The Study Of The Soft Sensing Of Oxygen Content In Flue Gases Based On Multivariable Generalized Predictive Control

Now the world is short of energy, maintaining boiler combustion efficiencybecame an important part of energy saving in power plant. The detection of flue gasoxygen content is an important means to reduce coal consumption and maintain thecombustion boiler economy, therefore, the accurate measurement of oxygen contenthas great significance for thermal power plant energy efficiency and environmentalprotection. Generally the industrial process measures oxygen content by the sensor,because of large investment and poor work environment, people began to usesoft-sensing technique to achieve the purpose of direct measurement.After researching the present of the soft sensing of oxygen content in flue gases,in this paper, multivariable generalized predictive control is applied to soft sensing ofoxygen content in flue gases. The model’s parameters are less, it inherited thecharacteristics of the adaptive technologies, and its many steps forecast, rollingoptimization and feedback correction mechanism is more suitable for industrial field.The main tasks are: Put the data pre-treat, enhance data reliability, reduce dimensionby the factor analysis, find out the variables which have the most close relations withoxygen content in flue gases as input of the model, improve the model accuracy;Construct model of multivariable generalized predictive control, estimate parametersof the model, derive the output model and control law, predict oxygen content in fluegases, and the forecasted values are compared with actual oxygen content. Usingparticle swarm optimization (PSO) to optimize the parameters combination of themodel, select the best parameters combination quickly to predict oxygen content.After the experiment, simulation results show that the oxygen content predictedvalues by optimization are closer to the real value.