Fouling resistance prediction based on GA–Elman neural network for circulating cooling water with electromagnetic anti-fouling treatment

Dynamic simulation experiments were conducted on calcium carbonate fouling formation in shell and tube heat exchangers by using a self-designed online evaluation experimental platform of the electromagnetic anti-fouling effect to obtain the experimental data of conductivity, pH, dissolved oxygen and fouling resistance with the electromagnetic anti-fouling treatment (EAT). And the Elman neural network (Elman NN) was optimized using the genetic algorithm (GA) to derive the GA–Elman neural network (GA–Elman NN). On the basis of GA–Elman NN, a fouling resistance prediction model was established with conductivity, pH, and dissolved oxygen as the input variables and fouling resistance as the output variable. Prediction results indicated that GA–Elman NN improved the weight and threshold, overcame the drawback of falling into the local minimum, and strengthened the capability of finding the optimal solution, thereby improving the prediction accuracy significantly. Moreover, the GA–Elman NN prediction model presented enhanced generalization capability. The mean absolute percent error was 6.07%, and the total error was 8.78% with the experimental system uncertainty. These values indicate that the GA-Elman NN prediction model possesses the high prediction accuracy and is rational and feasible in predicting fouling resistance.     

Hourly solar radiation forecasting using optimal coefficient 2-D linear filters and feed-forward neural networks

In this work, the hourly solar radiation data collected during the period August 1, 2005–July 30, 2006 from the solar observation station in Iki Eylul campus area of Eskisehir region are studied. A two-dimensional (2-D) representation model of the hourly solar radiation data is proposed. The model provides a unique and compact visualization of the data for inspection, and enables accurate forecasting using image processing methods. Using the hourly solar radiation data mentioned above, the image model is formed in raster scan form with rows and columns corresponding to days and hours, respectively. Logically, the between-day correlations along the same hour segment provide the vertical correlations of the image, which is not available in the regular 1-D representation. To test the forecasting efficiency of the model, nine different linear filters with various filter-tap configurations are optimized and tested. The results provide the necessary correlation model and prediction directions for obtaining the optimum prediction template for forecasting. Next, the 2-D forecasting performance is tested through feed-forward neural networks (NN) using the same data. The optimal linear filters and NN models are compared in the sense of root mean square error (RMSE). It is observed that the 2-D model has pronounced advantages over the 1-D representation for both linear and NN prediction methods. Due to the capability of depicting the nonlinear behavior of the input data, the NN models are found to achieve better forecasting results than linear prediction filters in both 1-D and 2-D.     

Effect of assembly error of bipolar plate on the contact pressure distribution and stress failure of membrane electrode assembly in proton exchange membrane fuel cell

In practice, the assembly error of the bipolar plate (BPP) in a PEM fuel cell stack is unavoidable based on the current assembly process. However its effect on the performance of the PEM fuel cell stack is not reported yet. In this study, a methodology based on FEA model, “least squares-support vector machine (LS-SVM)” simulation and statistical analysis is developed to investigate the effect of the assembly error of the BPP on the pressure distribution and stress failure of membrane electrode assembly (MEA). At first, a parameterized FEA model of a metallic BPP/MEA assembly is established. Then, the LS-SVM simulation process is conducted based on the FEA model, and datasets for the pressure distribution and Von Mises stress of MEA are obtained, respectively for each assembly error. At last, the effect of the assembly error is obtained by applying the statistical analysis to the LS-SVM results. A regression equation between the stress failure and the assembly error is also built, and the allowed maximum assembly error is calculated based on the equation. The methodology in this study is beneficial to understand the mechanism of the assembly error and can be applied to guide the assembly process for the PEM fuel cell stack.     

