Data fusion based wireless temperature monitoring system applied to intelligent greenhouse

Data Fusion Based on Temperature Monitoring of Aquaculture Ponds With Wireless Sensor Networks

“…Many parts of research, including filtering technology and neural networks, have been merged with data fusion technology. For example, to improve the performance of an intelligent greenhouse temperature monitoring system, Xia et al [25] proposed a hierarchical WSN real-time fusion strategy, which uses neural networks for global fusion to ensure the real-time efficiency and accuracy of data fusion. The proposed parallel inverse covariance intersection (PICI) algorithm fuses 600 sets of data in only 1.45 s. Compared with 1.972 s of sequential inverse covariance intersection (SICI) and 5.766 s of batch inverse covariance intersection (BICI), the speed is increased by 1.36× and 3.97×, respectively.…”

“…Aquaculture areas are mostly in harsh environments with more external disturbances and strong noise interference, so it is proposed to adopt the MP probability weighting to improve the UKF. For the traditional UKF algorithm, see [25].…”

“…Yang et al [27] introduced the fading memory matrix into UKF to improve the identification of nonlinear modeling errors, and Xia et al [25] replaced the original Cholesky decomposition with singular value decomposition (SVD) while introducing the fading memory matrix, which effectively avoided the collapse of UKF, and Xia et al [25] and Yang et al [27] both computed errors for validation and achieved better results, but the aquaculture region has strong noise interference, the measured data deviation is large, and the quality of the underlying data directly affects the subsequent fusion effect. This article introduces the MP estimate to effectively suppress the field noise and establish the MP estimate of the noise statistics, used to adjust the system noise statistics, effectively avoiding the noise too strong to cause the UKF calculation dispersion, while combined with [49], it can be known LU matrix decomposition that has the advantages of robustness, high accuracy, decomposition speed, and simple operation.…”

“…Shi et al [31] proposed the PDCI method based on the CI algorithm, which can effectively predict nonstationary time series, and according to [51], it is known that ICI not only inherits the advantages of CI with less computation but also effectively solves the problem of CI fusion method being too conservative, which makes the fusion accuracy higher, so this article ICI is used as the base algorithm for local fusion. The PICI algorithm proposed in [25] focuses on solving the local fusion order problem with a parallel fusion idea, work [27] is the traditional SICI fusion algorithm with a sequential fusion idea, and work [32] is the traditional BICI algorithm with one-time fusion when the data all reach the fusion center. The BICI algorithm differs from the PICI algorithm and SICI algorithm only in the fusion structure, and all three are based on the ICI algorithm.…”

“…Shi et al [59] used ELM for dissolved oxygen prediction and achieved better results, but ELM generates random input-layer connection weight and hidden-layer neuron offset during training, which leads to bias in the final results. Xia et al [25] used an improved sparrow algorithm to optimize ELM, which is a hybrid model that effectively solves the problem of slow global fusion; ELM has faster computational speed and computational accuracy than BP, and the introduction of LF makes it easy for the sparrow algorithm to jump out of the local optimum and accelerate the convergence speed. In this article, ELM is adopted as the neural network algorithm, SOA is selected as the optimization algorithm, the LF mechanism is introduced to speed up the search speed and avoid falling into local optimum, the cosine factor is introduced in the attack process to adapt the nonlinear search process of the seagull algorithm, and the sine cosine operator is introduced to balance the local search and global search ability of the algorithm during the iteration of the optimal position of the seagull.…”