Research On Wind Turbine Fault Prediction And Diagnosis Technology Based On Key Data Mining

Wind energy is a kind of environmentally friendly,clean and endless renewable energy source.With the depletion of traditional energy sources such as oil and coal and political emphasis on environmental governance,wind energy surpasses photovoltaic power generation,bio-power generation,etc.in new energy power generation with the advantages of high efficiency,high cost performance,less space demand and wide application,showing its leader role in energy market.However,traditional wind turbines often trigger various faults without warning,which causes the wind turbines to be shut down for repair,in this case,it will seriously affect the running of the whole power generation process.Therefore,it is urgent to predict the faults based on the historical operation of the wind turbines and the key data of the actual measurements,as well as predict the parts and the time of the failure in order to arrange the maintenance plan and purchase certain equipment in advance.In this way,the power generation efficiency of the wind turbine can be improved,and the extensive management of the wind turbine is completely changed.Seen form the economic aspect and the security aspect,it also changed the passive position of practitioners into active one.This paper proposes the function of predicting the fault of wind turbines through the key data mining method.By employ this project,the wind farm can make maintenance plans ahead of time and minimize the losses caused by faults.After data mining of the historical records from SCADA,a project used to predict the fan trouble diagnosis is proposed,which is using MATLAB to establish a temperature prediction model that based on the oil temperature of the gears of the BP neural network and the temperature of the No.2 generator bearing.According to the comparison of the prediction curve and the real curve comparison,outliers can be found and used as early warning signals.What is more,a wind turbine fault monitoring system based on LABVIEW platform and a fault prediction system using LABVIEW graphical development tools also be built.The system can predict the faulty components and fault time of the wind turbine by calling the control algorithm to analyze the imported data.After that,it can display different levels of fault warning with an intuitive color warning light,which can remind the operator of the time and parts of the fault without complicated programming.This method can make a diagnosis be pre-judgable before the SCADA system sent the fault message.By harnessing this system,the on-site technicians can formulate effective measures according to the system prompts,to reduce maintenance costs and to avoid safety hazards.This scheme has been tested in a wind farm in eastern Yunnan.The test results shown that the fault prediction solution is reliable and it has certain promotion and use value.