摘要
针对新能源监控系统中信号点关联作业依赖人工配置，工作强度大、效率低、易出错等实际问题，本文提出了一种融合深度语义理解与多维特征编码的数据智能关联方法。该方法首先设计了一种基于BERT模型的信号名称编码模型，并创新性地采用三元组网络架构进行对比学习微调，针对应用场景精准提取语义特征；设计融合机组编号、信号类型等多维特征的补充编码模型，并通过基于性能指标的权重分配策略实现模型有效融合，提升匹配精度。此外，方法引入“泛型”概念设计了可复用元模板，进一步提升工程效率。最后，选择国内三个实际风电场数据进行了实验和测试。结果表明，本算法使单台风机配置时间缩短73.4%，有效减少手动配置点数，显著提升工程效率，为新能源场站智能化配置提供有效解决方案。
关键词： 新能源监控；数据关联；三元组网络；BERT模型；自然语言处理
Abstract
To address the practical issues of manual configuration—such as high workload, low efficiency, and error susceptibility—in signal point association operations within renewable energy monitoring systems, this paper proposes an intelligent data association method that integrates deep semantic understanding and multi-dimensional feature encoding. First, a signal name encoding model based on the BERT model is designed, and a triplet network architecture is innovatively employed for contrastive learning and fine-tuning to accurately extract semantic features tailored to specific application scenarios. Additionally, a supplementary encoding model is developed to incorporate multi-dimensional features, such as turbine numbers and signal types, with a performance-based weight allocation strategy to enable effective model fusion and enhance matching accuracy. Furthermore, the method introduces the concept of “generics” to design reusable meta-templates, thereby further improving engineering efficiency. Finally, experiments and tests were conducted using data from three actual wind farms in China. The results demonstrate that this algorithm reduces the configuration time for a single wind turbine by 73.4%, effectively decreases the number of manually configured points, significantly enhances engineering efficiency, and provides an effective solution for the intelligent configuration of renewable energy stations.
Key words： Renewable energy monitoring; Data association; Triplet network; BERT model; Natural language processing
参考文献 References
[1] 王慧芳, 曹靖, 罗麟. 电力文本数据挖掘现状及挑战[J]. 浙江电力, 2019, 38(3): 1-7.
[2] 邱剑. 电力中文文本数据挖掘技术及其在可靠性中的应用研究[D]. 杭州: 浙江大学, 2016.
[3] 汪崔洋, 江全元, 唐雅洁, 等. 基于告警信号文本挖掘的电力调度故障诊断[J]. 电力自动化设备, 2019, 39(4): 126-132.
[4] 李明, 张伟. 基于BERT和CNN的电力设备缺陷文本分类方法[J]. 电力系统自动化, 2020, 44(10): 156-163.
[5] 贾宗泽. 基于知识图谱的电力知识问答与实现技术[D]. 北京: 华北电力大学, 2024.
[6] 刘沿娟,张栋栋,于海亮, 等. 基于知识图谱的电力标准智能问答系统研究[J]. 电工技术, 2024(16):143-146.
[7] Li, J., Wang, Y., & Sun, H. A Survey on Natural Language Processing for Industrial Systems[J]. Engineering Applications of Artificial Intelligence, 2024, 133: 108-125.
[8] Yujia Wu, Jun Wan. A survey of text classification based on pre-trained language model[J]. Neurocomputing, 2025, 616(1): 128921.
[9] Li, G., & Sun, C. Bag-of-Words vs. Graph vs. Sequence vs. Convolutional Models for Text Classification[J]. ACM Transactions on Intelligent Systems and Technology, 2023, 14(3): 1-25.
[10] DEVLIN J, CHANG M W, LEE K, et al. BERT: Pre-training of deep bidirectional transformers for language understanding[C]// Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2019: 4171-4186.
[11] CHEN X, LI S, ZHOU F. Intelligent Recognition of Power Customer Complaint Text Using a BERT-CRF Model[J]. Applied Energy, 2022, 318: 119212.
[12] MINAEE, S., KALCHBRENNER, N., CAMBRIA, E., et al. From TF-IDF to BERT: A Technical Overview of Text Representation Methods in NLP[J]. IEEE Transactions on Artificial Intelligence, 2022, 3(2): 127-145.
[13] HOFFER, E., & AILON, N. Deep metric learning using triplet network[C]// Proceedings of the 3rd International Conference on Learning Representations (ICLR). 2015.
[14] 吴桂龙, 杨志敏, 黄昱. 电力通信运行管理中典型业务数据的智能关联方法[J]. 电信科学, 2021, 2: 164-172.