Controlling the microscopic morphology and permeability of catalyst layers in proton exchange membrane fuel cells by adjusting catalyst ink agglomerates

Since agglomerates in catalyst inks affect the catalyst layers (CL) and membrane electrode assemblies (MEA) of proton exchange membrane fuel cell (PEMFC), it is important to study the connection among catalyst agglomerates, CL structure, and MEA performance. This study investigates the effect of Pt/GC catalyst agglomerates on the morphology and permeability of the CL by modulating the properties of the catalyst ink in two different ways. Additionally, MEA was further electrochemically tested to understand the relationship between the catalyst agglomerates and MEA performance. The result shows that High-pressure homogenization is more effective than mechanical shear mixing in dispersing the agglomerates in catalyst inks. However, the excessive homogenization pressure produced larger agglomerated particles, probably because more effective dispersion caused by higher homogenization pressure supplies new chain carriers for polymerization and higher temperature caused by higher homogenization pressure. Moreover, the surface of the CL fabricated in inks prepared by a homogenizer is more uniform, neat, and hydrophilic. But the number of secondary pores in the catalyst layer decreases at excessive homogeneous pressure, and the water permeability becomes poor, which in turn result in lower performance and higher mass transfer resistance. The electrochemical performance test results showed that the MEA with a relatively hydrophobic CL had a performance of 0.707 V at 1000 mA cm⁻², which was 30 mV higher than that with a relatively hydrophilic CL. This study provides insights for better tuning the properties of catalyst ink, CL morphology, and permeability to obtain better performance of MEA.     

基于主成分分析的神经网络在需水预测中的应用

需水量预测是水资源规划与管理的重要内容,建立高精度的需水量预测模型,可为区域水资源规划提供技术支撑。以南昌市为例,利用2005~2015年的实际用水量数据,对南昌市16个用水总量影响因子进行主成分分析,以提取的主成分作为BP神经网络的输入样本,建立BP神经网络需水预测模型。结果表明,经过主成分分析,16个用水量影响因子可用8个主要影响因子代替。建立的BP神经网络模型的需水预测平均相对误差仅为1.37%,预测精度较高,可作为区域需水预测的一种可靠方法。研究成果可为类似区域提供借鉴。     

Self-humidifying membrane electrode assembly prepared by adding PVA as hygroscopic agent in anode catalyst layer

In this work, a novel self-humidifying membrane electrode assembly (MEA) with addition of polyvinyl alcohol (PVA) as the hygroscopic agent into anode catalyst layer was developed for proton exchange membrane fuel cell (PEMFC). The MEA shows good self humidification performance, for the sample with PVA addition of 5 wt.% (MEA PVA5), the maximum power density can reach up to 623.3 mW·cm⁻², with current densities of 1000 mA·cm⁻² at 0.6 V and 600 mA·cm⁻² at 0.7 V respectively, at 50 °C and 34% of relative humidity (RH). It is interesting that the performance of MEA PVA5 hardly changes even if the relative humidity of both the anode and cathode decreased from 100% to 34%. The MEA PVA5 also shows good stability at low humidity operating conditions: keeping the MEA discharged at constant voltage of 0.6 V for 60 h at 34% of RH, the attenuation of the current density is less than 10%, whilst for the MEA without addition of PVA, the attenuation is high up to 80% within 5 h.     

An effective layout of polyaniline nanofibers incorporated in membrane-electrode assembly as methanol transport regulator for direct methanol fuel cells

This study reports a novel strategy by using polyaniline nanofibers (PANFs) to modify membrane-electrode assembly (MEA) for improving direct methanol fuel cell (DMFC) performance. First of all, a series of PANFs emeraldine salt was synthesized and characterized. Then, we investigated the effect of PANFs layout in MEA on DMFC performance. Three different placements to incorporate the as-synthesized PANFs in anodes include (1) placing a layer of PANFs between catalyst layer (CL) and proton exchange membrane (PEM), (2) mixing with catalyst slurry and coating onto gas diffusion layer (GDL), and (3) placing a layer of PANFs between CL and GDL. Polarization curves indicate that the third method is superior to the others and is adopted as the incorporation layout thereafter. Both methanol transport resistance and methanol crossover of the PANFs-modified MEA are studied further. The DMFC incorporated with H₂SO₄-doped PANFs obtained after the re-doping process with 2 mol L⁻¹ H₂SO₄ performs a power density as high as 53 mW cm⁻², about 20% higher than that of the pristine one without PANFs incorporation. However, an excessive doping level may result in a higher methanol transport resistance due to PANFs aggregation and thus deteriorate DMFC performance. This study provides a simple and effective way by placing a layer of PANFs between CL and GDL in anode to act as methanol transport regulator and improve DMFC performance consequently.     

