基于专家知识与大语言模型混合增强的水电站告警智能诊断

孙国强; 史学恒; 罗哲清; 程礼临; 臧海祥; 卫志农

PDF(898 KB)

电力建设 ›› 0

基于专家知识与大语言模型混合增强的水电站告警智能诊断

孙国强¹, 史学恒¹, 罗哲清², 程礼临¹, 臧海祥¹, 卫志农¹

作者信息 +

Intelligent Diagnosis of Hydroelectric Power Station Alarm Based on Hybrid-augmentation of Expert Knowledge and Large Language Model

SUN Guoqiang¹, SHI Xueheng¹, LUO Zheqing², CHENG Lilin¹, ZANG Haixiang¹, WEI Zhinong¹

Author information +

文章历史 +

摘要

【目的】为改善传统的水电站监控告警事件诊断中样本不平衡度大且诊断效率较低,而基于人工智能的诊断方法缺乏可解释性且诊断准确度难以保证的问题,设计了一种基于大语言模型的水电站告警事件样本增强与故障诊断的深度学习框架,同时引入了基于正则表达的专家知识实现对告警诊断结果的校核,提高告警事件诊断结果的可信度。【方法】首先,针对水电站监控事件样本不均衡的问题,采用SimBERT模型对样本数较少的故障类型进行数据增强并将增强数据与原始数据相混合构成深度学习网络的输入样本;其次,基于电网监控告警规则构建典型故障的正则表达,实现基于正则表达式的典型故障准确判别;最后,基于ERNIE大语言模型构建水电站监控事件的词向量样本,并输入层级注意力网络中学习样本特征并输出分类结果,并与正则判断结果进行校核,得到最终的故障诊断结果。【结果】测试实例表明,相比于传统水电站监控告警事件诊断方法,本文模型能够提升诊断精度2%以上,模型训练时间减低30%且能够在0.1 s内实现时间诊断。【结论】所提的基于大语言模型的水电站告警事件样本增强与故障诊断的深度学习框架能够实现了系统故障时对故障事件类型的快速精准诊断,有利于保障电力系统的长期稳定安全运行。

Abstract

[Objective] In order to address the problem of large sample imbalance and low diagnostic efficiency in traditional hydropower station monitoring and alarm event diagnosis, as well as the lack of interpretability and difficulty in ensuring diagnostic accuracy in artificial intelligence based diagnostic methods, a deep learning framework based on large language model for hydropower station alarm event sample enhancement and fault diagnosis is designed. At the same time, regular expressions based on expert knowledge are introduced to verify the alarm diagnosis results and improve the credibility of the alarm event diagnosis results. [Methods] Firstly, considering the issue of imbalanced monitoring event samples in hydropower station, the SimBERT model is used to enhance the number of data for types with fewer samples. The enhanced data is then mixed with the original data to form the input samples of the deep learning network; Secondly, based on power grid monitoring alarm rules, constructing regular expressions for typical faults to achieve accurate discrimination of typical faults; Finally, based on the ERNIE large language model, word vector embeddings of hydropower station monitoring events are constructed, and are input into the hierarchical attention network to learn features of each event. The classification results are then produced. The final fault diagnosis results are then obtained by verifying with results produced by regular expressions. [Results] Test results show that compared with traditional monitoring alarm event diagnosis methods for hydropower stations, this model can improve diagnostic accuracy by more than 2%, reduce model training time by 30%, and achieve diagnosis time within 0.1 seconds. [Conclusions] The proposed deep learning framework based on large language model for enhancing alarm event samples and fault diagnosis of hydropower stations can achieve fast and accurate diagnosis of fault event types during system failures, which is conducive to ensuring the long-term stable and safe operation of the power system.

