Since the sequence data of water temperature in industrialized Litopenaeus vannamei breeding is somehow non-linear, unsteady, and such problems like lower precision in the predicting results, low robustness will appear when utilizing the traditional single item predicting method, a combined non-linear prediction model based on empirical mode decomposition (EMD), phase space reconstruction and extreme learning machine (ELM) was proposed. In modeling, the EMD method was adopted to decompose the original time sequence data of water temperature in industrialized Litopenaeus vannamei breeding into a series of intrinsic mode function (IMF), and reconstruct the phase space of each component, set models of ELM training in phase space, predict each IMF sequence, and then combine and reconstruct the predicted values of each component to get the predicted value of original water temperature sequence. EMD—ELM was tested and compared with other algorithms by applying it to predict water temperature in industrialized Litopenaeus vannamei breeding pond of Zhanjiang City. The experimental results showed that the proposed combination prediction model of EMD—ELM had better prediction effect than the standard extreme learning machine (ELM), least squares support vector regression (LSSVR) and BP neural network methods. And the relative mean absolute percentage error (MAPE), root mean square error (RMSE) and mean absolute error (MAE) between the EMD—ELM and standard LSSVR models were 62.82%、45.62% and 42.77%, respectively, under the same experimental conditions. The relative MAPE, RMSE and MAE between the EMD—ELM and standard ELM models were 34.44%、28.94% and 25.37%, respectively. The relative MAPE, RMSE and MAE between the EMD—ELM and BPNN models were 77.0%、60.83% and 54.77%, respectively. It was obvious that the EMD—ELM had high forecast accuracy and generalization ability. The research results provided a new effective technical support for water temperature management and control in the industrialized cultivation of Litopenaeus vannamei.