Multi-spectral remote sensing images feature coverage classification based on improved convolutional neural network

Yufeng Li; Chengcheng Liu; Weiping Zhao; Yufeng Huang; Yufeng Li; Chengcheng Liu; Weiping Zhao; Yufeng Huang

doi:10.3934/mbe.2020245

Mathematical Biosciences and Engineering

2020, Volume 17, Issue 5: 4443-4456. doi: 10.3934/mbe.2020245

Previous Article Next Article

Research article Special Issues

Multi-spectral remote sensing images feature coverage classification based on improved convolutional neural network

Key Laboratory of Aerospace Information Perception and Intelligent Processing, College of Electronic and Information Engineering, Shenyang Aerospace University, Shenyang 110122, China

Received: 03 April 2020 Accepted: 14 June 2020 Published: 23 June 2020

With the continuous development of the earth observation technology, the spatial resolution of remote sensing images is also continuously improved. As one of the key problems in remote sensing images interpretation, the classification of high-resolution remote sensing images has been widely concerned by scholars at home and abroad. With the improvement of science and technology, deep learning has provided new ideas for the development of image classification, but it has not been widely used in remote sensing images processing. In the background of remote sensing huge data, the remote sensing images classification based on deep learning proposed in the study has more research significance and application value. The study proposes a high-resolution remote sensing images classification method based on an improved convolutional neural network. The traditional convolutional neural network framework is optimized and the initial structure is added. The actual classification results of radial basis functions and support vector machine are compared horizontally. The classification results of hyperspectral images were presented that the improved method can perform better in overall accuracy and Kappa coefficient. The commission errors of support vector machine classification method are more than 6 times of that of the improved convolutional neural network classification method and the overall accuracy of the improved convolutional neural network classification method has reached 97% above.

Keywords:

Citation: Yufeng Li, Chengcheng Liu, Weiping Zhao, Yufeng Huang. Multi-spectral remote sensing images feature coverage classification based on improved convolutional neural network[J]. Mathematical Biosciences and Engineering, 2020, 17(5): 4443-4456. doi: 10.3934/mbe.2020245

Related Papers:

[1]	Thomas Wetzel, Julio Pacio, Luca Marocco, Alfons Weisenburger, Annette Heinzel, Wolfgang Hering, Carsten Schroer, Georg Muller, Jurgen Konys, Robert Stieglitz, Joachim Fuchs, Joachim Knebel, Concetta Fazio, Markus Daubner, Frank Fellmoser . Liquid metal technology for concentrated solar power systems: Contributions by the German research program. AIMS Energy, 2014, 2(1): 89-98. doi: 10.3934/energy.2014.1.89
[2]	Espen Olsen, Maria Hansen, Heidi S. Nygård . Hydrolysis of molten CaCl₂-CaF₂ with additions of CaO. AIMS Energy, 2017, 5(6): 873-886. doi: 10.3934/energy.2017.6.873
[3]	Joanna McFarlane, Jason Richard Bell, David K. Felde, Robert A. Joseph III, A. Lou Qualls, Samuel Paul Weaver . Performance and Thermal Stability of a Polyaromatic Hydrocarbon in a Simulated Concentrating Solar Power Loop. AIMS Energy, 2014, 2(1): 41-70. doi: 10.3934/energy.2014.1.41
[4]	Kokou Aménuvéla Toka, Yawovi Nougbléga, Komi Apélété Amou . Optimization of hybrid photovoltaic-thermal systems integrated into buildings: Impact of bi-fluid exchangers and filling gases on the thermal and electrical performances of solar cells. AIMS Energy, 2024, 12(5): 1075-1095. doi: 10.3934/energy.2024051
[5]	Hong-Yu Pan, Chuang Sun, Xue Chen . Transient thermal characteristics of infrared window coupled radiative transfer subjected to high heat flux. AIMS Energy, 2021, 9(5): 882-898. doi: 10.3934/energy.2021041
[6]	Mowffaq M. Oreijah, Mohammed Yunus . A parametric analysis to evaluate the performance metrics of power generation system involving Trilateral Flash Cycle using three different working fluids for low grade waste heat. AIMS Energy, 2019, 7(4): 483-492. doi: 10.3934/energy.2019.4.483
[7]	Eman S. Maayah, Hamzeh M. Duwairi, Banan Maayah . Analytical model of solar energy storage using non—Newtonian Fluid in a saturated porous media in fully developed region: carboxymethyl cellulose (CMC) and graphite model. AIMS Energy, 2021, 9(2): 213-237. doi: 10.3934/energy.2021012
[8]	Khaled M. Bataineh, Assem N. AL-Karasneh . Direct solar steam generation inside evacuated tube absorber. AIMS Energy, 2016, 4(6): 921-935. doi: 10.3934/energy.2016.6.921
[9]	Khadim Ndiaye, Stéphane Ginestet, Martin Cyr . Thermal energy storage based on cementitious materials: A review. AIMS Energy, 2018, 6(1): 97-120. doi: 10.3934/energy.2018.1.97
[10]	Wei An, Yifan Zhang, Bo Pang, Jun Wu . Synergistic design of an integrated pv/distillation solar system based on nanoﬂuid spectral splitting technique. AIMS Energy, 2021, 9(3): 534-557. doi: 10.3934/energy.2021026

