基于SARIMA-VMD-LSSVM的水产养殖溶解氧质量浓度预测

Prediction of dissolved oxygen mass concentration in aquaculture based on SARIMA-VMD-LSSVM

为了充分利用溶解氧质量浓度的数据特征,进一步提高水产养殖中溶解氧质量浓度预测的准确性,提出“线性与非线性”与“分解-预测-集成”相结合的溶解氧质量浓度预测模型。该模型首先由季节性差分自回归滑动平均(SARIMA)模型对溶解氧质量浓度随着时间变化而组成的数据序列(简称溶解氧质量浓度的时间序列)进行线性拟合,使用变分模态分解(VMD)对残差序列进行分解,然后将各残差分量代入经改进的灰狼算法(IGWO)优化的最小二乘支持向量机模型(LSSVM)中,得到非线性分量的预测结果。最后集成线性与非线性预测结果,得到最终的溶解氧质量浓度预测值。结果表明,与SARIMA、LSSVM、VMD-LSSVM模型相比,基于SARIMA-VMD-LSSVM模型对溶解氧质量浓度进行预测的精度显著提高,预测的均方根误差(RMSE)为0.0787,平均相对误差(MAPE)为0.0226,说明该组合模型可有效提取溶解氧质量浓度的时间序列的多尺度特征,从而更精准地进行溶解氧质量浓度的预测。

In order to make full use of the data characteristics of dissolved oxygen mass concentration and further improve the accuracy of dissolved oxygen mass concentration prediction in aquaculture,a dissolved oxygen mass concentration prediction model combining"linear and nonlinear"and"decomposition-prediction-integration"was proposed.Firstly,the seasonal auto regressive integrated moving average(SARIMA)model was used to linearly fit the dissolved oxygen mass concentration time series,and the residual sequence was decomposed using variational mode decomposition(VMD).Then,each residual component was substituted into the least square support vector machine(LSSVM)model optimized by the improved gray wolf algorithm(IGWO)to obtain the prediction results of the nonlinear component.Finally,the linear and nonlinear prediction results were integrated to obtain the final dissolved oxygen mass concentration prediction value.Experimental results showed that compared with SARIMA,LSSVM,and VMD-LSSVM models,the prediction accuracy of SARIMA-VMD-LSSVM model was significantly improved.The root mean square error(RMSE)was 0.0787,and the mean absolute percentage error(MAPE)was 0.0226,indicating that the combined model could effectively extract the multi-scale features of the time series of dissolved oxygen mass concentration,and achieve more accurate prediction.