A novel anode catalyst layer with multilayer and pore structure for improving the performance of a direct methanol fuel cell

A novel anode catalyst layer (CL) has been prepared by ultrasonic‐spray process which combines directly spraying method and catalyst‐coated membrane switchover method, and heated‐stereoscopic process has been used to enhance bond force between CLs and proton exchange membrane in this paper. The scanning electron microscopy, electrochemical impedance spectra and polarization curves show that: the anode outer CL with pores and meshwork structure has increased the electrochemical active surface area and retained the transfer of protons and electrons, and the anode inner CL with compact structure has prevented methanol crossover. And the gradient catalysis for methanol electrochemical catalytic oxidation reaction has been achieved. The open circuit voltage has reached 0.697 V, and the performance has increased from 116.8 mW cm^?2 of traditional membrane electrode assembly (MEA) to 202.6 mWcm^?2 of novel MEA at 80°C. Copyright © 2012 John Wiley & Sons, Ltd.     

Investigation of water accumulation and discharge behaviors with variation of current density in PEMFC by high-resolution soft X-ray radiography

The management of water is a challenging issue to achieving high power density, long-term operation, and increased robustness in PEMFCs. Development of in-situ diagnostic techniques to understand the dynamic behavior of liquid water is needed. In this study, visualization of liquid water across the membrane electrolyte assembly (MEA) of operating PEMFCs was performed by using high-resolution soft X-ray radiography, with which was possible to distinguish the catalyst layer (CL), polymer electrolyte membrane, and gas diffusion layer (GDL) of the MEA. Dynamic liquid water transport behavior in the cathode CL and GDL was observed at high spatial and temporal resolution (1 frame/s). At low current density, generation of liquid water was recognized under the rib, while at high current density, liquid water was observed both under the rib and the channel. Temporal behaviors of liquid water accumulation in GDL under the rib and discharge process of liquid water to under the channel were also visualized.     

Characteristics of water transport through the membrane in direct methanol fuel cells operating with neat methanol

The water required for the methanol oxidation reaction in a direct methanol fuel cell (DMFC) operating with neat methanol can be supplied by diffusion from the cathode to the anode through the membrane. In this work, we present a method that allows the water transport rate through the membrane to be in-situ determined. With this method, the effects of the design parameters of the membrane electrode assembly (MEA) and operating conditions on the water transport through the membrane are investigated. The experimental data show that the water flux by diffusion from the cathode to the anode is higher than the opposite flow flux of water due to electro-osmotic drag (EOD) at a given current density, resulting in a net water transport from the cathode to the anode. The results also show that thinning the anode gas diffusion layer (GDL) and the membrane as well as thickening the cathode GDL can enhance the water transport flux from the cathode to the anode. However, a too thin anode GDL or a too thick cathode GDL will lower the cell performance due to the increases in the water concentration loss at the anode catalyst layer (CL) and the oxygen concentration loss at the cathode CL, respectively.     

Nonlinear model identification of wind turbine with a neural network

A nonlinear model of wind turbine based on a neural network (NN) is described for the estimation of wind turbine output power. The proposed nonlinear model uses the wind speed average, the standard deviation and the past output power as input data. An anemometer with a sampling rate of one second provides the wind speed data. The NN identification process uses a 10-min average speed with its standard deviation. The typical local data collected in September 2000 is used for the training, while those of October 2000 are used to validate the model. The optimal NN configuration is found to be 8-5-1 (8 inputs, 5 neurons on the hidden layer, one neuron on the output layer). The estimated mean square errors for the wind turbine output power are less than 1%. A comparison between the NN model and the stochastic model mostly used in the wind power prediction is done. This work is a basic tool to estimate wind turbine energy production from the average wind speed.     

