Research On The Application Of Kalman Filter In Multivariate Signal Fusion

As the renewable energy represented by the wind incorporated into the grid with large-scale integration,thermal power unit load frequency control task is increasingly difficult.In order to make the power unit response the change of the power load instruction quickly and ensure its safety,economy and environmental at the same time,more effective monitoring and control should be applied for the unit critical system.It is a difficult problem for state detection and optimal control because of difficulties in getting accurate parameters of the key state.In fact,such as coal calorific value,boiler efficiency,heat signals and many other measurements or soft measurements cannot satisfy the accuracy and real-time requirements at the same time.According to the same signal,the static error of the result obtained by some kind of measurement methods is small,but the dynamic error is large.However,the dynamic error which is obtained by adopting a different kind of measurement method is small but the static error is large.It is successful for Kalman Filters to be applied in an Integrated Navigation System.GPS positioning signal has good static accuracy but real-time performance is poor.Conversely,the inertial positioning signal has good real-time performance but the static accuracy is poor.However,Kalman Filters can combine the advantages of these two methods so as to obtain a positioning signal with high static accuracy and real-time performance.This method can be used to the field of thermal signal fusion for reference.For the measurement of heat of coal in the furnace,laboratory tests results have high static accuracy,but the measurement results are discrete points and there exists hysteresis.By the combination of positive and negative balance analysis method of boiler efficiency,the accuracy of the unit is high,but there is a huge dynamic error in the condition of variable load.The results combined with thermodynamic calculation of coal pulverizing system and the composition analysis of flue gas are only effective under stable condition.The results that based on the load-pressure simplified nonlinear dynamic model can reflect the changes of coal calorific value,but static value is difficult to determine.There are different characteristics in static accuracy and real-time performance for different methods.In this paper,two kinds of coal calorific information fusion methods are studied:(1)Combining the essence of prediction and correction on Kalman Filter and the data of coal calorific value obtained by laboratory test,it is efficient for real time correction of the real time coal calorific value based on dynamic model of unit load pressure model.(2)According to the basis of the complementary characteristics between real-time of the dynamic method and static accuracy of the static method which is used to combine thermodynamic calculation of pulverizing system and the analysis of flue gas composition,two methods are used for information fusion.By using real time operation data validation,the integrated heat output signal with static accuracy and dynamic error less than 4% is obtained,which meets the practical requirements of the project.