Compound Adaptives Kalman Filtering And Its Application

In many practical engineering applications,due to the limitations of people's understanding of the working conditions and the complexity of the application environment,the parameters describing the state space model in the Kalman filtering system are usually partially known or unknown.When discretizing a continuous stochastic system of observability,the noise may have related characteristics;and it does not necessarily ensure that the state of the system is still observable.These features will make the application of observable degree theory based on standard Kalman filtering and the application capabilities of existing adaptive Kalman filtering techniques very challenging.In response to the above problems,a series of research on the theory of observable degree and its adaptive Kalman filtering algorithm have been carried out,and great progress has been made.At the same time,expanding the application fields of adaptive Kalman filtering algorithms,such as power load forecasting,has potential practical application value.Regarding the issue above,this paper has carried out the following five aspects of research work in a targeted manner:(1)The factors of observable degree analysis in the theory of state estimation are analyzed in depth.The observable degree analysis method for the existing analytical stochastic system does not consider the problem of noise influence.This paper first discusses that noise is an important factor affecting the observable degree analysis.It expounds the necessity of considering the influence of noise and the difficulties and challenges of the research,so as to provide a new solution to the analysis of the observable degree theory of stochastic systems.(2)A computational method for effectively characterizing the relationship between observable degree and filtering accuracy of noise-related systems is proposed.For the least square observable degree calculation method,the noise-related system is not considered.In this paper,a method for calculating the observable degree of the noise-related system is obtained by considering the noise-related system.(3)An adaptive filtering algorithm based on Sage-Husa and observable degree analysis is proposed.The problem of time-varying process noise and observable degree difference in the Kalman filtering system leads to a problem of filter performance degradation.Based on the Sage-Husa technology,a composite adaptive Kalman filtering algorithm is established based on the estimation error minimum performance Cramerome lower bound(CRLB)in the estimation theory.(4)An adaptive filtering method based on adaptive technology combination is proposed.The problem of the process noise and measurement noise time-varying in the Kalman filter system and the adverse effect of the state variable of the observable degree difference on the filter,and the Sage-Husa noise estimator can not estimate the defects at the same time.Based on the third chapter,the variational Bayesian method is introduced to estimate the measurement noise,and a composite adaptive Kalman filtering method with wider application range is established.(5)Application of adaptive filtering algorithm in power load forecasting.The modeling difficulties encountered by traditional model prediction methods and the artificial neural network load prediction of black box operations do not provide a good insight into the physical and economic reasons behind load changes.In this chapter,adaptive Kalman filtering algorithm is combined with BP neural network technology to reconstruct a load forecasting model for predicting power load.