Hierarchical thermal management for PEM fuel cell with machine learning approach
Abstracts:Thermal management is crucial for the mass transport and water balance of proton exchange membrane fuel cell (PEMFC). Inspired by this, a hierarchical thermal management strategy (TMS) is proposed for fuel cell hybrid electric vehicle (FCHEV). In particular, the transient TMS demands are determined by a well-designed energy management strategy (EMS) taking health and thermal safety into consideration. Furthermore, along with the high-efficiency heat dissipation, a hydrogen consumption minimization strategy (HCMS) is proposed via optimal temperature tracking, which investigates the desirable trace offline. These parallel strategies are incorporated through the deep reinforcement learning (DRL)-based deep deterministic policy gradient (DDPG) algorithm. With the help of its self-adaptive ability, DDPG deals with the complicated TRS problem in multidimensional coupled cooling system, through a mutually updated actor-critic framework. Results suggest the superiority and reliability of proposed TMS with respect to the stack efficiency, fuel economy and tracking performance.
Improved convolutional neural network chiller early fault diagnosis by gradient-based feature-level model interpretation and feature learning
Abstracts:For chillers, fault diagnosis (FD) is important for maintaining system reliability and performance. Deep learning methods, such as convolutional neural network (CNN), have been widely studied for chiller FD for its more significant diagnosis accuracy. But CNN model with deep layers and complex structures is black-box and difficult to interpret, which would greatly limit its practical FD applications for chillers. Traditional CNN model interpretation method may be not sensitive to interpret the chillers systematic faults especially at their early stages. Hence, to further obtain better interpretation of the CNN FD model, this study proposed a high-sensitivity gradient-based interpretation method. The proposed method adopts a softsign-forward-ReLU-backward manner to interpret the CNN model from the prospective of fault-discriminative feature, which localizes the fault-related feature variables and visualizing the diagnosis criteria for the CNN identified faults. The ASHRAE research project 1043 (RP-1043) chiller fault dataset was used to validate the proposed model interpretation and explanation method with higher feature-level sensitivity for the incipient fault. Based on the feature-level explanation and feature learning results, different feature combinations were investigated to improve diagnosis accuracy of some early-stage faults. If only small sizes of training data were available for modelling, 17 fault-related features were selected from the original 64 features to re-develop the CNN model and achieved diagnosis accuracy improvement of 9% for the early-stage improper refrigerant charge faults at most.
Filling time-series gaps using image techniques: Multidimensional context autoencoder approach for building energy data imputation
Abstracts:Building energy prediction and management has become increasingly important in recent decades, driven by the growth of Internet of Things (IoT) devices and the availability of more energy data. However, energy data is often collected from multiple sources and can be incomplete or inconsistent, which can hinder accurate predictions and management of energy systems and limit the usefulness of the data for decision-making and research. To address this issue, past studies have focused on imputing missing gaps in energy data, including random and continuous gaps. One of the main challenges in this area is the lack of validation on a benchmark dataset with various building and meter types, making it difficult to accurately evaluate the performance of different imputation methods. Another challenge is the lack of application of state-of-the-art imputation methods for missing gaps in energy data. Contemporary image-inpainting methods, such as Partial Convolution (PConv), have been widely used in the computer vision domain and have demonstrated their effectiveness in dealing with complex missing patterns. Given that energy data often exhibits regular daily or weekly patterns, such methods could be leveraged to exploit the regularity of the data to learn underlying patterns and generate more accurate predictions for missing values. To study whether energy data imputation can benefit from the image-based deep learning method, this study compared PConv, Convolutional neural networks (CNNs), and weekly persistence method using one of the biggest publicly available whole building energy datasets, consisting of 1479 power meters worldwide, as the benchmark. The results show that, compared to the CNN with the raw time series (1D-CNN) and the weekly persistence method, neural network models with reshaped energy data with two dimensions reduced the Mean Squared Error (MSE) by 10% to 30%. The advanced deep learning method, Partial convolution (PConv), has further reduced the MSE by 20%–30% than 2D-CNN and stands out among all models. Based on these results, this study demonstrates the potential applicability of time-series imaging in imputing energy data. The proposed imputation model has also been tested on a benchmark dataset with a range of meter types and sources, demonstrating its generalizability to include additional accessible energy datasets. This offers a scalable and effective solution for filling in missing energy data in both academic and industrial contexts.
