It is important to find out the distribution pattern of particulate matter in livestock houses to reduce dust particle pollution and improve air quality in livestock houses. The concentrations of different particle sizes (PM1, PM2.5, PM10 and TSP) in a layer house in Jilin Province were monitored and the distribution characteristics of different particle sizes were analyzed. The temperature, humidity, airflow and particulate matter concentration fields of a closed layer house with negative pressure ventilation were simulated by using computational fluid dynamics (CFD) technology and the design was verified through experiments, and the unreasonable areas were optimized for simulation. The results showed that temperature was positively correlated with PM1 and PM2.5（P<0.01); humidity was negatively correlated with particulate matter; TSP concentration was mainly affected by chicken activity. The simulated temperature values were generally large relative to the measured values, with a relative error range of 1.23% to 8.88%; while the simulated relative humidity values were generally small relative to the measured values, with a relative error range of 5.11% to 14.00%. The simulated values of particulate matter concentration were small compared with the measured values, which verified the accuracy of the model. By optimizing the angle between the deflector and the wall, the angle was increased to 67.5°. After optimization, the airflow uniformity in the house was obviously enhanced, and the airflow distribution was more reasonable, and the concentration of PM1 in the house was decreased by 17.4%, and that of PM2.5 was decreased by 15.9%, PM10 was decreased by 18.1% and TSP was decreased by 21.6%, which was more conducive to the growth and development of chickens and better air quality improvement of the layer house. The research result revealed the distribution patterns of temperature, humidity, airflow and particle concentration fields in northern layer houses in winter, which can provide a reference basis for environment optimization in layer houses.