Prediction algorithm solar irradiance based on ARIMA-TCN-LSTM-AM-CatBoost model in wireless sensor network

摘要：

在无线传感器网络中，节点的供能问题是制约其实际应用的重要因素，将太阳能转化成电能为节点供电是一种有效的解决方案.对可收集的太阳能进行预测，有助于提高无线传感器网络的使用寿命.针对太阳能的不稳定性和波动性，提出了一种混合模型来预测太阳能.该模型首先利用自回归差分移动平均模型(ARIMA)对序列中线性成分的敏感性来对太阳辐照度中的线性成分进行提取，保留非线性成分于残差序列中;融合时间卷积网络(TCN),长短期记忆神经网络(LSTM)和注意力机制(AM).提取光伏数据和由ARIMA生成的残差数据中更深层次的时间依赖关系和复杂模式;最后，利用Catbo0st决策树算法对预测结果进行集成和综合分析.实验结果表明，在太阳辐照度预测中，所提出的方案在精度、鲁棒性和泛化能力上相较于其他方法具有明显优势。

In wireless sensor networks, the energy supply problem of sensor nodes is an important constraint for their practical application, and converting solar energy into electricity to power nodes is an effective solution. The Prediction of the harvestable solar energy helps to improve the lifetime of the wireless sensor network. A combination model is proposed to predict solar energy based on its instability and volatility. The model first extracts the linear component of solar irradance using the auto-regressive differential moving average model (ARIMA), while retaining the nonlinear component in the residual series.Next, temporal conyolutional network(TCN). long and short-term memory neural network(LSTM), and attention mechanism(AM) are combined to further extract deeper temporal dependencies and complex patterns in the PV data and the residual data generated by ARIMA. Finally, the CatBoost decision tree algorithm is integrated to synthesize the prediction results. The experimental results show that in solar irradiance prediction, the scheme proposed in this paper has obvious advantages in terns of accuracy, robustness, and generalization ability compared with other methods.

作者：

龙晨，徐震，文士元

Long Chen, Xu Zhen, Wen Shiyuan

机构地区：

武汉轻工大学电气与电子工程学院；成都理工大学计算机科学与网络安全学院

引用本文：

龙晨，徐震,文士元。基于ARIMA-TCN-LSTM-AM-CatBoost的WSN太阳辆照度预测算法[J].河南师范大学学(自然科版),2026,54(3):76-84. (Long Chen, Xu Zhen, Wen Shiyuan.Prediction algorithm solar irradi-ance based on ARIMA-TCN-LSTM-AM-CatBoost model in wireless sensor network[J].Journal of Henan Normal University(Natural Science Edition) .2026.54(3):76-84.DOI: 10.16366/j.cnki.1000-2367.2024.12.18.0003.）

基金：

国家社会科学基金

关键词：

无线传感器网络;太阳辐照度;自回归差分移动平均模型;长短期记忆网络;CatBoost

wireless sensor network; solar radiation; auto-regressive differential moving average model; long short-term memory network; CatBoost

分类号：

TN919.2

基于ARIMA-TCN-LSTM-AM-CatBoost的WSN大阳辐照度预测算法.pdf