导出引用

孙国强, 史学恒, 罗哲清, 程礼临, 臧海祥, 卫志农. 基于专家知识与大语言模型混合增强的水电站告警智能诊断[J]. 电力建设. 0

SUN Guoqiang, SHI Xueheng, LUO Zheqing, CHENG Lilin, ZANG Haixiang, WEI Zhinong. Intelligent Diagnosis of Hydroelectric Power Station Alarm Based on Hybrid-augmentation of Expert Knowledge and Large Language Model[J]. Electric Power Construction. 0

中图分类号： TM734

参考文献

[1] 闪鑫, 戴则梅, 张哲, 等. 智能电网调度控制系统综合智能告警研究及应用[J]. 电力系统自动化, 2015, 39(1): 65-72.
SHAN Xin, DAI Zemei, ZHANG Zhe, et al.Research on and application of integrated smart alarm based on smart grid dispatching and control systems[J]. Automation of Electric Power Systems, 2015, 39(1): 65-72.
[2] 王臻, 刘东, 徐重酉, 等. 新型电力系统多源异构数据融合技术研究现状及展望[J]. 中国电力, 2023, 56(4): 1-15.
WANG Zhen, LIU Dong, XU Chongyou, et al.Status quo and prospect of multi-source heterogeneous data fusion technology for new power system[J]. Electric Power, 2023, 56(4): 1-15.
[3] 王静智, 赵磊, 邓方明, 等. 居民用户参与电网调峰激励智慧用能策略研究[J]. 电力科学与技术学报, 2024, 39(4): 121-127.
WANG Jingzhi, ZHAO Lei, DENG Fangming, et al.Research on smart energy consumption strategy of residents participating in peak load regulation[J]. Journal of Electric Power Science and Technology, 2024, 39(4): 121-127.
[4] 李卓, 王胤喆, 叶林, 等. 从感知-预测-优化综述图神经网络在电力系统中的应用[J]. 中国电力, 2024, 57(12): 2-16.
LI Zhuo, WANG Yinzhe, YE Lin, et al.The application of graph neural networks in power systems from perspective of perception-prediction-optimization[J]. Electric Power, 2024, 57(12): 2-16.
[5] 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.
WANG Cuiyang, JIANG Quanyuan, TANG Yajie, et al.Fault diagnosis of power dispatching based on alarm signal text mining[J]. Electric Power Automation Equipment, 2019, 39(4): 126-132.
[6] 刘梓权, 王慧芳, 曹靖, 等. 基于卷积神经网络的电力设备缺陷文本分类模型研究[J]. 电网技术, 2018, 42(2): 644-651.
LIU Ziquan, WANG Huifang, CAO Jing, et al.A classification model of power equipment defect texts based on convolutional neural network[J]. Power System Technology, 2018, 42(2): 644-651.
[7] BAI Z Y, SUN G Q, ZANG H X, et al.Identification technology of grid monitoring alarm event based on natural language processing and deep learning in China[J]. Energies, 2019, 12(17): 3258.
[8] 康佳宇, 张沈习, 张庆平, 等. 基于ANOVA和BO-SVM的变压器故障诊断方法[J]. 高电压技术, 2023, 49(5): 1882-1891.
KANG Jiayu, ZHANG Shenxi, ZHANG Qingping, et al.Fault diagnosis method of transformer based on ANOVA and BO-SVM[J]. High Voltage Engineering, 2023, 49(5): 1882-1891.
[9] 冯斌, 张又文, 唐昕, 等. 基于BiLSTM-Attention神经网络的电力设备缺陷文本挖掘[J]. 中国电机工程学报, 2020, 40(S1): 1-10.
FENG Bin, ZHANG Youwen, TANG Xin, et al.Text mining of power equipment defects based on BiLSTM-attention neural network[J]. Proceedings of the CSEE, 2020, 40(S1): 1-10.
[10] 戴志辉, 张富泽, 张近月, 等. 基于MacBERT-BiLSTM-CRF模型的继电保护装置缺陷知识图谱构建方法[J]. 电力系统保护与控制, 2024, 52(20): 131-143.
