多传感器数据融合的风机状态预测算法设计与实验  

Design and Experiments of Prediction Algorithm for Wind Turbines State Based on Multi-Sensor Data Fusion

作　　者：寇志伟 靳乐乐 孔哲 齐咏生[1,2,3] 刘利强 KOU Zhi-wei;JIN Le-le;KONG Zhe;QI Yong-sheng;LIU Li-qiang(College of Electric Power,Inner Mongolia University of Technology,Hohhot,010051;Inner Mongolia Key Laboratory of Electromechanical Control,Hohhot,010051;Engineering Research Center of Large Energy Storage Technology,Ministry of Education,Hohhot,010051)

机构地区：[1]内蒙古工业大学电力学院,内蒙古呼和浩特010051 [2]内蒙古自治区机电控制重点实验室,内蒙古呼和浩特010051 [3]大规模储能技术教育部工程研究中心,内蒙古呼和浩特010051

出　　处：《制造业自动化》2025年第2期114-122,共9页Manufacturing Automation

基　　金：内蒙古自治区科技计划(2021GG0256)。

摘　　要：风机运行状态的准确预测与判断可以提前预警故障,维持风机的稳定运行,实现风电功率的合理调度,保障电力生产的稳定和安全。提出了一种多传感器数据融合的风机状态预测方法。首先,研究了风机多传感器数据的处理与特征提取方法,应用四分位法与Relief-F算法清洗了多传感器数据,并且依据数据权重选择了多传感器数据源。其次,设计了基于BP神经网络和基于D-S证据理论的信息融合算法,并在MATLAB中进行了实验验证,得到的风机状态预测准确率分别为80.35%及78.72%。再次,基于双层容错数据融合思想,改进了D-S证据理论方法,设计了基于FTDF-TCR的多传感器数据融合算法,并应用相同的样本数据集进行实验验证。最后,经实验验证,算法风机状态预测的准确率为89.36%,相较于原有算法分别提升了9.01%及10.64%,有效提高了风机运行状态预测的准确率。Accurate prediction and judgement of wind turbine operation status can provide the early warning of faults,maintain the stable operation of the wind turbine,realize the reasonable scheduling of wind power and guarantee the stability and the safety of power production.In this paper,a wind turbine state prediction method with multi-sensor data fusion is proposed.First,the processing and feature extraction methods of wind turbine multi-sensor data are studied,the multi-sensor data are cleaned by applying the quaternion method and Relief-F algorithm,and the multi-sensor data source is selected based on the data weights.Second,an information fusion algorithm based on BP Neural Network and D-S evidence theory is designed and verified in MATLAB,and the accuracy of wind turbine state prediction is 80.35% and 78.72% respectively.Next,based on the idea of two-layer fault-tolerant data fusion,the D-S evidence theory method is improved.The multi-sensor data fusion algorithm based on FTDF-TCR is designed and verified by applying the same sample dataset.Finally,as verified by experiments,the accuracy of wind turbine state prediction is 89.36%,an increase by 9.01% and 10.64% respectively,compared with that of the original algorithm,which shows that the accuracy of prediction has been effectively improved.

关 键 词：多传感器 数据融合 风机 D-S证据理论 FTDF-TCR 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]