Research On Early Warning Method Of Power Station Equipment State Based On Data Mining

Power plant equipment is the core of power production,which is of great significance to the normal and safe operation of thermal power plants.In actual production,the operating environment of power plant equipment is poor,which leads to frequent failures,brings huge economic losses to thermal power plants,and even leads to major production safety accidents.It is of great significance for the safety and economy of power production in thermal power plants to monitor the operation status of power plant equipment on-line and capture the early development process of faults.In this paper,the induced draft fan is selected as a typical power plant equipment,and the induced draft fan status early warning model based on multivariate state estimation technology is established.In the early stage of equipment deterioration,abnormalities are found and early warning is given,and the operation and maintenance personnel are guided to take corresponding measures to avoid shutdown hazards.The method is extended to coal mills for verification,which proves that the method is universal.The construction method of historical memory matrix based on cluster analysis to distinguish typical working conditions is proposed,which improves the prediction accuracy of the model.Firstly,this paper introduces the mechanical structure and common faults of the induced draft fan,and analyzes the abnormal monitoring parameters of different fault types.On this basis,according to the principles of "fault sensitivity","easy access" and "simplest",the multivariate State Estimation Modeling variables of the induced draft fan are preliminarily selected,and the simplest modeling variables are obtained through correlation analysis.A process memory matrix construction method based on cluster analysis to distinguish typical working conditions is proposed.By using hierarchical sampling,the redundancy and unevenness of working conditions are avoided,and the multivariate state estimation model of the induced draft fan health state is established.Finally,the simulation experiment using the induced draft fan health state history data proves that the established state prediction model has high accuracy and can meet the application requirements of state early warning technology.Define the similarity rules,and use the analytic hierarchy process and sliding window statistical method to accurately capture the subtle anomalies of equipment measuring points.The difference between the new observation vector and the model prediction vector of the induced draft fan reflects the state information of the induced draft fan.In order to more easily capture the early development process of the fault,the similarity rules of the observation vector and the prediction vector are defined to quantitatively judge the difference between them.In order to distinguish the contribution rate of different variables to the information of equipment deterioration process,the weight of each variable in the similarity function is determined by analytic hierarchy process.Then,the sliding window statistical method is used to eliminate the influence of random interference,and the reasonable state early warning threshold is determined according to the minimum value of average similarity under critical normal state,so as to establish the induced draft fan state early warning method based on multivariate state estimation.Finally,the state degradation process of the induced draft fan is simulated by artificially accumulating and increasing the deviation.The simulation results show that the method can detect the state degradation process of the induced draft fan and achieve accurate state early warning.