Research On Adaptive Fault Early Warning Method Of Power Generation Equipment Under Variable Conditions

With the development of flexible transformation of thermal power units in China,more and more thermal power units participate in peak regulation operation.Most thermal power units were originally designed as base-load units,and participation in peak regulation operation will inevitably involve frequent start-up and large range of load changes,leading to equipment damage and acceleration.There are many generating equipment in thermal power units.Once a major fault occurs,it will not only lead to unplanned shutdown,affect the safety and economy,but also destroy the stability of power grid operation.This paper mainly studies the adaptive fault warning method of power generation equipment under variable working conditions.On the one hand,it can adapt to the variation of unit working conditions and reduce false positives.On the other hand,compared with the traditional alarm method,it can issue fault warning in advance.Firstly,the structure,working principle,common faults and fault symptoms of coal mill are analyzed.On this basis,aiming at the selection of modeling variables of power generation equipment,four principles of selecting modeling variables are proposed.According to the principle and expert knowledge,five variables,including the current of coal mill,the outlet temperature of coal mill,the loading oil pressure of the roller,the primary air flow rate and the differential pressure at the outlet of coal mill,were selected as the modeling variables of coal mill state.Then,in view of the complex operating environment of thermal power plants resulting in a large number of noise interference in the collected data,an improved wavelet analysis method is proposed to preprocess the data.In order to lay a foundation for the subsequent establishment of sub-models to more accurately characterize the status of power generation equipment,the fuzzy soft clustering(SFCM)algorithm was proposed to divide the data conditions.Through the example analysis of a coal mill,it is found that the improved wavelet analysis algorithm can effectively remove the noise in the data.The data of each working condition divided by SFCM algorithm contains the data of the transition part between its close working conditions,which lays a good foundation for improving the accuracy of the equipment model.Then,introduce methods of building process memory matrix based on factor analysis and isometry sampling.Contrasting the effect of different methods for building process memory matrix using coal mill data,it is found that model which results from each dimension of equidistant sampling gets the highest accuracy,at the same time size of process memory matrix is relatively small,and could effectively improve the speed of computing.The validity of the coal mill state model is verified through the analysis of a coal mill example.Finally,in view of running state of power equipment under variable working condition prone to entailing false positives,first define a similarity function to quantitatively characterize the differences between the observed and estimated values,and then put forward a kind of adaptive threshold value method which can adapt to the working conditions change.The analysis of coal mill operation mode data shows that the adaptive threshold value could change with similarity updating,and effectively reduce false positives.This paper presents an adaptive fault warning method for power generation equipment under variable working conditions.Through examples of analysis of coal mills,it can be seen that the adaptive threshold can be reasonably determined according to false positives,missed positives and warning time.The results show that the proposed method can detect the early fault of equipment and give accurate warning.Compared with the traditional warning method,the proposed method could trigger warning earlier.