Research And Practice On Partition Of Large Francis Hydropower Unit Under Full Operating Range

Francis turbine unit has complex structure.The factors causing vibration include hydraulic,mechanical,electrical and other aspects.Various factors are coupled with each other.The operating conditions of the unit are changeable.When the unit runs in the non-optimal design operating conditions,the vibration will be more severe.Longterm operation in this area will endanger the safety and stability of the unit,damage the life of the unit and reduce economic benefits.At present,dividing the unit operation area and avoiding bad working conditions have become the common countermeasures in the industry.Taking Unit 1 of T Hydropower Station as an object,this paper first realizes the division of unit operation area based on classical method.Then the shortcomings of the classical methods are summarized and the following research is carried out:1)The classical method is based on unit stability test,which is difficult to cover all operating conditions.The result of partition is rough.In this paper,on-line monitoring state data are introduced.Combined with test data,the refinement and the coverage of all working conditions of partition data are realized.The operating areas of all working conditions are divided by using full operating range data and peak-peak value standard thresholds.2)The classical method is based on the signal eigenvalue of the time-domain peakpeak value.This algorithm ignores the frequency-domain information and cannot fully reflect the signal vibration intensity characteristics.In this paper,the ability of EEMD energy entropy to extract time-frequency characteristics of vibration signals is verified by rotor test-bed.Then the statistical distribution law of EEMD energy entropy of testbed signals is analyzed,and the threshold of distinguishing normal and abnormal states is set.3)The operation limits of some measuring parameters in national and industrial standards are not clear,and the setting methods of the limits do not take into account the individuality of units and the parameters.Based on the measured vibration signals of units,the EEMD energy entropy distribution law of each monitoring parameter of historical monitoring data in a maintenance period is analyzed.Referring to the set value of confidence probability in the national standard,the relative threshold of EEMD energy entropy of each state parameter is set up,and the operation area of units under all operating conditions is divided accordingly.The results show that,compared with the classical method,the full-workingcondition method based on peak-peak standard threshold can reflect the real vibration zone more comprehensively and accurately.Compared with the peak-peak value,EEMD energy entropy can effectively distinguish the abnormal and normal state of the unit,and also has a certain ability to distinguish different kinds of faults.This method is suitable for signal feature extraction in the partition of hydropower units.The method based on EEMD energy entropy relative threshold can be used as a supplement and improvement of the classical partition method.The partition results can reflect the working region of the unit with larger vibration relative to the historical operation state.