A novel NN based rotor flux MRAS to overcome low speed problems for rotor resistance estimation in vector controlled IM drives

This paper presents a new neural network based model reference adaptive system (MRAS) to solve low speed problems for estimating rotor resistance in vector control of induction motor (IM). The MRAS using rotor flux as the state variable with a two layer online trained neural network rotor flux estimator as the adaptive model (FLUX-MRAS) for rotor resistance estimation is popularly used in vector control. In this scheme, the reference model used is the flux estimator using voltage model equations. The voltage model encounters major drawbacks at low speeds, namely, integrator drift and stator resistance variation problems. These lead to a significant error in the estimation of rotor resistance at low speed. To address these problems, an offline trained NN with data incorporating stator resistance variation is proposed to estimate flux, and used instead of the voltage model. The offline trained NN, modeled using the cascade neural network, is used as a reference model instead of the voltage model to form a new scheme named as “NN-FLUXMRAS.” The NN-FLUX-MRAS uses two neural networks, namely, offline trained NN as the reference model and online trained NN as the adaptive model. The performance of the novel NN-FLUX-MRAS is compared with the FLUX-MRAS for low speed problems in terms of integral square error (ISE), integral time square error (ITSE), integral absolute error (IAE) and integral time absolute error (ITAE). The proposed NN-FLUX-MRAS is shown to overcome the low speed problems in Matlab simulation.     

Failure of PEM water electrolysis cells: Case study involving anode dissolution and membrane thinning

Polymer electrolyte membrane (PEM) water electrolysis is an efficient and environmental friendly method that can be used for the production of molecular hydrogen of electrolytic grade using zero-carbon power sources such as renewable and nuclear. However, market applications are asking for cost reduction and performances improvement. This can be achieved by increasing operating current density and lifetime of operation. Concerning performance, safety, reliability and durability issues, the membrane-electrode assembly (MEA) is the weakest cell component. Most performance losses and most accidents occurring during PEM water electrolysis are usually due to the MEA. The purpose of this communication is to report on some specific degradation mechanisms that have been identified as a potential source of performance loss and membrane failure. An accelerated degradation test has been performed on a MEA by applying galvanostatic pulses. Platinum has been used as electrocatalyst at both anode and cathode in order to accelerate degradation rate by maintaining higher cell voltage and higher anodic potential that otherwise would have occurred if conventional Ir/IrO_x catalysts had been used. Experimental evidence of degradation mechanisms have been obtained by post-mortem analysis of the MEA using microscopy and chemical analysis. Details of these degradation processes are presented and discussed.     

Investigation of a cost-effective strategy for polymer electrolyte membrane fuel cells: High power density operation

Polymer electrolyte membrane (PEM) fuel cell technology needs to overcome the cost barrier in order to compete with the internal combustion engines (ICEs) for transportation application. A viable approach is to raise fuel cell's power output without increasing its size and Pt loading in the catalyst layers (CLs). In this strategy, the cost per kW power output can be proportionally reduced due to the increased power density. This paper examines this strategy by exploring several important aspects that influence fuel cell performance under high power or current density using a three-dimensional (3-D) fuel cell model. It is shown that local CLs may be subject to low oxygen concentration under a high current density of 2 A/cm², causing low reaction rate near the outlet, especially under the land. Additionally, the oxygen reduction reaction (ORR) rate may be subject to a large through-plane variation under 2 A/cm², raising ohmic voltage loss in the CL. Two additional cases are investigated to improve fuel cell performance under 2 A/cm²: one has a 5 times thinner CL with the same ORR kinetics per membrane electrode assembly (MEA) area and the other has a 5 times thinner CL with 5 times higher ORR kinetics. The results show the output voltage is raised approximately from 0.5 V to 0.554 V in the former CL case and further to 0.606 V for the latter CL. To enable high-efficiency operation (e.g. >50%), thinner CLs with high ORR kinetics and GDLs with better transport properties are one research and development (R&D) direction.     

基于KG-BP神经网络在秦淮河洪水水位预测中的应用

为了提高BP神经网络模型的预测精度,提出了一种基于KNN算法及GA算法优化的BP神经网络的水位预测方法(KG-BP),即通过KNN邻近算法从全样本数据中剔除与待测点相关度较低的样本集,并允许保留K个"优质"训练数据集;将筛选出的"优质"训练数据集代入GA算法中实现初始权阈值的优化;再将"优质"的样本和初始权阈值代入BP模型中进行训练。将该预测方法应用于东山站水位实际预测中,并与BP模型、GA-BP模型的预测结果进行对比分析,验证了KG-BP模型具有较高的预测精度。     