Abstract

1. Introduction

The use of various sources of renewable energy becomes increasingly important in the worldwide effort of ameliorating problems associated with the use of fossil fuels. No doubt that solar energy is one of the most attractive sources of clean energy. Among different solar energy technologies,concentrated solar thermal power(CSP)is believed to be very promising for large-capacity power generation. Due to the possibility of incorporating solar thermal storage into the power generation system,CSP can better meet the power dem and at the time sun is down [1,2,3,4].

In the past decade we have seen a worldwide significant increase in solar thermal power generation capacity with the combination of solar thermal storage. The number of solar thermal power plants constructed in Southern Europe,USA,Africa,and Australia is increasing,which generate electricity by replacing conventional fuel-combustion facilities in power plants with solar thermal collection devices. Solar concentrators collect thermal energy and raise the temperature of the heat transfer fluid(HTF),which is fed to a heat exchanger and boils water into steam of high temperature and high pressure. The steam subsequently drives steam turbines,which generate electricity. There are four types of solar concentration technologies,parabolic trough,solar power tower,Fresnel reflectors,and solar dish Stirling engines. The first three use a HTF to take away the heat from the collectors and use the heat in power generation systems.

Heat transfer and heat storage are the two important roles of a HTF in concentrated solar power systems. The well-known HTF used in CSP systems in the earlier stage is made of organic substances,which is an eutectic mixture of biphenyl(C₁₂H₁₀) and diphenyl oxide(C₁₂H₁₀O),sold under the br and name of Therminol VP-1 and Dowtherm A [5,6]. It exhibits a low melting point of 12 °C(285 K)but is limited to an upper temperature of 390 °C(673 K)due to chemical dissociation above this temperature. This is not sufficiently high for the increasing dem and of thermal efficiency of a CSP system. The high vapor pressure(10 atm at 390 °C) and high cost of this HTF also significantly restrict its application.

Increasing the high temperature in a CSP plant from 390 ℃ to 500 ℃(using molten salts)would increase the Rankine cycle efficiency to the 40% range(compared to the efficiency of 37.6% using Therminol VP-1) and thereby reducing the levelized electricity cost by 2 cents/kWh [7,8]. For more development of CSP technologies,finding a heat transfer fluid working at much high temperatures is important.

There are multiple dem and s if a fluid serves as a heat transfer fluid in a large range of temperature variation. First the fluid should have a low freezing point to avoid solidification in the circulation system. Second,the fluid should be chemically stable at a high temperature,and at the same time to have low vapor pressures(below 1.0 atm)due to the safety requirement of containers and pipes. Third,the fluid must have minimum corrosion to the metal pipes and containers that hold the fluids. Forth,the heat transfer fluid should have favorable transport properties(low viscosity and high thermal conductivity)for efficient heat exchange and low pressure loss in the flow and circulation.

Molten salts have been studied for their possibility of high working temperatures,and in the meantime with low melting points,moderate density,high heat capacity,and high thermal conductivity. Other than favorable thermal and transport properties,long term thermal stability(or chemical stability with less corrosion to containers) and low cost of molten salt HTF is also very critical [9,10,11,12].

Due to the very strong dem and ,the research work for suitable molten salts for HTFs as well as thermal energy storage materials in solar thermal power plants [13,14] is very active recently. The studies on some inorganic salts used for thermal energy storage materials are available in several papers [15,16,17,18,19]. Zalba et al. [16] summarized thermal characteristics of some phase-change materials. Kenisarin [17] and Gil et al. [18,19] analyzed some phase change materials and the practical applications for solar thermal storage. So far,several well-recognized commercial molten salts by eutectic mixtures of nitrates or nitrites have been used in concentrated solar power systems mainly for thermal storage purpose,but can also be used as HTF. A binary mixture,Solar Salt(60 wt.% NaNO₃/ 40 wt.% KNO₃)has been used as thermal storage material in the 10 MWe solar-II central receiver project in California [20,21,22],in the 2-tank direct system of the Archimede project in

