通信与信息技术Issue(5)：90-94,137,6.
基于混合CNN-LSTM注意力机制基站电费稽核方法
Audit method of base station electricity cost using a CNN-LSTM-attention hybrid model
摘要
Abstract
In the field of mobile communications,with the mature construction and operation of 5G networks,electricity costs ac-count for a significant proportion of telecom operators'maintenance expenses.To strengthen risk management and improve the efficiency of electricity cost auditing,this paper proposes a audit method of Base Station Electricity Cost using a CNN-LSTM-attention hybrid mod-el.This approach integrates Convolutional Neural Networks(CNN),Long Short-Term Memory networks(LSTM)and Attention Mecha-nisms to address electricity cost auditing for base stations.In this method,a preprocessed data of 12 months electricity cost and corre-sponding room equipments power consumption are first input into the network model through 15 channels.Then,the CNN is utilized to learn correlations between multi-year electricity cost and power consumption data across different channels.Subsequently,the LSTM cap-tures temporal sequential features of the signals,while the attention mechanism enhances the contribution of key months to prediction ac-curacy.Finally,an auditing framework is established to effectively verify actual electricity cost reimbursement data reported by mainte-nance personnel.The proposed hybrid framework not only accounts for inter-channel correlations but also filters irrelevant data points,thereby improving the precision of multi-base station electricity cost auditing.Experimental results demonstrate that the CNN-LSTM model with the hybrid attention mechanism achieves superior base station electricity cost prediction performance.Compared to existing methods,it delivers higher coefficients of determination(R2),lower root mean square error(RMSE),and reduced mean absolute error(MAE),with promising practical application outcomes.关键词
电费稽核/LSTM/CNN/注意力机制/评价指标Key words
Audit electricity costs/LSTM/CNN/Attention Mechanisms/Evaluation method分类
信息技术与安全科学引用本文复制引用
王云花,邓翠艳,姚旭清..基于混合CNN-LSTM注意力机制基站电费稽核方法[J].通信与信息技术,2025,(5):90-94,137,6.基金项目
2023年山西省高等学校科技创新项目"基于大数据的客群精准筛选与营销的研究"(项目编号:2023L519) （项目编号:2023L519）
