Research On Turbine Generator Monitoring System Based On State Characteristics

Electric energy is an indispensable and important energy for economic life and social production,and turbo-generator provides 80% of the electric energy used by human beings.How to effectively improve the safety and reliability of the continuous operation of power equipment and ensure the continuous stability of power supply has become an important issue in the development of energy in the power industry and even in society.At present,most of the power equipment in our country adopts more traditional state monitoring means,that is,judging the state by checking the operating parameters of steam turbine units;a few large power plants have used condition monitoring system to monitor the power equipment,but the monitoring system used can’t make full use of all the obtained state information,so it is difficult for field technicians to quickly grasp the operating state of the equipment.In view of the above practical problems,this paper analyzes five typical faults of turbo-generator by combing the functional structure of turbo-generator.and the method of combining vibration signal analysis and machine learning state monitoring is adopted in the monitoring and identification of the system to provide a theoretical basis for the subsequent research.Through the analysis of the acquired vibration signal,the running state of the equipment is judged.After the vibration signal is preprocessed and its effect is evaluated,the feature extraction can be carried out.In order to realize the full and effective use of vibration signals,we use Frequency-domain integral processing,Time-domain statistical characteristic parameters,FFT and STFT to analyze them.the equipment state information is gradually extracted from many angles,and the state monitoring conclusion based on vibration analysis is obtained.By comparing the prediction effect of linear regression and LSTM neural network on some state parameters of generator,the LSTM neural network is selected as the state data analysis algorithm.According to the implementation method of machine learning,the construction flow of equipment health prediction model is designed,and the model parameters are determined in the process of implementation,and the health prediction model that meets the requirements is obtained.Based on the analysis data of generator state parameters,the results of data analysis based on machine learning state monitoring are obtained,and the comparison and verification of the conclusions of the two monitoring methods are completed.The vibration analysis and machine learning state monitoring methods are integrated into the monitoring system,combined with the system requirements,the overall architecture,function composition and interface planning of the system are designed and implemented,designed and implemented the basic function operation of the system,so that the field operator can quickly and conveniently obtain the equipment running state parameters,vibration signal analysis results,machine learning state monitoring conclusions.Various data processing results of the system can provide valuable information reference for daily maintenance and fault maintenance,so as to improve the reliability and stability of generator operation and reduce the probability of equipment downtime damage caused by failure.