Cascaded Fractional Kalman Filtering for State and Current Estimation of Large-Scale Lithium-Ion Battery Packs

Abstract

In this paper, a cascaded fractional Kalman filter for state of charge and branch current estimation of largescale battery systems is proposed. As a centralized approach for the estimation of a large-scale system is costly in terms of effort and time, a partition into smaller and, therefore, simpler subsystems is applied. Since the overall system is divided into smaller units, a local computation is allowed and complexity reduced. In these distributed systems, usually, the subsystems communicate with each other to exchange relevant data. Using a model based on mesh currents, we receive a cascaded system structure which results in a hierarchical arrangement of all subsystems. This concludes in a one-directional information flow and, therefore, reduces the overall communication effort. Using this proposed approach, it is not only possible to estimate the states of each branch locally but also to calculate the branch currents when the total current is known. Finally, a practical test with real measurement data is presented.

Fractional Extended and Unscented Kalman Filtering for State of Charge Estimation of Lithium-Ion Batteries

“…The FKF and FEKF have been applied to SOC estimation of battery cells employing piecewise linear [4] and nonlinear [5] fractional order models based on the RQ chain equivalent circuit. In [25] the FEKF has been applied for SOC-and current estimation of battery packs. Also, in [5] a comparison between integer and fractional order battery models and filters has been conducted, showing higher accuracy of the fractional models as well as faster convergence rates of the FEKFs.…”

Current and State of Charge Estimation of Lithium-Ion Battery Packs Using Distributed Fractional Extended Kalman Filters

“…, i P,k ) . The result is a state space model of the entire battery pack (upper index b), identical to [29] as…”

“…Another approach has been given in [27] which employs a filtered terminal voltage for SOC estimation without current measurement, but modeling and measurement errors are not considered. In [29], a real-time capable approach is given which estimates the string currents as well as the SOCs of all cells in a battery using cascaded fractional models. However, since an unidirectional communication procedure between the strings is used, an error propagation to all following subsystems is likely possible.…”

“…However, since an unidirectional communication procedure between the strings is used, an error propagation to all following subsystems is likely possible. An algorithm where each subsystem receives the same information is expected to be more robust, but the mesh current based model of [29] is not suitable for this. Summarizing, except [29], there exist no real-time capable, model based methods for current estimation of battery packs, yet.…”

“…An algorithm where each subsystem receives the same information is expected to be more robust, but the mesh current based model of [29] is not suitable for this. Summarizing, except [29], there exist no real-time capable, model based methods for current estimation of battery packs, yet. However, there exists an approach for distributed estimation of the temperature field across a Liion battery [30] showing that a distribution scheme achieves reduced computational costs.…”