| 基于人工神经网络的海洋锚系浮标表层水温序列异常检测研究 |
| Anomaly detection of surface water temperature time series by ocean mooring buoy based on artificial neural network |
| 投稿时间：2018-08-23 修订日期：2018-09-01 |
| DOI： |
| 中文关键词: 海洋观测 异常检测 人工神经网络 锚系浮标 |
| 英文关键词:ocean observing anomaly detection artificial neural network mooring buoy |
| 基金项目:国家自然科学基金项目 |
|
| 摘要点击次数: 227 |
| 全文下载次数: 0 |
| 中文摘要: |
| 锚系浮标是业务化海洋观测系统的代表性设备，通常在海洋恶劣环境下运行，数据序列易受到影响而发生异常。本文以人工神经网络模型预测区间为判定阈，对山东褚岛锚系浮标表层水温序列进行了异常检测案例研究。结果表明，本方法检测自然环境因素带来的表层水温数据序列有效，未出现假阴性（漏报）或假阳性（误报）。对电源、通信等间接设备故障带来的异常有一定延迟，但能够识别出设备故障带来的所有极大异常和少部分非极值异常，异常检测率约为97.7%。对锚系浮标观测序列开展异常检测研究并分析设备故障特征，对保障海洋锚系浮标的长期稳定运行具有重要实际意义。 |
| 英文摘要: |
| Ocean mooring buoy (OMB) is a representative equipment of ocean observing system. It usually operates in harsh ocean environment that the observing sequence is easily affected. In this paper, an anomaly detection case study of surface water temperature series obtained by OMB in Chu Island in Shandong Province is carried out with the prediction intervals of artificial neural network model as the decision threshold. The results show that the method is effective in detecting the surface water temperature data sequence caused by natural environmental factors, and no false negative or false positive are detected. There is a certain delay in the power equipment, communications and other indirect equipment failures. But all the maximum anomalies and a few non-extreme anomalies caused by equipment failure can be identified. The detection rate of anomalies is about 97.7%. Anomaly detection and analysis of equipment fault characteristics for the observation sequence of OMB are of great practical significance to ensure the long-term stable operation of OMB. |
| 查看/发表评论 下载PDF阅读器 |
