面向WSN的高精度预测算法和分簇路由研究

Research on High-Precision Prediction Algorithm and Clustering Routing for WSN

摘要: 无线传感器网络(WSNs)中的节点通常由电池供电，能量是限制网络性能的重要因素。尽管部分节能路由和低功耗协议的提出在一定程度上缓解了此问题，但是无法从根源上解决。太阳能是一种可再生清洁能源，是解决网络能量受限问题的有效途径，但由于能源具有不稳定性，因此针对此类能量收集无线传感器网络(EH-WSN)进行充能预测和路由规划的研究具有重要意义，它是EH-WSN中低功耗技术的重要研究内容。针对充能预测，本文提出一种基于分类和递归的高精度预测算法k-LSTM，以实现更高效的能源管理，为WSN的路由规划提供有力支撑。此外，针对分簇路由中簇首节点选择不合理导致节点耗能加剧，网络性能不佳的问题，本文改进簇首适应度函数提出了一种基于粒子群的分簇算法，并对簇间路由下一跳的选择提出新的代价函数，进一步优化网络能源消耗。实验证明了所提方案对网络性能的提升优势，延长了网络寿命，并提高了网络吞吐量。

Abstract: Wireless Sensor Networks (WSNs) are usually powered by batteries, while energy is a critical factor limiting network performance. Although some energy-saving routing and low-power protocols have been developed to alleviate this problem to some extent, they cannot solve it fundamentally. Solar energy is a renewable and clean energy source that offers an effective solution to the problem of energy limitation in networks. However, due to the instability of energy, research on charging prediction and routing planning for energy-harvesting wireless sensor networks (EH-WSNs) is of significant importance and is a crucial research area in low-power technologies for EH-WSNs. For charging prediction, this paper proposes a high-precision prediction algorithm, k-LSTM, based on classification and recursion to achieve more efficient energy management and provide strong support for WSN routing planning. Moreover, to address the problem of energy consumption due to the unreasonable selection of cluster heads in cluster routing, which leads to poor network performance, we propose an improved cluster head fitness function and a PSO-based clustering algorithm. We also introduce a new cost function for selecting the next hop in inter-cluster routing to further optimize network energy consumption. Our experimental results demonstrate the advantages of the proposed solutions in improving network performance, significantly reducing the energy consumption of nodes, prolonging network lifetime, and increasing network throughput.