DAI Zhihui, ZHANG Fuze, ZHANG Jinyue, et al.Construction method of a defect knowledge map of a relay protection device based on a MacBERT-BiLSTM-CRF model[J]. Power System Protection and Control, 2024, 52(20): 131-143.
[11] 汪洋, 杨仕伟, 王宝华, 等. 基于深度置信网络的交直流配电网直流故障检测技术[J]. 电力工程技术, 2023, 42(1): 251-259.
WANG Yang, YANG Shiwei, WANG Baohua, et al.DC fault detection technology for AC/DC distribution network based on DBN[J]. Electric Power Engineering Technology, 2023, 42(1): 251-259.
[12] 赵妍, 孙延, 聂永辉. 基于格拉姆角差场和迁移残差网络的HVDC线路故障识别[J]. 电力建设, 2024, 45(8): 118-127.
ZHAO Yan, SUN Yan, NIE Yonghui.HVDC line fault identification based on the gram angle difference field and transfer residual network[J]. Electric Power Construction, 2024, 45(8): 118-127.
[13] DEVLIN J, CHANG M, LEE K, et al. BERT:Pre-training of deep bidirectional transformers for language understanding[EB/OL]. (2019-05-24)[2023-11-17]. https://arxiv.org/abs/1810.04805.
[14] LAN Z Z, CHEN M D, GOODMAN S, et al. ALBERT: a lite BERT for self-supervised learning of language representations[EB/OL].2019: 1909.11942. https://arxiv.org/abs/1909.11942v6.
[15] LIU Y H, OTT M, GOYAL N, et al. RoBERTa: a robustly optimized BERT pretraining approach[EB/OL].2019: 1907.11692. https://arxiv.org/abs/1907.11692v1.
[16] REIMERS N, GUREVYCH I.Sentence-BERT: sentence embeddings using Siamese BERT-networks[C]//Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). Stroudsburg, PA, USA: ACL, 2019: 3980-3990.
[17] SUN Z J, LI X Y, SUN X F, et al. ChineseBERT: Chinese pretraining enhanced by glyph and pinyin information[EB/OL].2021: 2106.16038. https://arxiv.org/abs/2106.16038v1.
[18] 王洪彬, 周念成, 黄睿灵, 等. 基于深度学习的110kV电网监控信号语义解析及态势感知模型[J]. 电力系统保护与控制, 2023, 51(2): 160-168.
WANG Hongbin, ZHOU Niancheng, HUANG Ruiling, et al.110 kV signal semantic analysis and situation awareness model based on deep learning theory for a power system monitoring system[J]. Power System Protection and Control, 2023, 51(2): 160-168.
[19] 贾骏, 杨强, 付慧, 等. 基于电力设备大数据的预训练语言模型构建和文本语义分析[J]. 中国电机工程学报, 2023, 43(3): 1027-1037.
JIA Jun, YANG Qiang, FU Hui, et al.Research on pre-training language model construction and text semantic analysis based on power equipment big data[J]. Proceedings of the CSEE, 2023, 43(3): 1027-1037.
[20] SUN Y, WANG S H, LI Y K, et al. ERNIE: enhanced representation through knowledge integration[EB/OL].2019: 1904.09223. https://arxiv.org/abs/1904.09223v1.
[21] 皮俊波, 齐世雄, 孙文多, 等. 基于UIE框架的电网故障处置预案实体和事件识别方法[J]. 中国电力, 2023, 56(12): 138-146.
PI Junbo, QI Shixiong, SUN Wenduo, et al.Entity and event recognition method for power grid fault handling plan based on UIE framework[J]. Electric Power, 2023, 56(12): 138-146.
[22] LI J, ZHANG D Z, WULAMU A.Chinese text classification based on ERNIE-RNN[C]//2021 2nd International Conference on Electronics, Communications and Information Technology (CECIT). IEEE, 2021: 368-372.
[23] 张金营, 王哲峰, 谢华, 等. 基于知识图谱与大语言模型的电力行业知识检索分析系统研发与应用[J]. 中国电力, 2024, 57(12): 198-205.
