基于参数自适应学习UKF的非线性受约束状态估计

This project is mainly on the theories, algorithms and applications for constrained nonlinear state estimation. The unscented Kalman filter is combined with the H-infinity filter to make it suitable for state estimation in nonlinear non-Gaussian systems. The typical types of constraints encountered in nonlinear systems and the description methods are firstly clarified; the forementioned improved unscented Kalman filter, with a projection procedure or a shrinkage procedure added, is applied to handle the bounded state estimation problem; theoretical analysis is finally done to depict the performance of the proposed algorithm. The influence of the free parameters of the unscented Kalman filter to the performance of the state estimation will be studied, and adaptive learning algorithms based on the Gaussian process optimizaiton for the selection of the free parameters will also be investigated. All of the proposed algorithms will be tested and evaluated in a platform of Lithium-ion battery state of charge estimation system..The research of this project can provide theoretical support and basic algorithms to the constrained state estimation problem, which is encountered in many nonlinear applications. New ideas and methods for higher SOC estimation accuracy are also hopefully to be brought forward.

本项目主要研究非线性系统中受约束状态的估计理论、算法与应用。采用将Unscented卡尔曼滤波器（UKF）与H-∞滤波器(HF)相结合的方法，开展改进的UKF在非线性非高斯系统状态估计中的应用研究；明确非线性系统中典型的状态约束类型及其描述方法，基于投影过程和收缩过程，采用改进的UKF完成非线性系统中有界约束状态的估计，并进行其性能的理论分析；揭示UKF算法中自由参数的选取对状态估计性能的影响，并基于高斯过程优化方法完成算法自由参数的自适应学习；最终以电动汽车锂电池剩余电量估计系统为平台，对所提出的算法进行测试与性能评估。.通过本项目的研究，可为非线性动态系统中受约束状态的估计应用提供一定的理论支撑和算法基础，有望为获取更高准确度的电动汽车锂电池SOC在线估计提供一种新的思路和方法。

本项目主要研究非线性系统中受约束状态的估计理论、算法与应用。项目采用将Unscented卡尔曼滤波器（UKF）与H-∞滤波器(HF)相结合的方法，提出一种新的Unscented H-∞滤波器（UHF），并开展UHF在非线性非高斯系统状态估计中的应用研究；明确非线性系统中典型的状态约束类型及其描述方法，基于投影过程和收缩过程，采用UHF完成非线性系统中有界约束状态的估计；研究UHF算法中自由参数的选取对状态估计性能的影响，基于粒子群优化(PSO)算法，采用三种不同策略完成UHF算法中自由参数的自适应学习；最终以电动汽车动力电池剩余电量在线估计系统为应用，针对五种不同电动汽车实际工况，对所提出的受约束状态估计算法进行了测试与性能评估。仿真和实验结果表明，所提出的基于PSO参数优化的UHF受约束状态估计算法，可以实现非线性系统中受约束状态的有效估计，状态估计精度均高于现有的常规UKF算法等。项目研究成果可面向其他非线性系统中的受约束状态估计应用进行推广，具有较好的理论意义和实用价值。