Distributed Unknown Input and State Estimation for Nonlinear Multi-agent Systems with Applications to Battery Management

This paper proposes a novel filtering algorithm for simultaneous estimation of unknown inputs and states of a class of nonlinear discrete-time heterogeneous multi-agent systems. Based on the Taylor approximation of the nonlinear multi-agent system, a distributed semi-cooperative switch-mode filter is developed to achieve the minimum-variance unbiased (MVU) estimation of the unknown inputs and states. Compared with the conventional decentralized EKF-based unknown input filter, the proposed distributed filter has a more relaxed existence condition of the filter, which makes it more applicable in reality. This new type of filter is then successfully applied to the simultaneous estimation of state of charge (SOC) and temperature of a battery pack for battery management of electric vehicles and grid-tied energy storage systems.