基于 Attention与改进SCINet模型的无线传感器网络能量预测与分簇路由算法
- 分享到:
摘要：
针对能量收集无线传感器网络中,能量预测精度不佳、节点能量利用效率过低和网络难以持续运行等问题,提出了一种改进样本卷积交互神经网络(sample convolution and interaction network,SCINet)预测模型,并引人概率稀疏自注意力机制,在新特征序列的每个时间步上计算注意力权重,捕捉重要特征,提高模型预测精度,最后,根据节点剩余能量、预测未来可收集的太阳能能量,对分簇路由算法进行改进,仿真实验结果表明,该能量预测模型具备更高的预测精度和泛化能力,在能量预测模型的基础上,改进的分簇路由算法,能有效地延长无线传感器网络的生命周期.
To address the issues such as poor energy prediction accuracy,low energy utilization efficiency of nodes, and difficulty in sustaining operations in energy-harvesting wireless sensor networks, an improved sample convolutional and interaction net work(SCINet) prediction model is proposed. This model incorporates the Probabilistic Sparse Self-Attention mechanism, which calculates attention weights at each time step of the new feature sequence to capture important features and enhance model prediction accuracy. Finally, the clustering routing algorithm is improved based on the remaining energy of the nodes and the predicted solar energy that can be collected in the future, Simulation results demonstrate that this energy prediction model has higher prediction accuracy and generalization capability. Based on the energy prediction model, the improved clustering routing algorithm can effectively extend the lifespan of wireless sensor networks.
作者:
金崇强，徐震，王雪山
Jin Chongqiang，Xu Zhen，Wang Xueshan
机构地区：
武汉轻工大学电气与电子工程学院,武汉；河南中烟工业有限责任公司驻马店卷烟厂
引用本文：
金崇强,徐震,王雪山.基于 Attention 与改进 SCINet模型的无线传感器网络能量预测与分簇路由算法[J].河南师范大学学报(自然科学版),2025,53(5):52-59.(Jin Chongqiang,Xu Zhen,Wang Xueshan,Energy prediction and cluster routing algorithm for wireless sensor networks based on attention and improved SCINet modeling[J].Journal of Henan Normal University(Natural Science Edition),2025,53(5):52-59.D01:10.16366/j.cnki1000-2367.2024.04.20.0001.)
基金：
国家自然科学基金
关键词：
能量预测;样本卷积交互神经网络;概率稀疏自注意力机制;分簇路由算法
energy prediction; sample convolutional and interaction network; probabilistic sparse selfattention mechanism; clustering routing algorithm
分类号：
TP393
