Analysis of the Robust Operation of a Hybrid Power Station Incorporating Diesel Generators and Battery Energy Storage Systems in a Novel Data Center Electrical Topology

(GenAI), is driving hyperscale data center operators to adopt more flexible and resilient electrical infrastructures. These electrical networks must operate reliably in both grid-connected and islanded modes to ensure uninterrupted service continuity. However, the highly dynamic, sub-second power demands associated with GenAI training introduce significant stress on power supply and stability, especially during off-grid operation relying on alternative sources.

This paper presents the methodology and key findings of an analysis conducted to evaluate the operation of a hybrid power station (HPS) incorporating diesel generator units and battery energy storage systems (BESSs) to support data center operation under islanded conditions.

The study is based on a detailed simulation model and documentation provided by an equipment manufacturer.

Historically, continuous operation in data centers has been ensured through dedicated lowvoltage (LV) back-up diesel generators located in close electrical proximity to critical information technology (IT) loads during off-grid conditions. However, the evolution of computational workloads and IT architectures, particularly those associated with GenAI, has driven a transition toward more centralized and flexible back-up power generation facilities

(PGFs).

This study pursues two key objectives:

•

• First, a comprehensive and reliable simulation model of a HPS is developed.

Second, the resulting analyses are evaluated to determine whether they provide credible insights into the dynamic security of the system, with particular emphasis on the HPS as a critical asset. To achieve the first objective, detailed and customized simulation models were developed for each component of the HPS. Alternator parameters for the diesel generators were obtained from the manufacturer, together with complete documentation of the associated control systems, including the automatic voltage regulator (AVR), excitation system, and speed governor. The

BESS model was also provided by the manufacturer. To ensure accurate representation of real system dynamics, systematic fine-tuning based on Site Acceptance Test (SAT) was applied to the key power components throughout the model development process.

To address the second objective, the proposed electrical network with the HPS was evaluated through dynamic time-domain analysis under islanded operation. Simulations covered multiple disturbance scenarios, including pulse load behavior, loss of one power generation unit (PGU),

IT load connection and disconnection, etc., and were assessed against relevant performance criteria, such as voltage and frequency stability. The results confirm stable operation of the electrical network.

The main contribution of this work is summarized below:

• • This study underlines the importance of accurate and up-to-date modelling of the dynamic behaviour of the PGUs of a data center, and especially of the HPS, in order to capture the system performance, provide operational solutions and advice that ensure the reliable operation of the facility.

A comprehensive set of simulation-based case studies is proposed to evaluate the security of the electrical network and the continuity of service in data center’s operation during islanded conditions. Based on this analysis, the work provides high-level recommendations on optimal modes of operation specifically Grid-Forming (GFM),

Grid-Following (GFL), for the PGUs of the hybrid power generation facility.