Air Liquide (United States)

The water electrolysis process requires a high DC current supply that can sustain the desired hydrogen production rate over a large period of operation at a competitive cost.During the conversion of electricity from AC to DC, power quality may be affected because of the non-linear effect caused by the power electronics.Most of the recent research has focused on exploring different rectifier topologies.None of them have investigated the influence of cell stack degradation on the performance of power electronics.In this work, we built a one-way interaction model to predict the influence of electrolyzer degradation on power electronics output over multiscale operational time (from milliseconds to years) for proton exchange membrane electrolyzer (PEM).In this model, we assume a constant degradation rate on the electrolyzer that results in a linear increase of internal resistance over time.Counterintuitively, rather than the power quality decreasing, results show that the power quality increased with the electrolyzer degradation for both the AC (power factor and THD) and DC side (ripple) for the 6-pulse thyristor.Furthermore, the influence of three variables (degradation rate, load current, and topology) on AC (power factor and THD) and DC (ripple factor) side power output were investigated.Finally, results were partially validated with experimental data from a 20 MW scale PEM electrolyzer.

Abstract Ground-based thermodynamic and kinematic profilers were placed adjacent to the western shore of Lake Michigan at two sites as part of the 2017 Lake Michigan Ozone Study. The southern site near Zion, Illinois, hosted a microwave radiometer (MWR) and a sodar wind profiler, while the northern site in Sheboygan, Wisconsin, featured an Atmospheric Emitted Radiance Interferometer (AERI), a Doppler lidar, and a High Spectral Resolution Lidar (HSRL). Each site experienced several lake-breeze events during the experiment. Composite time series and time–height cross sections were constructed relative to the lake-breeze arrival time so that commonalities across events could be explored. The composited surface observations indicate that the wind direction of the lake breeze was consistently southeasterly at both sites regardless of its direction before the arrival of the lake-breeze front. Surface relative humidity increased with the arriving lake breeze, though this was due to cooler air temperatures as absolute moisture content stayed the same or decreased. The profiler observations show that the lake breeze penetrated deeper when the local environment was unstable and preexisting flow was weak. The cold air associated with the lake breeze remained confined to the lowest 200 m of the troposphere even if the wind shift was observed at higher altitudes. The evolution of the lake breeze corresponded well to observed changes in baroclinicity and calculated changes in circulation. Collocated observations of aerosols showed increases in number and mass concentrations after the passage of the lake-breeze front.

Community detection decomposes large-scale, complex networks ‘optimally’ into sets of smaller sub-networks. It finds sub-networks that have the least inter-connections and the most intra-connections. This article presents an efficient community detection algorithm that detects community structures in a weighted network by solving a multi-objective optimization problem. The whale optimization algorithm is extended to enabe it to handle multi-objective optimization problems with discrete variables and to solve the problems on parallel processors. To this end, the population’s positions are discretized using a transfer function that maps real variables to discrete variables, the initialization steps for the algorithm are modified to prevent generating unrealistic connections between variables, and the updating step of the algorithm is redefined to produce integer numbers. To identify the community configurations that are Pareto optimal, the non-dominated sorting concept is adopted. The proposed algorithm is tested on the Tennessee Eastman process to show its application and performance.

Gas separation membranes are expected to play a significant role in green house reduction, renewable energy production, and energy savings for separations. However, newly developed membrane materials are constrained by the trade-off between selectivity and permeability and existing body of literature lacks the clear guidance on directions for the investigation. In this paper, we first systematically reviewed the effects of mixed gases and contaminants on membrane selectivities from gas transport mechanism. We then examined cascade membrane process design to achieve the desired product recovery and purity, utilizing two types of membranes and utilizing pressure ratio as the guidance to design the stages. From these examinations, we concluded that high membrane selectivity is critically needed, while high membrane permeability has limited impact for practical applications. Process design and economics for biogas purification was utilized as an example to demonstrate the need for high selectivity membranes. We further identified some gas separation applications that are critically in need of high membrane selectivities, such as CO2 capture from flue gases, and highlighted recent progress in membrane materials with high gas selectivities for these applications.

More and more healthcare data are becoming readily available nowadays. These data can help the healthcare professionals and patient themselves to better understand the patient status and potentially lead to improved care quality. However, the analysis of these data are challenging because they are large-scale and heterogeneous, high-dimensional and sparse, temporal but irregularly sampled. In this paper, we propose a method called Double Core Memory Networks (DCMN) to integrate information from different modalities of the longitudinal patient data and learn a joint patient representation effective for downstream analytical tasks such as risk prediction. DCMN is designed not only to disentangle the temporal and non-linear intra-modal dependencies for the data within each modality but also to capture the long-term inter-modal interactions. DCMN models are the end-to-end memory networks with two external memory cores where each modality of data is compressed and stored. Each memory core has an information-flow controller named query to interact with an external memory module. In addition, we incorporate a gating mechanism into basic DCMN model to perform dynamic regulation of memory interaction. DCMN models have multiple computational layers (hops) allowing data of different modalities interacting with each other recurrently along with a mechanism of alternating access of external memory for each memory core hop-by-hop. We evaluate DCMN models on two outcome prediction tasks, including a mortality prediction on the public Medical Information Mart for Intensive Care III (MIMIC-III) database and a cost prediction on the Hospital Quality Monitoring System (HQMS) dataset. Experimental results demonstrate that our DCMN models are more competitive over the baseline methods in the multimodal prediction setting.

Inthis work, the shock sensitivity of hexachlorodisilane (HCDS)hydrolysis products was studied. The hydrolysis conditions includedvapor and liquid HCDS hydrolysis in moist air. Shock sensitivity wasdetermined by using a Fall hammer apparatus. Extensive infrared studieswere done for the hydrolysis products. It was found that the Si–Sibond in HCDS during hydrolysis is preserved and can be cleaved byshock, leading to intramolecular oxidation of the neighboring silanol(Si–OH) groups to form a networked Si–O–Si structureand hydrogen gas. The limiting impact energy for shock sensitivitywas also found proportional to the oxygen/silicon ratio in the deposit.Finally, recommendations are given for controlling the shock sensitivityof the hydrolyzed deposit.