Italy [23],and the indirect thermal energy storage system for the Andasol plant in Spain [24,25].Solar Salt has a thermal stability(below 600 ℃) and a relatively high melting point(220 ℃). A new heat transfer fluid called Hitec,which is a ternary salt mixture of 53 wt.% KNO₃/ 7wt.% NaNO₃/

40 wt.% NaNO₂,has been considered to replace the Solar Salt because of its low freezing point of

142 ℃ [26]. Hitec is thermally stable at temperatures up to 454 ℃,and may be used at temperature up to 538 ℃ for a short period [27]. A modified version,Hitec XL,is a mixture of 48 wt.% Ca(NO₃)₂/ 7 wt.% NaNO₃/ 45 wt.% KNO₃ which melts at about 133 ℃ and may be used at a temperature up to 500 ℃ [28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43]. Different compositions of Ca(NO₃)₂/ NaNO₃/ KNO₃ have been identified in the open literature as eutectic salts [29,30,31,32,33,34,35,36,37,38,39,40,41,42,43]. The ternary eutectic salt with composition of 44 wt.% Ca(NO₃)₂/ 12 wt.% NaNO₃/ 44 wt.% KNO₃ melts at 127.6 ℃ and its thermal stability is good at up to 622 ℃ [43]. Different phase diagrams also have been published for the Ca(NO₃)₂/ NaNO₃/ KNO₃ system [44,45,46,47].

There are also other ternary salts being developed and undergoing tests for industrial application. The eutectic salt in composition of 25.9 wt.% LiNO₃/ 20.0 wt.% NaNO₃/ 54.1 wt.% KNO₃ melts at about 118 ℃ and thermal stability is up to 435 ℃ [48,49]. The salt in composition of 30wt.% LiNO₃/ 18 wt.% NaNO₃/ 52 wt.% KNO₃ is reported to have thermal stabilities up to 550 ℃ and melting point of 120 ℃[50,51,52]. Another ternary nitrate salt mixtures consisting of 50-80 wt.% KNO₃/ 0-25 wt.% LiNO₃/ 10-45 wt.% Ca(NO₃)₂ melts below 100 ℃ and thermal stability is up to

500℃ [53]. S and ia National Laboratories developed a low-melting heat transfer fluid made of a mixture of four inorganic nitrate salts: 9-18 wt.%NaNO₃/ 40-52 wt.%KNO₃/ 13-21 wt.%LiNO₃/20-27 wt.% Ca(NO₃)₂ [54]. This quaternary salts mixture has a melting temperature less than 100 ℃ and the thermal stability limit is greater than 500 ℃. The quaternary salt 17.77wt.% LiNO₃/ 15.28 wt.% NaNO₃/ 35.97wt.% KNO₃/ 30.98 wt.% 2KNO₃·Mg(NO₃)₂ melts at about 100.9 ℃ but there is no data about thermal stability [55]. Another quaternary nitrate salt mixture consisting of 17.5 wt.% LiNO₃/ 14.2 wt.% NaNO₃/ 50.5 wt.% KNO₃/ 17.8 wt.% NaNO₂ has a melting point of 99 ℃ and thermal stability at up to 500 ℃ [56]. Eutectic mixture with five species of salts also has been developed,for example,the salt with compositions of 6wt.% NaNO₃/ 23 wt.% KNO₃/ 8 wt.% LiNO₃/ 19 wt.% Ca(NO₃)₂/ 44 wt.% CsNO₃ melts at 65 ℃ and has thermal stability of up to 561 ℃ [29]. Nevertheless,nitrate and nitrite salt systems have been found not thermally stable at temperature above 600 ℃ from a large amount of research works. A summary of the above mentioned HTFs is in Table 1.

Table 1. Summary of the key data of some heat transfer fluids.

Name	Formula	T_melt(℃)	T_max(℃)
Therminal VP-1	(C₁₂H₁₀) and (C₁₂H₁₀O). Percentage not know.	12	390
Solar Salt	wt. 60% NaNO₃ /40% KNO₃	220	600
Hitec	wt. 53% KNO₃/7% NaNO₃/40% NaNO₂	142	454-538
Hitec XL	wt. 48% Ca(NO₃)₂/7% NaNO₃/45% KNO₃	133	500
NS-1	wt. 44% Ca(NO₃)₂/12% NaNO₃/44% KNO₃	127.6	622
NS-2	wt. 25.9% LiNO₃/20.0% NaNO₃/54.1% KNO₃	118	435
NS-3	wt. 30% LiNO₃/18% NaNO₃/52% KNO₃	120	550
NS-4	wt. 50-80% KNO₃/0-25% LiNO₃/10-45% Ca(NO₃)₂	100	500
NS-5	wt. 17.77% LiNO₃/15.28% NaNO₃/35.97% KNO₃/ 30.98% 2KNO_3·Mg(NO₃)₂	100
NS-6	wt. 17.5% LiNO₃/14.2% NaNO₃/50.5% KNO₃/ 17.8% NaNO₂	99	500
NS-7	wt. 6% NaNO₃/23% KNO₃/8% LiNO₃/ 19% Ca(NO₃)₂/44% CsNO₃	65	561