A discussion of internal flow characteristics and performance of super long gravity heat pipes for deep geothermal energy extraction
Abstracts:Super long gravity heat pipe (SLGHP) provides a novel way for harvesting thermal energy from a deep geothermal reservoir. Previous research has demonstrated the feasibility of this innovative technique, most studies have focused on the overall performance of the heat extraction system, whereas detailed investigations into the internal flow characteristics remains ambiguous. This hampers the application and further advancement of SLGHPs. In this study, a preliminary analysis of the internal flow resistance characteristics is carried out using of the homogeneous model for two SLGHPs, a regular one and a modified one with an inner coaxial pipe. It indicates that the regular SLGHP has higher flow resistance under the same thermal condition of the reservoir. Specifically, the regular SLGHP experiences approximately 1.75 times larger flow resistance than the modified SLGHP (average temperature of the evaporation section and condenser is 95.6 °C and 55 °C, respectively, with water as the working fluid). While on the other hand, the heat transfer capability of the modified SLGHP was improved 29.73% ∼77.77% at various condensation temperature compared with the regular SLGHP. The dominant factor contributing to pressure drop in the regular SLGHP is the frictional loss resulting from the counter-current flow of rising vapor and falling liquid film. Conversely, the SLGHP with a coaxial pipe effectively separates the vapor and condensate flows, reducing frictional resistance and improving heat transfer performance. It is speculated that explosive nucleate pool boiling may occur in the liquid pool at the bottom of the SLGHP, enhancing heat transfer from the liquid bulk to the upward-flowing vapor.
A numerical simulation of the distribution and the variation law of the liquid water content in icing wind tunnel
Abstracts:To solve the aircraft icing problem, simulation of the high-altitude clouds as the icing environment becomes necessary. Clouds always contain many micro-scale subcooled droplets. The accurate control of the icing environment simulation is challenging. To investigate the distribution and the variation law of the liquid water content in the icing wind tunnel, a simulation model of the icing wind tunnel was developed with the reprogramming software FLUENT. The model was validated with the parameters of ice shape and maximum ice thickness on the model of SC0710 airfoil with a constant cross-section at zero angle of attack. The errors of the maximum ice thickness, the maximum and mean deviation are 5.06%, 0.22 mm and 0.13 mm, respectively. Then, the influences of the parameters of water droplets at the nozzle exit and airflow were simulated and analyzed. The results indicate that increasing the initial velocity or temperature of water droplets at nozzle exit will promote the uniformity of the liquid water content (LWC) distribution and reduce the average LWC in the test section. The average LWC of the test section is significantly reduced due to a large amount of water droplets hitting the wall under the action of inertia once the medium volume diameter (MVD) exceeds 40 μm.
An experimental study on the effect of CO2 laser powers on melting characteristics of ice with trapped air bubbles under vertical irradiation
Abstracts:Icing occurs frequently and plays negative effects in natural and industrial fields. Air bubbles are always trapped in the ice during freezing, and then affect the thermal and mechanical properties of ice. As a typical non-contact and high-efficiency deicing method, laser deicing technology is tested on melting ice with trapped air bubbles, under vertical CO2 laser irradiation and natural convection, with the laser power varied at a range of 20 ∼ 60 W. As resulted, the axial melting rate tends to increase and then decrease with increasing laser power. The maximum values of average and instantaneous axial melting rate at 40 W are 5.46 mm/s and 12.32 mm/s, respectively. At 40 ∼ 60 W, the higher the laser power, the larger the average melting length after stabilization. At 40 W and 60 W, the average melting length are 3.60 mm and 5.59 mm, respectively. It is found little correlation between the melting angle and laser power. At 30 W and 50 W, the maximum melting angles in the front view are 2.49° and 1.15°, respectively. During the whole melting process, the highest energy efficiency is maintained at 20 W, and the maximum energy efficiency is 69.1%. Results of this study could provide a reference for the application of laser deicing technology.
Space heating performance analysis of air source heat pump integrated with image gray recognition-based defrosting controller
Abstracts:In practice, owing to the unreasonable defrosting initiating method of air source heat pumps (ASHPs), mal–defrost occurs frequently, further leading to the increase of building heating energy consumption. In recent years, with the development of image recognition technology in many fields, it has been expected to be used in defrosting initiating control methods of ASHPs. Although former investigations demonstrate that it is feasible and promising, there is still lack of a control strategy to make it become reality. To promote the application of image recognition in defrosting initiating method of ASHPs, a control strategy for the defrosting initiating method based on image gray recognition was proposed in the present work and used in an image gray recognition equipment (IGRE). Then, nine groups of experiments were conducted based on this new IGRE method. Results indicate that the proposed control strategy for the IGRE method is feasible and practical. By employing this new IGRE method, the defrosting accuracy rate and the average COP of the ASHPs are respectively 42.86% and 36.60% higher, compared with those using the conventional T–T method. In complex and variable environments, the defrosting accuracy rate of this new IGRE method reaches 93.33% in a 10–day test.