ZHANG Jinying, WANG Zhefeng, XIE Hua, et al.Development and application of a knowledge retrieval and analysis system for the power industry based on knowledge graph and large language model[J]. Electric Power, 2024, 57(12): 198-205.
[24] 李刚, 方鸿, 刘云鹏, 等. 新型电力系统中的大模型驱动技术: 现状、机遇与挑战[J]. 高电压技术, 2024, 50(7): 2864-2878.
LI Gang, FANG Hong, LIU Yunpeng, et al.Large-model drive technology in new power system: status, challenges and prospects[J]. High Voltage Engineering, 2024, 50(7): 2864-2878.
[25] ISSIFU A M, GANIZ M C.A simple data augmentation method to improve the performance of named entity recognition models in medical domain[C]//2021 6th International Conference on Computer Science and Engineering (UBMK). IEEE, 2021: 763-768.
[26] 唐子卓, 刘洋, 许立雄, 等. 基于负荷数据频域特征和LSTM网络的类别不平衡负荷典型用电模式提取方法[J]. 电力建设, 2020, 41(8): 17-24.
TANG Zizhuo, LIU Yang, XU Lixiong, et al.Imbalanced-load pattern extraction method based on frequency domain characteristics of load data and LSTM network[J]. Electric Power Construction, 2020, 41(8): 17-24.
[27] 游文霞, 梁皓, 杨楠, 等. 基于重采样和混合集成学习的不平衡窃电检测[J]. 电网技术, 2024, 48(2): 730-742.
YOU Wenxia, LIANG Hao, YANG Nan, et al.Class imbalanced electricity theft detection based on resampling and hybrid ensemble learning[J]. Power System Technology, 2024, 48(2): 730-742.
[28] 王绪亮, 顾媛丽, 张鸿儒, 等. 基于知识集成流形的电力设备缺陷文本数据增强方法与应用研究[J]. 电网技术, 2024, 48(4): 1690-1702.
WANG Xuliang, GU Yuanli, ZHANG Hongru, et al.Data augmentation and application of defect texts for power equipment based on knowledge integration manifold[J]. Power System Technology, 2024, 48(4): 1690-1702.
[29] 卫志农, 石东明, 张明, 等. 考虑样本类别不平衡的电网故障事件智能识别方法[J]. 电力自动化设备, 2021, 41(11): 93-99.
WEI Zhinong, SHI Dongming, ZHANG Ming, et al.Intelligent identification method of power grid fault events considering sample classification imbalance[J]. Electric Power Automation Equipment, 2021, 41(11): 93-99.
[30] WEI Z L, WANG C T, YANG X X, et al.Imbalanced sentiment classification of online reviews based on SimBERT[J]. Journal of Intelligent & Fuzzy Systems, 2023, 45(5): 8015-8025.
[31] YANG Z C, YANG D Y, DYER C, et al.Hierarchical attention networks for document classification[C]//Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. Stroudsburg, PA, USA: ACL, 2016: 1480-1489.
[32] VASWANI A, SHAZEER N, PARMAR N, et al.Attention is all you need[J]. Advances in Neural Information Processing Systems, 2017, 30:20-26.
[33] LAMPLE G, CONNEAU A. Cross-lingual language model pretraining[EB/OL].2019: 1901.07291. https://arxiv.org/abs/1901.07291v1.
[34] 孙国强, 章逸舟, 唐杰阳, 等. 基于数据增强和深度学习的水电站告警事件诊断[J]. 电力自动化设备, 2023, 43(8): 88-95.
SUN Guoqiang, ZHANG Yizhou, TANG Jieyang, et al.Diagnosis method of hydropower alarm events based on data augmentation and deep learning[J]. Electric Power Automation Equipment, 2023, 43(8): 88-95.