skip to main content

Title: Weather Analysis Using Neural Networks for Modular Data Centers
The objective of this work is to introduce the application of an artificial neural network (ANN) to assist in the evaporative cooling in data centers. To achieve this task, we employ the neural network algorithms to predict weather conditions outside the data center for direct evaporative cooling (DEC) operations. The predictive analysis helps optimize the cooling control strategy for maximizing the usage of evaporative cooling thereby improving the efficiency of the overall data center cooling system. A typical artificial neural network architecture is dynamic in nature and can perform adaptive learning in minimal computation time. A neural network model of a data center was created using operational historical data collected from a data center cooling control system. The neural network model allows the control of the modular data center (MDC) cooling at optimum configuration in two ways. First way is that the network model minimizes time delay for switching the cooling from one mode to the other. Second way, it improves the reaction behavior of the cooling equipment if an unexpected ambient condition change should come. The data center in consideration is a test bed modular data center that comprises of information Technology (IT) racks, Direct Evaporative cooling (DEC) and Indirect Evaporative Cooling (IEC) modules; the DEC/IEC are used together or in alternative mode to cool the data center room. The facility essentially utilizes outside ambient temperature and humidity conditions that are further conditioned by the DEC and IEC to cool the electronics, a concept know as air-side economization. Various parameters are related to the cooling system operation such as outside air temperature, IT heat load, cold aisle temperature, cold aisle humidity etc. are considered. Some of these parameters are fed into the artificial neural network as inputs and some are set as targets to train the neural network system. After the training the process is completed, certain bucket of data is tested and further used to validate the outputs for various other weather conditions. To make sure the analysis represents real world scenario, the operational data used are from real time data logged on the MDC cooling control unit. Overall, the neural network model is trained and is used to successfully predict the weather conditions and cooling control parameters. The prediction models have been demonstrated for the outputs that are static in nature (Levenberg Marquardt method) as well as the outputs that are dynamic in nature i.e., step-ahead & multistep ahead techniques.  more » « less
Award ID(s):
Author(s) / Creator(s):
; ; ;
Date Published:
Journal Name:
ASME 2018 International Technical Conference and Exhibition on Packaging and Integration of Electronic and Photonic Microsystems
Page Range / eLocation ID:
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. With an increase in the need for energy efficient data centers, a lot of research is being done to maximize the use of Air Side Economizers (ASEs), Direct Evaporative Cooling (DEC), Indirect Evaporative Cooling (IEC) and multistage Indirect/Direct Evaporative Cooling (I/DEC). The selection of cooling configurations installed in modular cooling units is based on empirical/analytical studies and domain knowledge that fail to account for the nonlinearities present in an operational data center. In addition to the ambient conditions, the attainable cold aisle temperature and humidity is also a function of the control strategy and the cooling setpoints in the data center.The primary objective of this study is to use Artificial Neural Network (ANN) modelling and Psychrometric bin analysis to assess the applicability of various cooling modes to a climatic condition. Training dataset for the ANN model is logged from the monitoring sensor array of a modular data center laboratory with an I/DEC module. The data-driven ANN model is utilized for predicting the cold aisle humidity and temperatures for different modes of cooling. Based on the predicted cold aisle temperature and humidity, cold aisle envelopes are represented on a psychrometric chart to evaluate the applicability of each cooling mode to the territorial climatic condition. Subsequently, outside air conditions favorable to each cooling mode in achieving cold aisle conditions, within the ASHRAE recommended environmental envelope, is also visualized on a psychrometric chart. Control strategies and opportunities to optimize the cooling system are discussed. 
    more » « less
  2. Abstract Airside economizers lower the operating cost of data centers by reducing or eliminating mechanical cooling. It, however, increases the risk of reliability degradation of information technology (IT) equipment due to contaminants. IT Equipment manufacturers have tested equipment performance and guarantee the reliability of their equipment in environments within ISA 71.04-2013 severity level G1 and the ASHRAE recommended temperature-relative humidity (RH)¬†envelope. IT Equipment manufacturers require data center operators to meet all the specified conditions consistently before fulfilling warranty on equipment failure. To determine the reliability of electronic hardware in higher severity conditions, field data obtained from real data centers are required. In this study, a corrosion classification coupon experiment as per ISA 71.04-2013 was performed to determine the severity level of a research data center (RDC) located in an industrial area of hot and humid Dallas. The temperature-RH excursions were analyzed based on time series and weather data bin analysis using trend data for the duration of operation. After some period, a failure was recorded on two power distribution units (PDUs) located in the hot aisle. The damaged hardware and other hardware were evaluated, and cumulative corrosion damage study was carried out. The hypothetical estimation of the end of life of components is provided to determine free air-cooling hours for the site. There was no failure of even a single server operated with fresh air-cooling shows that using evaporative/free air cooling is not detrimental to IT equipment reliability. This study, however, must be repeated in other geographical locations to determine if the contamination effect is location dependent. 