Experimental investigation on the melting process using freezing concentration for heat source tower heat pump system
Abstracts:Facing with problem of the dilute antifreeze solution in winter, it is necessary for the heat source tower heat pump system to develop a new kind of regenerator to improve the utilization rate of waste liquid. In this article, a solution regenerator unit using freeze concentration with a heat pump system was tested to research variations in the melting process under different operation conditions. From the separation efficiency, the low valve opening degree and high freezing point enhanced the separation process. The maximum separation ratio was all at 50 % valve opening degree operation at each concentration group. As the inlet concentration increased, the separation ratio was downtrend at different opening degree operations. For the icing distribution, the low covering ice area shortened the running time of melting process. The average percentage slope can reflect the influence changes of the ice thickness, the ice covering area and the inlet concentration on the melting process. The average percentage slope was inversely linear with the mass of calculated melted ice. The increase in average percentage slope effectively reduced amount of melting ice.
Thermo-fluid dynamic analysis of the air flow inside an indoor vertical farming system
Keywords:Indoor vertical farming system;Heat and mass transfer;CFD;Measurements;Turbulence
Abstracts:In this study, computational fluid dynamics (CFD) is used to assess the thermo-fluid dynamic behaviour of an indoor vertical farming system. Experimentally driven three-dimensional k-ε steady simulations are performed using a species transport model to account for relative humidity. In-field sensor measurements are used to set the boundary conditions for the simulations and to validate the results in terms of temperature and humidity distribution within the growing cell. The entire cell, consisting of eight levels of growing tables is simulated under daylight conditions, including the heat source from LED lights in the thermo-fluid dynamic air flow field. Results help to understand the main features of the air flow distribution. The comparison of the numerical results with measurements demonstrate the ability of the numerical approach to characterize the thermo-fluid dynamic flow field of the indoor farming system and the reliability of well-calibrated CFD simulations in controlling the air flow distribution, which is necessary to reduce energy consumption and to improve plant growth quality. Critical regions for the plant growth are identified within the farm based on the study of air speed uniformity and on the analysis of vapor pressure deficit (VPD). A growing table efficiency index is introduced based on the definitions of two dimensionless objective uniformity parameters. The analysis of these indices helps identifying differences in the fluid dynamic behaviour of the upper and lower floors of the cell and some deficiencies in the ventilation system ability to provide uniform air conditions to all the growing tables. Results show a symmetric behaviour of the left and right layers for upper floors, with tables in front of the evaporators showing better efficiencies. An opposite behaviour is observed on the lower floors, exhibiting a strong asymmetry between left and right layers. On the other hand, the VPD analysis reveale that certain regions near the evaporators and walls (with good flow uniformity) experience extreme conditions that could affect plant growth. The approach presented in this paper provides a detailed understanding of the indoor vertical farming environment and its impact on growth and yield of leafy greens. This, in turn, helps in the effort to optimize the vertical farm ventilation system and thus its energy, which remains the main issue of such cultivation plants.
Applications of weighted-sum-of-gray-gas model with soot to an experimental aero-engine combustion chamber
Abstracts:Due to the importance of radiation during combustion in the aero-engine, the weighted-sum-of-gray-gas (WSGG) model considering both gas and soot is applied to the combustion modelling for an experimental aero-engine in this work. Five gray gases are chosen to resolve the absorption coefficients of CO2-H2O mixture, with which the Planck-mean absorption coefficients of soot are used for the spectral consideration of gas-soot mixture. Together with the discrete ordinate method, the radiation effects using WSGG model on the combustion are presented. Comparisons are made between full-scale combustion simulations without radiation and those with radiation from gas only or gas-soot mixture. Results show that radiation from either gas or gas-soot mixture leads to no more than 3% differences in temperature and 5% differences in species concentrations. However, the temperature of combustor liner walls in the primary zone (i.e., the main combustion region) is significantly altered with an average of 100 K rise when considering radiation from gas only and a maximum of 500 K rise when considering radiation from gas-soot mixture. Results also show that the temperature of combustor liner walls with radiation from gas-soot mixture becomes more well-distributed than that either without radiation or with radiation from gas only.