Aiming at the problem of high difficulty and low prediction accuracy during dissolved oxygen prediction modeling due to the nonlinear and unstable characteristics of dissolved oxygen time series in intensive aquaculture,a complete ensemble empirical mode de-composition with adaptive noise-clustering reconstitution and partial least squares optimized extreme learning machine model(CEEM-DAN-CPELM)is proposed.The CEEMDAN method is used to decompose the dissolved oxygen data stream into mode functions of differ-ent frequencies,and the complexity of each component is evaluated based on its fuzzy entropy value.The K-medoids method is used to cluster all functions according to their fuzzy entropy complexity,achieving the data decomposition reconstruction process and reducing the difficulty of dissolved oxygen prediction.Then,partial least squares algorithm is used to optimize the extreme learning machine and improve the predictive performance of the prediction model.Finally,the CEEMDAN-CPELM model is applied in aquaculture production at Changshu aquaculture experimental base.The experimental results show that the root mean square error value of the proposed CEEM-DAN-CPELM model is 0.959,which is significantly lower than the comparison models such as GA-SELM,LSSVM,and ELM,verifying the feasibility and effectiveness of the prediction model.
关键词
传感网络/溶解氧/预测/自适应完备集合经验模态分解/偏最小二乘法
Key words
sensor network/dissolved oxygen/prediction/complete ensemble empirical mode decomposition with adaptive noise/partial least squares
维普
万方数据