    more » « less
  3. Two self-developed control schemes, ON/OFF and supervisory control and data acquisition (SCADA), were applied on a hybrid evaporative and direct expansion (DX)-based model data center cooling system to assess the impact of controls on reliability and energy efficiency. These control schemes can be applied independently or collectively, thereby saving the energy spent on mechanical refrigeration by using airside economization and/or evaporative cooling. Various combinations of system-level controls and component-level controls are compared to a baseline no-controls case. The results show that reliability is consistently met by employing only sophisticated component-level controls. However, the recommended conditions are met approximately 50% of the simulated time by employing system-level controls only (i.e., SCADA) but with a reduction in data center cooling system power usage effectiveness (PUE) values from 3.76 to 1.42. Moreover, the recommended conditions are met at all averaged times with an even lower cooling system PUE of 1.13 by combining system-level controls only (SCADA and ON/OFF controls). Thus, the study introduces a simple method to compare control schemes for reliable and energy-efficient data center operation. The work also highlights a potential source of capital expenses and operating expenses savings for data center owners by switching from expensive built-in component-based controls to inexpensive, yet effective, system-based controls that can easily be imbedded into existing data center infrastructure systems management. 
    more » « less
  4. null (Ed.)
    Model predictive control (MPC) has been widely investigated for climate control of commercial buildings for both energy efficiency and demand flexibility. However, most MPC formulations ignore humidity and latent heat. The inclusion of moisture makes the problem considerably more challenging, primarily since a cooling and dehumidifying coil model which accounts for both sensible and latent heat transfers is needed. In our recent work, we proposed an MPC controller in which humidity and latent heat were incorporated in a principled manner, by using a reduced-order model of the cooling coil. Because of the highly nonlinear nature of the process in a cooling coil, the model needs to be modified based on certain weather/climatic conditions to have sufficient prediction accuracy. Doing so, however, leads to a mixed-integer nonlinear program (MINLP) that is challenging to solve. In this work, we propose an MPC formulation that retains the NLP (nonlinear programming problem) structure in all climate zones/weather conditions. This feature makes the control system capable of autonomous operation. Simulations in multiple climate zones and weather conditions verify the energy savings performance, and autonomy of the proposed controller. We also compare the performance of the proposed MPC controller with an MPC formulation that does not explicitly consider humidity. Under certain conditions, it is found that the MPC controller that excludes humidity leads to poor humidity control, or higher energy usage as it is unaware of the latent load on the cooling coil. 
    more » « less
  5. When operating in direct evaporative cooling (DEC) mode, the amount of moisture added to a system can be controlled by frequently modulating water supply to the wet cooling media. Though many challenges arise due to geographical and site conditions, this concept can be applied to data centers to serve as a cost-effective alternative for maintaining the operating temperature of the facility at any weather condition. However, this method results in scale and mineral build up on the media because of an irregular water distribution. To prevent the scale formation, the operators allow the water supply continuously on the cooling media ultimately leading towards the high consumption of facility water and significantly deteriorating the Wet cooling media life. This challenge has been addressed for the first time by experimentally characterizing the vertically split distribution wet cooling media. These systems allow some section of the media to be wetted while other sections remain dry. Various configuration of vertically staged media may be achieved by dividing the full width of the media into two, three, four or more number of equal and unequal sections and providing individually controlled water distribution headers. To increase the number of stages and provide smooth transition from one stage to the other, a MATLAB code is written to find width of DEC media sections for known total width of the media and number of sections. Here, an experimental design to characterize the performance characteristics of a vertically split wet cooling media which has separate water distribution setup has been presented. Apart from relative humidity and temperature, other parameters of interests like pressure drop across the media and saturation efficiency of the rigid media are presented. In the unequal configuration, the media was tested for 0%, 33%, 66%, and 100%. This research provides a potential solution towards the limitation of direct evaporative cooling in terms of energy savings, facility water, reliability and contaminants. 
    more » « less