The Zinc-Nickel single flow battery (ZNB) is a new and special type of flow batteries with a number of promising features, such as membrane free and high scalability, and thus has attracted substantial interests in recent years. However, little has been done so far to investigate how to effectively and reliably manage this new type of batteries. Significant developments are required from the engineering prospective: investigation of battery modelling and battery characterization techniques, accurate state estimation, the ability of instantaneous power acceptance and deliverance, the judgement of the battery health, the determination of battery maintenance time, and long-term performance characterization in real applications. This thesis consists of original contributions in the battery modelling and management at system level.
Three battery modelling techniques, namely the artificial neural network (ANN) based battery modelling, electrochemical mathematical battery modelling approaches and the equivalent circuit based battery modelling are examined and compared. Due to the timeliness of the state estimation, the state-of-charge (SoC) estimation needs to be conducted online in the battery management system. In order to improve computing efficiency, an open-circuit-voltage (OCV) observer based on-line joint estimation of both the state-of-charge (SoC) and the state-of-health (SoH) is proposed. At this point, the proposed open-circuit-voltage~(OCV) observer can not only enhance the estimation accuracy, but also provide a novel framework where the filter dimension is reduced to one, which offsets the increased complexity issue when higher order equivalent circuit models (ECMs) are introduced.
On the other hand, the performance of battery management system (BMS) is highly dependent on the accuracy of state estimation. By incorporating the merits of model predictive control (MPC) scheme, a novel model predictive control based observer (MPCO) for the working conditions monitoring is then proposed. Two remarkable advantages can be achieved against some current state-of-the-art studies. One benefit comes from the rolling horizon scheme and another is introduced by the imposed constraints on the optimization problem.
Due to the high variability of the intermittent renewable energy sources, load demands, and operating conditions, the state of charge (SoC) is not an ideal indicator to gauge the battery capability to deliver the required services. Alternatively, the peak power is more closely related to the instantaneous power acceptance and deliverance, and its real-time estimation plays a key role in grid-tied energy storage systems. In this thesis,a novel peak power prediction method is developed based on rolling prediction horizon. Four indices are proposed to capture the characteristics of the peak power capability over variable prediction windows. The consequent impact of the electrode material and applied flow rate on peak power deliverability are analysed qualitatively.
From the battery maintenance perspective, longer lifespan can be obtained by the periodic reconditioning. However, there are no indicators to explicitly identify the health status of Zinc-Nickel single flow battery and determine the moment of reconditioning. In this thesis, two health indices in terms of the growth of internal resistance and degradation of the battery capacity are compared. Experimental results confirm that the health status of Zinc-Nickel single flow battery is more sensitive to capacity variations. An indicator according to the capacity changes is thus proposed to judge the timing of reconditioning.
