Study On Soft Sensor Method For Furnace Temperature Of Thermal Power Plant Based On Support Vector Machine

The accurate measurement of furnace temperature is the key factor to ensure the stable operation of the thermal power unit control system and improve the efficiency of the boiler,but temperature measurement devices in high temperature, high pressure,strong airflow erosion, strong corrosive environment, are generally difficult to carry out long-cycle work.In case of failure of temperature measurement devices, real-time monitoring of thermal power of the furnace temperature has been a bottleneck problems,and the boiler system has many characteristics such as many process variables, nonlinear and time varying. Aiming at these problems,in order to accurately describe the characteristics of the furnace temperature process,the paper establishes multi-output furnace temperature intelligent soft sensor model support vector machine.Based on the sufficient and research of some classical predicition methods and the current hot forecasting methods, multi-output furnace temperature prediction model based on least squares support vector machines has been established.According to the mechanism analysis and operation condition of power plant boiler, the auxiliary variables are determined, and the selection of auxiliary variables is extracted by kernel principal component analysis. Then the sample data is optimized by the similarity between the data,and the basis function is used as the kernel function. The multi-input multi-output model of the furnace temperature and soft sensor is established by using the method of least squares support vector machine.In order to improve the prediction accuracy of the model,this paper uses pruning method and matrix method of combining theoretical knowledge of the model line calibration. Using the improved particle swarm algorithm based on parameter optimization, according to the principle of conservation of energy,the penalty coefficient, the penalty compensation mechanism for loss of energy balance of the particle swarm, local search and global the search ability, the paper obtaines the temperature soft measurement model parameter optimization.Taking a regional temperature as the representative, the offline LSSVM model is tested and compared with the model, the online model and the improved particle swarm optimization algorithm.Finally, the development and testing of the boiler furnace temperature calculation software is completed on the C++ Visual 6 development platform.According to the actual running data of #2 unit in Huaneng Jiaxiang power plant,the experimental results show that the soft measurement method can adapt to changes in unit operating conditions, real-time monitoring of the furnace temperature to achieve the desired effect, convenient operating personnel to monitor and control furnace temperature, in order to further improve boiler combustion optimization provides the basis.