半月刊

ISSN 1000-1026

CN 32-1180/TP

+高级检索 English
采用GA-BPNN与TLS模型的风电机组异常辨识方法
作者:
作者单位:

浙江大学电气工程学院,浙江省杭州市 310000

作者简介:

通讯作者:

基金项目:

国家重点研发计划资助项目(2017YFB0902600);国家自然科学基金资助项目(51877190);国家电网公司科技项目(52110418000T)。


Anomaly Identification Method of Wind Turbine Based on Gene Algorithm - back Propagation Neural Network and t-location Scale Model
Author:
Affiliation:

College of Electrical Engineering, Zhejiang University, Hangzhou 310000, China

Fund Project:

This work is supported by National Key R&D Program of China (No.2017YFB0902600), National Natural Science Foundation of China (No. 51877190) and State Grid Corporation of China (No. 52110418000T).

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
    摘要:

    基于反向传播神经网络(BPNN)建立了风电机组状态参数预测模型,并采用遗传算法(GA)对BPNN模型的初始权重与阈值进行优化,有效消除环境因素对风电机组状态参数的影响;采用t-location scale(TLS)分布模型刻画不同风速区间下预测残差的分布特性,基于矩估计方法实现TLS分布参数估计,并在此基础上提出了计及风速影响的状态残差异常程度量化指标。以某风场的1.5 MW双馈风电机组为例进行了异常分析,结果验证了模型的有效性和准确性。

    Abstract:

    Based on back propagation neural network (BPNN), a wind turbine(WT) state parameter prediction model is established, and the initial weight and threshold of BPNN model are optimized by genetic algorithm(GA), which effectively eliminates the influence of environmental factors on WT state parameters. The t-location scale (TLS) distribution is used to characterize the distribution characteristics of state parameter prediction errors with different wind speed intervals. Moment estimation is used to calculate the TLS distribution parameters and error anomaly index (EAI) is defined to quantify the anomaly level of prediction errors, which is verified to be an indicator of the WT anomalies. The proposed method is used for 1.5 MW WT with doubly fed induction generators and the results show that the proposed method is effective and accuracy.

    表 4 3种分布模型对发电机前端轴承温度的拟合精度Table 4 Fitting accuracy of generator bearing temperature (front) with three different
    表 1 风电机组状态参数Table 1 Monitoring parameters of wind turbine
    表 6 各数据样本拟合误差的分布参数Table 6 Distribution parameters of fitting errors with different data samples
    表 10 风电机组异常辨识结果Table 10 Anomaly identification results of wind turbines
    表 2 样本量对BPNN预测过程的影响Table 2 Impact of sample data on BPNN training process
    表 7 各数据样本的TLS拟合精度Table 7 Fitting accuracies of TLS models for different data samples
    表 8 3种预测时间间隔下不同工况拟合误差分布参数Table 8 Distribution parameters fitting errors with three prediction time intervals in different operation conditions
    表 9 风电机组故障案例及目标状态参数Table 9 Fault cases of wind turbines and the corresponding target condition parameters
    表 3 齿轮箱输入轴温度样本数据提取方法Table 3 Method for selecting the sample data of gearbox input shaft temperature
    图1 GA-BPNN训练方法流程Fig.1 Flow chart of GA-BPNN training method
    图2 3种分布模型对发电机前端轴承温度预测误差的拟合曲线Fig.2 Fitting curves of generator bearing temperature (front) with three distribution models
    图3 发电机前端轴承温度预测误差概率分布Fig.3 Probability distribution of prediction errors of generator bearing temperature (front)
    图4 异常辨识流程Fig.4 Flow chart of anomaly identification
    图5 EAI时间序列序列Fig.5 The time series of EAI
    图 某风场1-10号风电机组风速分布盒须图Fig. Boxplot of the wind speeds of different WTs in the studied wind farm
    表 5 3种预测时间间隔下不同工况拟合误差分布参数Table 5 Distribution parameters of fitting errors with three prediction time intervals in different operation conditions
    图 风电机组的主要组件与传感器位置Fig. Main components and location of sensors of wind turbines
    图 风电机组状态参数数据样本划分Fig. Different intervals of the WT condition parameters
    图 BPNN与GA-BPNN模型预测精度对比Fig. Prediction accuracies of BPNN and GA-BPNN model
    图 采用近期样本与本文样本的模型预测精度对比Fig. Prediction accuracies of model trained by current data and data presented in this paper
    图 预测残差的区间划分Fig. The division method for prediction error
    参考文献
    相似文献
    引证文献
引用本文

李泽宇,郭创新,朱承治.采用GA-BPNN与TLS模型的风电机组异常辨识方法[J/OL].电力系统自动化,http://doi.org/10.7500/AEPS20190203002.
LI Zeyu,GUO Chuangxin,ZHU Chengzhi.Anomaly Identification Method of Wind Turbine Based on Gene Algorithm - back Propagation Neural Network and t-location Scale Model[J/OL].Automation of Electric Power Systems,http://doi.org/10.7500/AEPS20190203002.

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2019-02-03
  • 最后修改日期:2020-02-18
  • 录用日期:2019-10-06
  • 在线发布日期:
  • 出版日期: