基于预测的风机齿轮箱状态监测研究-水利水电工程专业论文.docx

万 万方数据 、 Abstract Title: A STUDY ON FORECAST BASED WIND TURBINE MONITORING Major: Water Resources and Hydropower Engineering Name: Jian ZHANG Signatu阳 ZNmCf J(àf) Supervisor: Prof. Yuxia LI Signature: 车 2ω Abstract As clean energy，wind energy is paid more and more attention by the world ，in this context， wind power has a rapid development around all the world，which is growing faster and faster proportion in the power system. Existing wind farms are generally through regular maintenance and troubleshooting a丘erwards for repairing and maintenance of wind fa口口s，which increases downtime of wind turbine，and makes wind farms a lower cost.Putting forw町d the beforehand fault monitoring and diagnosis could solve the above problems . This article is in connection with one of the most important components of wind turbine -gearbox，and proposes forecast based wind turbine monitoring model.Firstly，according to the problem that choosing wind turbine gearbox model input variables is not scientific，orthogonal least squares(OLS) method is proposed ，and make a comparative analysis with correlation coefficient and experience chosen method ，which confirms the effectiveness of the OLS when choosing wind turbine gearbox model input variables. On this basis，this article puts forward GA-BP Neural Networks and applying this method in wind turbine gearbox model，according to making a comparative analysis with traditional BP Neural Networks ，which proves that GA-BP has a lower eηors and stronger s阳bility. Lastly，this article combines GA-BP Neural Networks and statistical analysis of the residuals，and forming forecast based wind turbine temperature monitoring model，according to analyzing the operation mechanism ，running process and determining the important p缸ameters during the operation process ，and introduces the relationship between the model and SCADA system and the task of the model in the condition-bas