Batteries In Data Centers Research Articles

VRLA battery fault prediction for data center based on random forest model and feature enhancement method

The Valve-Regulated Lead-Acid (VRLA) battery is an important part of data center power supply system. Battery failure will threaten the safe operation of the data center. How to predict the impending battery failure in advance is an urgent problem to be solved to ensure the operation safety of the data center. However, batteries in such precise data centers rarely fail. Few faulty samples cause extreme imbalance between normal samples and faulty samples. In the data center site, the battery is usually in a floating state and the battery charge-discharge cycle times are less. As a result, the obtained battery data has a single working condition. In addition, VRLA batteries are faced with the problems of limited observation information (large amount of data but low data dimension, only voltage, resistance and temperature collected directly from sensors).In this paper, a feature enhancement method is proposed by analyzing the working characteristics of VRLA batteries in the data center. This method extends the two-dimensional characteristics (voltage, resistance) of battery to nine-dimensional characteristics to solve the problem of limited observation information of VRLA battery. The problem of extreme imbalance between normal samples and faulty samples of battery is solved based on clustering undersampling method. Based on the above two methods, a VRLA battery fault classification prediction model is proposed. The nine-month operation data of 1000 VRLA battery were randomly selected from a data center and combined with the simulated fault samples to form a test set. The test results show that the F-score value of the model is increased from 54.5 % to 97.5 % after the clustering undersampling method and the feature enhancement method proposed in this paper. Compared with the VRLA battery replacement strategy recommended in IEEE STD 1188-2005 on this test set, the method can predict the impending battery failure at least 3 days in advance.

Network thermal-runaway propagation model – An experimental and modeling study

Thermal runaway is a major hazard of Li-ion batteries (LIBs). This study investigates thermal runaway propagation (TRP) of LIB modules made of 18650 form-factor LIB cells. This type of LIB cell is widely used in applications such as data center battery backup units and electric vehicles. TRP experiments were conducted with single cells or nine-cell arrays for two types of LIBs. The cathode materials for the two LIB types were lithium cobalt oxide (LiCoO2 or LCO) and lithium iron phosphate (LiFePO4 or LFP). Both battery types had graphite anodes and were tested at 100% state-of-charge. The TRP rates across the LCO cell arrays were significantly faster than those across the LFP cell arrays. Experiments also showed that heat transfer by ejected material within an enclosure can significantly accelerate TRP and needs to be included the TRP model. In addition to the TRP experiment, heat transfer between copper slugs with the same shape as an 18650 LIB cell was experimentally investigated to support the heat transfer modeling. The TRP modeling of the 18650 cell arrays included modeling LIB cell thermal runaway behavior and heat transfer between cells. A previous study developed a lumped thermal runaway model for single 18650 cells. Based on this approach, the current study developed a network TRP model coupling the lumped cell-level thermal runaway model with heat transfer among cells. The model successfully predicts the average TRP rate in LFP arrays and illustrates the importance of different energy release and heat transfer modes for different battery types. The modeled TRP rates for LCO arrays showed some deviations from the experiments due to the increased complexity and greater importance of heat transfer by ejected material.

Economic scheduling of virtual power plant in day-ahead and real-time markets considering uncertainties in electrical parameters

The virtual power plant (VPP) concept, which offers energy efficiency, benefits significantly from the tremendous potential of integrating distributed energy resources (DERs). This work proposes optimal economic dispatch of virtual power plant (VPP), and the problem is formulated in a novel way considering the potential of energy storage systems (ESS) of electric vehicles (EVs) and data center (DC) in a data center facility test model using optimized energy management system (EMS). The problem is evaluated for maximizing the revenue for VPP considering market prices and the uncertainties in the distributed energy resources. To this purpose, a techno-economic analysis of the proposed test system of VPP is evaluated using mixed-integer linear programming (MILP). Renewable generation, electricity price, load demand, and real-time EV behavior uncertainty analysis are forecasted using Long Short-Term Memory (LSTM) along with DC battery degradation analysis which constitutes the optimization model. The results showed that the VPP EMS can perform optimal power scheduling with the day ahead and real-time electricity price scenarios. Moreover, further analysis including battery degradation cost of DC UPS ESS proved the capability of the proposed framework by up to 2064.4 $/h revenue for day-ahead price and 562 $/h revenue for the real-time price compared to revenue without battery degradation cost. A simulation scenario is also carried out where the constant load demand and the battery storage powers are varied between 50% and 150%. It can be concluded that the 500 kW and 750 kW ESS (100% and 150%) are better choices as compared to 250 kW (50%) during increased load demand for investors due to their higher revenue. Hence it is concluded that this research highlights the importance of revenue benefits from deploying VPP with the collaborative operation of DERs.

Data driven battery anomaly detection based on shape based clustering for the data centers class

Abstract Batteries are a significant part of data centers, which ensure the uninterrupted working of a data center. Using online measurement to find out odd batteries in data centers is challenging due to lack of training samples since there are only a very few full charging-discharging cycles during the lifetime of batteries. In this paper, a new battery anomaly detection method based on time series clustering is proposed. This method uses only battery operating data and does not depend on offline testing data, thus provides a way to improve the maintenance efficiency and lessen batteries operating risks in data centers. Effectiveness of the proposed method is demonstrated and confirmed by a case study for 40 batteries in an existent data center.