考虑预报因子选择的神经网络降雨径流模型

为优化神经网络模型的应用效果,研究了基于神经网络的降雨-径流模型,根据Copula熵法确定预报因子,并与传统的线性相关法进行比较分析,采用BP、RBF、GRNN三种神经网络建立降雨-径流模型,应用均方根误差、合格率、确定性系数三个指标为模型选取评价准则。通过对金沙江流域的径流预报,发现基于Copula熵法的BP模型预报结果更接近实测值,精度更高。     

Wind Farms as Reactive Power Ancillary Service Providers—Technical and Economic Issues

This paper examines the possibility of providing reactive power support to the grid from wind farms (WFs) as a part of the ancillary service provisions. Detailed analysis of the WF capability curve is carried out considering maximum hourly variation of wind power from the forecasted value. Different cost components are identified, and subsequently, a generalized reactive power cost model is developed for wind turbine generators that can help the independent system operator (ISO) in managing the system and the grid efficiently. Apart from the fixed cost and the cost of loss components, a new method is proposed to calculate the opportunity cost component for a WF considering hourly wind variations. The CigrÉ 32-bus test system is used to demonstrate a case study showing the implementation of the developed model in short-term system operations. A finding is that higher wind speed prediction errors (a site with high degree of wind fluctuations) may lead to increased payments to the WFs for this service, mainly due to the increased lost opportunity cost (LOC) component. In a demonstrated case, it is found that 2340 $/h is paid to the WF as the LOC payment only, when the wind prediction error is 0.5 per unit (p.u.), whereas 54 $/h is the expected total payment to the WF when the prediction error is 0.2 p.u. for its reactive power service.      

基于BP神经网络的混凝土抗硫酸盐侵蚀预测研究

鉴于新拌混凝土流变参数与硫酸盐侵蚀劣化指标之间存在的非线性映射关系,基于新拌混凝土流变参数,利用BP神经网络建立了硫酸盐干湿循环侵蚀后混凝土抗压强度损失率、质量损失率及相对动弹性模量的预测模型,预测了混凝土受硫酸盐侵蚀程度。结果表明,BP神经网络模型预测的硫酸盐侵蚀结果与实测结果吻合较好,且预测误差较小,说明BP神经网络模型对混凝土受硫酸盐侵蚀后的劣化程度具有良好的预测效果。     

Gradation of mechanical properties in gas-diffusion electrode. Part 2: Heterogeneous carbon fiber and damage evolution in cell layers

In PEM fuel cell, gas-diffusion electrode (GDE) plays very significant role in force transmission from bipolar plate to the membrane. This paper investigates the effects of geometrical heterogeneities of gas-diffusion electrode layer (gas-diffusion layer (GDL) and catalyst layer (CL)) on mechanical damage evolution and propagation. We present a structural integrity principle of membrane electrode assembly (MEA) based on the interlayer stress transfer capacity and corresponding cell layer material response. Commonly observable damages such as rupture of hydrophobic coating and breakage of carbon fiber in gas-diffusion layer are attributed to the ductile to brittle phase transition within a single carbon fiber. Effect of material inhomogeneity on change in modulus, hardness, contact stiffness, and electrical contact resistance is also discussed. Fracture statistics of carbon fiber and variations in flexural strength of GDL are studied. The damage propagation in CL is perceived to be influenced by the type of gradation and the vicinity from which crack originates. Cohesive zone model has been proposed based on the traction-separation law to investigate the damage propagation throughout the two interfaces (carbon fiber/CL and CL/membrane).     

基于AP-TS-SVR模型的热电厂工业热负荷预测

工业热负荷预测对保障工业生产过程安全和提高热电厂经济性具有重要意义,然而工业热负荷影响因素众多且有较强的随机性,因而采用常规单一负荷预测方法预测难度较大。建立了一种通过平均比例法将时间序列与支持向量机相结合的AP-TS-SVR混合模型,并对某热电厂工业热负荷进行了预测。结果表明,在预测的72h(3个预测日)中,只有6h的预测相对误差大于20%,单日平均相对误差最大为10.64%;与时间序列、支持向量机以及神经网络等单一方法的预测结果相比,AP-TS-SVR混合模型的预测结果较优。AP-TS-SVR混合模型仅需输入工业用户历史负荷数据即可进行热电厂工业热负荷短期预测,对热电厂按需供热具有重要的指导意义。