maximum error of relative humidity was 135%, and the mean error was 2.3%; the mean determination coefficients R2 of temperature and humidity were 0.821 and 0.645, respectively. The established prediction model achieved a higher performance, which can provide a feasible reference for an optimal environmental control strategy and the reduction of time lag for environmental control in pig house.
