Research On The Assessment Of Real-time Reliability Of Thermal Power Plant

The thermal power unit is still the main power generation mode in China and plays an important role in power industry of China.The real-time reliability assessment and prediction of thermal power units is beneficial to the power plant to make corresponding countermeasures according to the real-time reliability of the unit,which improves the safety and economy of the unit operation and helps the grid to optimize the dispatching plan,avoiding accidents and improving the social benefits and economic benefits of the grid.The thermal power unit has a complicated structure and many equipments,and it is difficult to directly establish its real-time reliability assessment and prediction model.This paper uses graph analysis and LSTM methods to evaluate and predict the real-time reliability of thermal power units.First of all,for the thermal power unit,there are lots of equipments,and the equipment structure is complex,it is difficult to establish the failure model;the equipment has few or no prior knowledge of faults,and it is difficult to obtain the fault data.This paper proposes two real-time reliability assessment method for equipments:AFNIS-based Real-time reliability assessment method and similarity-based real-time reliability assessment method.ANFIS requires a small amount of fault data and equipment operating experience to establish an accurate real-time reliability assessment model for the equipment.The similarity-based method only needs to use the normal operation data of the equipment,and solves the real-time reliability evaluation model modeling problem without prior knowledge of equipment failure.Taking the medium speed coal mill as an example,the example analysis verifies the correctness of the modeling method and real-time reliability assessment method,and shows that the ANFIS-based and similarity-based real-time reliability assessment model can be used for thermal power plant equipment.The real-time reliability is evaluated effectively.Secondly,in order to predict the real-time reliability of the equipment,the LSTM deep neural network is introduced to establish the time series prediction model of the equipment variable to predict the running trend of the equipment.Combined with the real-time reliability evaluation method,we can predict the real-time reliability of the equipment.The LSTM neural network is used to establish the time series model of the coal mill,which avoids the problem that the equipment mechanism modeling is difficult and the prediction accuracy of the traditional time series method may not meet the requirements.The example analysis verifies the accuracy of the LSTM time series model for future trend prediction of the equipment,and shows that the real-time reliability prediction model based on LSTM can accurately predict the real-time reliability of thermal power unit equipment.Finally,because there are too many equipments and operational parameters in thermal power unit,it is difficult to directly establish real-time reliability assessment and prediction model of thermal power unit.This paper uses the graph analysis method to establish the relationship between the real-time reliability of the unit and the equipment.Thus through the real-time reliability ofequipments,we can acquire real-time reliability of the unit.A 1000MW ultra-supercritical unit was taken as an example to carry out simulation experiments.Experiments show that the graph-based analysis method can accurately establish the model between the real-time reliability of unit and these of equipment.Combined with the real-time reliability assessment and prediction method of the equipment,the real-time reliability of the unit can be effectively evaluated and accurately predicted.