Prediction of pH value in industrialized aquaculture based on ensemble empirical mode decomposition and improved artificial bee colony algorithm

Abstract: The pH value in industrialized cultivation ponds is crucial to the survival of Litopenaeus Vannamei. Grasping the trend of the pH value timely and accurately is the key for the high-density healthy Litopenaeus Vannamei culture. Therefore, in order to solve the low prediction accuracy of the single model in pH value prediction, this paper proposes a pH value combination forecasting model in Litopenaeus Vannamei industrialized cultivation based on ensemble empirical mode decomposition (EEMD) and improved artificial bee colony (IABC) algorithm. In the modeling process, the non-linear time sequences of the original pH value are de-noised and decomposed into a series of stable and uncoupling sequences by the EEMD. Based on the changed characteristics of each sequence, the appropriate forecasting method is selected and a new independent prediction model is established. The independent prediction values are reconstructed to obtain the ultimate prediction results. But whether the weight of the combined forecasting model is appropriate restricts the prediction accuracy and performance seriously. Therefore, we choose the IABC optimized method to seek the optimal weight of the combined forecasting model, which overcomes the blindness and the impact of human factors in parameter selection of the combined forecasting model in order to accelerate its convergence rate and forecast accuracy. The combinations of the best weights are obtained automatically after the optimization, and in the process the nonlinear combination prediction model of pH value in industrialized cultivation is constructed. With this model, the pH value change has been predicted for industrialized cultivation pond from September 8 to September 15 in 2014 in Zhanjiang City, Guangdong Province. The experimental results show that the proposed combination prediction model of EEMD-IABC has better prediction effect than the optimized back propagation neural network based on simulated annealing algorithm (SA-BPNN) and genetic algorithm-least square support vector regression (GA-LSSVR) method. And the relative mean absolute percent error (MAPE), root mean square error (RMSE), mean absolute error (MAE) and determination coefficient (R2) for the pH values between the EEMD-IABC and GA-LSSVR models are 14.6%, 28.6%, 27% and 1.1% respectively under the same experimental conditions. The relative MAPE, RMSE, MAE and R2 for the pH values between the EEMD-IABC and SA-BPNN models are 56.8%, 61.4%, 62.8% and 6.16% respectively. It is obvious that the EEMD-IABC has high forecast accuracy and generalization ability. It can meet the actual demand for the pH value controlling in the Litopenaeus Vannamei industrialized cultivation and also provide a reference for water quality predictions in other fields.