Fluctuation of animal husbandry system (FAHS) can affect not only the sustainability of farmers' livelihood, but also the stability of regional ecosystem. Characteristics of FAHS and its driving factors in Yanchi County, Ningxia Hui Autonomous Region, which was a typical agro-pastoral zone, were studied and discussed in this paper. According to previous research results and our surveys, total number of sheep was defined as the dependent variable and 10 driving factors were defined as the independent variables, namely, average annual precipitation, average annual temperature, effective irrigated area, grain acreage, gross output of grain, afforestation area, grassland area, vegetation index, agricultural labor force and per capita net income of farmers. Then, the empirical mode decomposition (EMD), which is a powerful method for analyzing the nonlinear and non-stationary time series data, was used to analyze the variability of animal husbandry system and its driving factors. The time series data of total number of sheep and its 10 driving factors from 1954 to 2013 were decomposed into several intrinsic mode functions (IMFs) and a residual trend respectively. Meanwhile, the set pair analysis (SPA) method was used to analyze the contribution of each driving factor to FAHS. Here, 3 IMFs that were decomposed from total number of sheep and the corresponding IMFs that were decomposed from each driving factor were input in the set pair analysis model, and the relationship between FAHS and its driving factors was obtained by this method. The results showed: 1) The total number of sheep increased significantly with an annual fluctuation during the past 60 years and its coefficient of variation was moderate. The total number of sheep had 3 IMFs and its 10 driving factors had 2-5 IMFs. The most of IMFs had the fluctuant periods of 3-4, 7-15, 22-26 and 40-45 years respectively. The IMFs that had the fluctuant period of 22-26 years and the biggest variance contribution were the most important components that were decomposed from the total number of sheep. 2) The fluctuation of the total number of sheep mainly depended on grain acreage, followed by average annual precipitation, and the contribution rate of both driving factors accounted for 75%. The fluctuation of the number of sheep was also weakly affected by average annual temperature, gross output of grain and total output and afforestation area. 3) In these IMFs that were decomposed from total number of sheep, the one that had a longer fluctuation period was affected more seriously by agriculture. In contrast, the one that had a shorter fluctuation period was affected more seriously by climatic factors. Economic and social factors had limited effect on the fluctuations of the total number of sheep. 4) Climatic factors such as average annual precipitation and average annual temperature mainly drove the fluctuation of the total number of sheep. In contrast, grain total output and annual afforestation area had a reverse effect to the total number of sheep. Other 6 factors including effective irrigated area, grain acreage, grassland area, vegetation index, agricultural labor force and per capita net income of farmers had bi-directional effects on the total number of sheep. Based on above analyses, we suggest that the sustainable development for animal husbandry system in Yanchi County is constructing an effective agro-forestry-animal composite husbandry system in this typical agro-pastoral zone. In particular, developing appropriate scale and efficient farmland and feed shrub stands is very important to maintain the stability of animal husbandry system.