Research On Intelligent Fault Early Warning Method Of Boiler Equipment In Thermal Power Plant

With the rapid development of China's power industry,the application of boiler equipment in thermal power plants is becoming more and more widespread.As an indispensable and important power equipment in thermal power generation,various enterprises have put forward higher and higher safety and stability in the operation of boiler systems.Therefore,the boiler system structure is more complex and the operating parameters are more diversified.In order to reduce or even avoid the social and economic losses caused by the occurrence of faults,improve the safety and service life of boiler equipment,real-time monitoring and monitoring of equipment operating status data The timely and accurate early warning of various fault phenomena has extremely important theoretical significance and practical value.This paper mainly conducts research work on three levels of cleaning,data fuzzification and fault early warning methods of power plant boiler equipment online monitoring data,and studies intelligent fault early warning technology that can monitor the operation status of boiler equipment in real time.This paper proposes a data cleaning method based on stack noise reduction autoencoder.This method first introduces a hybrid optimization algorithm of Adam and SGD to continuously adjust the optimal network parameters of the stack noise reduction autoencoder model.The trained model can directly extract hidden features of normal state data to obtain reconstruction errors under normal state.The model is then used to analyze the effect of reconstruction errors under different types of abnormal state data on the model.Finally,"dirty data" and abnormal data reflecting equipment failures are quickly classified,cleaned and repaired.Through experimental analysis,this method can effectively and quickly classify and clean the boiler equipment online monitoring data,and has achieved good application results,verifying the effectiveness and practicability of this method.This paper presents a data fuzzification method based on I-KE-FCM algorithm.On the basis of data cleaning,this method first analyzes the correlation of multiple attribute data of boiler equipment;then it improves the idea of K-Means algorithm to select the initial class center,and combines it with information entropy theory to qualitatively Determine the best number of category partitions for all attribute data;finally assign the obtained category number and category center to the fuzzy c-means algorithm to divide the fuzzy interval,and express the data of each attribute in the database as the degree of membership at different levels.Through experimental analysis,it is verified that the method has significant efficiency in terms of time performance.This paper presents a method of early warning based on fuzzy association rule.On the basis of data fuzzification,this method first numbers the original database of boiler equipment according to the principle of maximum membership,and ranks each data in the level range where the maximum membership is.Then introduced the algorithm of fuzzy association rule based on transpose matrix,which can reduce the generation of candidate sets,and stipulated the selection principles of minimum fuzzy support and minimum fuzzy confidence degree.The database establishes the rule base under the normal operation data state.Finally,the fault data of the equipment is matched with the rule base,and the method is compared with the traditional fuzzy association rule algorithm.Through experimental analysis,the feasibility of this early warning method is confirmed.To sum up,the method proposed in this paper can quickly and accurately identify the fault state of boiler equipment,can effectively avoid the occurrence of safety accidents,can win more precious emergency repair time for operation and maintenance personnel,and ensure the safe and stable operation of power plant industrial production.