Research On Reliability Evaluation Technology Of Thermal Power Unit Based On Fault Early Warning

Thermal power generation is a special production systems with a variety of equipment.In order to meet the user’s actual demand for electric energy,in the long-term operation of the unit,it is necessary to constantly adjust the operation mode of the unit at the same time.Under the influence of such frequently-scheduled production modes,the correlation between various primary and secondary equipment leads to a continual increase in potential faults and a consequent decrease in reliability of the generating units.This paper mainly studies the reliability assessment of thermal power units based on early fault warning.This method can transmit early warning signals in the early equipment failure,and provide enough time to trace the fault point and determine the type of fault.The unit short-term reliability assessment can be realized according to the abnormal parameters.Degradation trend of unit reliability can provide accurate information reference for operators.In order to establish a more accurate fan failure warning model based on MSET,a method of constructing memory matrix was proposed based on probability density sampling of 2-norm.And the new method was compared with the traditional method.Taking a induced draft fan in power plant as an example,the operating parameters with strong correlation were selected one by one.The weight coefficient of each monitoring parameter to fan early fault warning was calculated by analytic hierarchy process(AHP).In order to more clearly characterize the correlation between the observed vector and the estimated vector,the weight coefficients of each variable were applied to the common similarity function of the project.The validity of the MSET estimation model was validated for the actual operation data.The early warning threshold was determined according to the averagc similarity between the observed value and the estimated value of the verification set processed by the sliding window method.Aimulation experiment was conducted based on abnormal data to realize fault warning simulation.Finally,the the possible type of fault was traced and determined by relative error.A short-term reliability weighted assessment model with multiple monitoring parameters was proposed based on the fault warning.When the reliability of the unit in the future was estimated based on the predicted value,the data in the five time periods after the failure warning signal was issued and before the shutdown were selected as the initial sequence respectively based on the unchanged deterioration range of abnormal parameters.The gray prediction method based on dynamically optimal weights was used to give a short-term forecast of anomalous monitoring parameters.The average reliability of unit was computed according to the parameter predictive value of each interval,and during the period before the induced draft fan stopped down,the unit reliability showed a gradual downward trend.In the case of known unit output in the current period of time to assess the reliability of the unit,the K-means clustering method was used to divide the load of the unit for a period of time before the induced draft fan stops,and the range of deterioration of abnormal parameters under different load ranges were determined.For the abnormal parameter monitoring value inputed,the unit reliability was calculated according to the deterioration range of the abnormal parameter corresponding to the current unit load.The average reliability of units under different load ranges was calculated.This gave a comprehensive assessment of the short-term reliability of the unit.