|
| 一种大语言模型支持的用电异常智能检测方法 |
| An intelligent detection method for electrical anomalies with large language models |
| 投稿时间：2025-01-01 修订日期：2025-04-02 |
| DOI： |
| 中文关键词: 用户用电量时间序列 异常检测 大语言模型 提示工程 |
| 英文关键词: Consumer electricity consumption time series Anomaly detection Large language modeling Prompt engineering |
| 基金项目:国家电网公司总部科技项目 |
|
| 摘要点击次数: 0 |
| 全文下载次数: 0 |
| 中文摘要: |
| 时间序列分析在工业领域极其重要。精准识别时间序列中的异常，能够有效提高生产效率，降低生产成本。针对电力使用，用户用电量时间序列的异常检测对于降低电力公司的运用成本、提升其服务质量具有重要的作用。随着大语言模型的快速发展，大语言模型被用在越来越多的领域，包括新兴的时间序列预测。本文提出一种基于大语言模型的用户用电时间序列异常检测方法，该方法通过提示工程与大语言模型交流，通过对时间序列进行预处理，使大语言模型能够理解输入数据信息，并正确执行异常检测任务。通过在用户用电时间序列数据集上进行实验分析，结果表明，与现有的异常检测模型相比，本文所提方法能够有效检测出异常数据。 |
| 英文摘要: |
| Time series analysis is extremely important in the industrial field. Accurately identifying anomalies in time series can effec-tively improve production efficiency and reduce production costs. For electricity usage, the anomaly detection of the time series of users' electricity consumption plays an important role in reducing the operating costs of power companies and im-proving their service quality. With the rapid development of large language models, they are being used in more and more fields, including the emerging time series prediction. This pa-per proposes a method for anomaly detection of users' electric-ity consumption time series based on large language models. This method communicates with large language models through prompt engineering. By preprocessing the time series, it enables large language models to understand the information of the input data and correctly execute the anomaly detection task. Through experimental analysis on the dataset of users' electricity consumption time series, the results show that com-pared with the existing anomaly detection models, the method proposed in this paper can effectively detect abnormal data. |
|
View Fulltext
查看/发表评论 下载PDF阅读器 |
| 关闭 |