摘要
受自然环境、计量仪器等影响,量测数据会出现异常,导致调度人员错误决策,威胁电力系统安全稳定运行。为保障电力系统安全稳定运行,提出了一种基于双向长短期记忆(bidirectional long short-term memory,Bi-LSTM)神经网络的配网电压无监督异常数据检测方法。利用Bi-LSTM神经网络处理时序数据的双向特性,建立时序预测模型,通过对比预测值和实际值的误差检测异常数据。最后,基于某实际配网电压数据进行仿真验证,仿真结果表明:所提方法在准确率、F1分数等指标方面均优于决策树、K近邻、支持向量机、长短期记忆(long short-term memory,LSTM)神经网络。
Due to the influence of natural environment,metering instruments,etc.,abnormal measurement data lead to dispatchers’wrong decisions,which threatens the safe and stable operation of the power systems.To ensure the safe and stable operation of power systems,an unsupervised abnormal voltage data detection method was proposed for distribution networks based on a bidirectional long short-term memory(Bi-LSTM)neural network.Considering the bidirectional characteristics of Bi-LSTM neural network in dealing with time series data,a time series prediction model was constructed to detect abnormal data by comparing the error values between the predicted and the actual values.Finally,based on the voltage data of actual distribution network,the simulation verification was ca-rried out.The simulation results show that the advantages of the proposed method in accuracy,F1 score and other indicators over the decision tree,K-nearest neighbor,support vector machine,and long short-term memory(LSTM)neural network.
作者
况华
何鑫
何觅
覃日升
姜訸
KUANG Hua;HE Xin;HE Mi;QIN Ri-sheng;JIANG He(Yunnan Power Grid Co., Ltd., Kunming 650011, China;Yunnan Power Grid Co., Ltd. Electric Power Research Institute, Kunming 650217, China;Yunnan Power Grid Co., Ltd. Kunming Power Supply Bureau, Kunming 650011, China)
出处
《科学技术与工程》
北大核心
2021年第24期10291-10297,共7页
Science Technology and Engineering
基金
云南电网科技项目(056200KK52190079)。
关键词
异常数据检测
配网电压
双向长短期记忆(Bi-LSTM)神经网络
时序
abnormal data detection
distribution network voltage
bidirectional long short-term memory(Bi-LSTM)neural network
time series