Cluster-Based Prediction for Batteries in Data Centers

Prediction of a battery’s health in data centers plays a significant role in Battery Management Systems (BMS). Data centers use thousands of batteries, and their lifespan ultimately decreases over time. Predicting battery’s degradation status is very critical, even before the first failure is encountered during its discharge cycle, which also turns out to be a very difficult task in real life. Therefore, a framework to improve Auto-Regressive Integrated Moving Average (ARIMA) accuracy for forecasting battery’s health with clustered predictors is proposed. Clustering approaches, such as Dynamic Time Warping (DTW) or k-shape-based, are beneficial to find patterns in data sets with multiple time series. The aspect of large number of batteries in a data center is used to cluster the voltage patterns, which are further utilized to improve the accuracy of the ARIMA model. Our proposed work shows that the forecasting accuracy of the ARIMA model is significantly improved by applying the results of the clustered predictor for batteries in a real data center. This paper presents the actual historical data of 40 batteries of the large-scale data center for one whole year to validate the effectiveness of the proposed methodology.

Data centers as a source of dynamic flexibility in smart girds

Data centers have a significant potential to become a major source of flexibility in smart girds. They consume currently roughly 3% of all the electricity produced globally and are expected to only increase their consumption as the world becomes more connected and digitalized.As data centers are required to operate without any interruptions, they use power protection systems and energy storages. This paper investigates the technical and economic feasibility of dual-purposing these power protection systems, the uninterruptible power supplies, and their batteries in data centers to perform primary frequency regulation services. While the topic of data centers and demand response has been extensively covered in the current scientific literature, the focus has been on the demand response enabled by server workload shifting or hardware-enabled peak shaving. Based on an extensive literature review, there is a knowledge gap in the literature concerning primary frequency regulation and dynamic response enabled by modern power electronics systems in data centers. In this paper, this knowledge gap is bridged by suggesting a novel approach of taking advantage of the bidirectional operations capabilities of the uninterruptible power supply systems, thereby enabling them to provide dynamic power response from their battery systems.The feasibility of this approach is examined with the proposed method, which includes (1) an analysis of the required energy for primary regulation and the availability of this energy in a typical data center, (2) a simulation of activation events and their impact on the service life of the battery systems, (3) reaction speed and reliability considerations of the operations, and (4) an economic feasibility and balancing market analysis.The results show that as primary frequency regulation is an energy nonintensive service and data center battery systems are by design oversized for redundancy reasons, typical data centers have more than ample amounts of energy to participate in the primary regulation without jeopardizing their own processes. The results also show that by maintaining reasonable levels of usage, the battery systems can be operated within their specifications, and the demand response operations will not cause premature aging of the battery systems. The reaction speed of the power electronics is found to be very high and easily meet the current market requirements.While the achievable revenue from the primary regulation service is small compared for example with the electricity costs of the data centers, it is still significant as there is little to no impact on the daily business of the data centers.

Some Constraints on the Reuse of Li-ion Batteries in Data Centers

Abstract The development of rechargeable electrochemical sources, especially those based on Li-ion technology, has opened the way for their use in various fields ranging from electronics and IT to electric cars. New ideas appear, such as using replaced the batteries from electrical vehicles in the development of energy storage devices for critical consumers such as Data Centers. New uses impose detecting the unfavorable operating conditions which may endanger the power supply of Data Centers. This paper, by modeling and simulation, analyzes some particular problems of Li-ion batteries that appear due to the differences in state of charge of the cells.

Optimization of Data Center Battery Storage Investments for Microgrid Cost Savings, Emissions Reduction, and Reliability Enhancement

This paper presents a methodology for optimizing investment in data center battery storage capacity. Utility grid managers spend significant resources toward predicting and matching available power generation capacity to demand in real time. It is therefore essential for the success of the power industry that economic dispatch, energy efficiency, and grid security be maintained as power requirements change. This is especially challenging for microgrids during periods of peak demand due to limited available capacity. Data centers possess a unique requirement for short-term battery power supply where cost savings, emissions reduction, and reliability enhancement can be achieved through investment in additional battery capacity. To maximize these benefits, an optimization methodology is presented through a case study for an existing data center and microgrid. Here, we discuss a case study demonstrating the effectiveness of the proposed approach. For the selected mid-size data center, our results indicate monetize monthly savings of up to 10 000$ and 0.5% reduction in loss of load probability while simultaneously reducing carbon footprints. The results of this work are directed toward large data centers at university and corporate campuses, microgrids, and military installations.

Benefits and limitations of tapping into stored energy for datacenters

Datacenter power consumption has a significant impact on both its recurring electricity bill (Op-ex) and one-time construction costs (Cap-ex). Existing work optimizing these costs has relied primarily on throttling devices or workload shaping, both with performance degrading implications. In this paper, we present a novel knob of energy buffer (eBuff) available in the form of UPS batteries in datacenters for this cost optimization. Intuitively, eBuff stores energy in UPS batteries during "valleys" - periods of lower demand, which can be drained during "peaks" - periods of higher demand. UPS batteries are normally used as a fail-over mechanism to transition to captive power sources upon utility failure. Furthermore, frequent discharges can cause UPS batteries to fail prematurely. We conduct detailed analysis of battery operation to figure out feasible operating regions given such battery lifetime and datacenter availability concerns. Using insights learned from this analysis, we develop peak reduction algorithms that combine the UPS battery knob with existing throttling based techniques for minimizing datacenter power costs. Using an experimental platform, we offer insights about Op-ex savings offered by eBuff for a wide range of workload peaks/valleys, UPS provisioning, and application SLA constraints. We find that eBuff can be used to realize 15-45% peak power reduction, corresponding to 6-18% savings in Op-ex across this spectrum. eBuff can also play a role in reducing Cap-ex costs by allowing tighter overbooking of power infrastructure components and we quantify the extent of such Cap-ex savings. To our knowledge, this is the first paper to exploit stored energy - typically lying untapped in the datacenter - to address the peak power draw problem.