| Show Table

DownLoad: CSV

A target of high temperature at 800 ℃ for CSP has been proposed by US Department of Energy. This is possible to accomplish Brayton power cycle,which can further increase the heat-to-electricity efficiency. Therefore,it is proposed recently to replace molten nitrate-nitrite salt with other kinds of salts to possibly increase the applicable temperatures to the level of 800 ℃ and even up to 1000 ℃. Hundreds of inorganic salts and salt composites for HTF and latent heat storage in the temperature ranging from 120 ℃ to 1000 ℃ are listed in Kenisarin’s review paper [57]. Those materials are on the basis of chlorides,fluorides,bromides,hydroxides,nitrates,carbonates and other salts. He found out that almost no single inorganic salt possesses decent properties to serve as a qualified HTF. Binary and ternary eutectic compositions based on fluorides and chlorides are the most prospective materials in term of their possibly favorable thermal and transport properties,as well as reasonably low cost in particular.

The eutectics of fluoride salts have been utilized in space solar power and molten salt nuclear reactors because of their favorable thermal and transport properties,especially heat storage capacity,but with the disadvantage of high cost,material compatibility and toxicity [58,59,60]. Carbonates may also be used for high temperature HTF and latent heat storage materials,but with drawbacks of high viscosity and easy degradation [61,62]. Consequently,chlorides salts are attractive due to their possibly favorable properties and especially low cost [63].

Funded by U.S. Department of Energy,a team by researchers from the University of Arizona,Georgia Institute of Technology,and Arizona State University has been conducting studies to ternary,quaternary and even higher order eutectic salts based on five key species of halide salts—AlCl₃,ZnCl₂,FeCl₃,NaCl,and KCl. These species are relatively inexpensive and also have great amount of reserve on the earth. The mixing of these ionic and covalent salts is expected to create favorable properties needed for HTF.

It is underst and able that ionic and covalent halide salts are different in molecule size,shape,and chemical bonding. The bonding of positive ion and negative ion can make disorder leading to eutectic mixture with low melting temperatures. In Figure 1,strong evidences of low melting points at some eutectic compositions are shown in the phase diagrams of some binary and ternary mixtures by halide ionic salt with covalent salt(NaCl-AlCl₃,KCl-AlCl₃,NaCl-ZnCl₂,KCl-ZnCl₂,NaCl-KCl-AlCl₃,NaCl-KCl-ZnCl₂). Although very promising,a significant amount of work needs to be conducted to fully underst and all the properties of these salts mixtures.

Figure 1. Phase diagrams of binary and ternary mixtures of ionic and covalent halide salts [64,65].

DownLoad: Full-Size Img PowerPoint

Obviously,identifying the eutectic compositions that have low melting points is only the first step of developing a HTF. As the final goal,a HTF should meet the target of thermal and transport properties in a relatively wide temperature range from below 250 ℃ to at least above 800 ℃. To obtain the thermal and transport properties(vapor pressure,density,viscosity,specific heat,thermal conductivity)for a high order mixture(of ternary and quaternary components),the properties of all individual components as well as all the low-order salt mixtures have to be identified. For this purpose,the present paper reviews the currently available experimental data for density,viscosity,

thermal conductivity,specific heat capacity,vapor pressure,and melting point of several halide salt single species as well as their binary to ternary mixtures. These data are expected to serve as the basis for further work of developing high order eutectic mixtures with underst and ing of their thermal and transport properties.

2. Properties of Several Popular Single Halide Salts in Molten State

In this section,the thermal and transport properties of the five single species of molten salt(AlCl₃,ZnCl₂,FeCl₃,NaCl,KCl)are provided for reference and evaluation on whether a salt can contribute to a better property of a possible eutectic salt mixture.

2.1. Aluminum chloride(AlCl₃)

This salt has a relatively low melting point which is T_m = 465 K(192 ℃)[66]. However,it sublimes early at a temperature of 180 ℃. The surveyed properties for molten AlCl₃ [66] are shown in Figure 2. The properties as functions of temperatures are given by the following equations:

$\rho = 3.7660038 - 1.3346 \times {10^{ - 2}}T + 2.7622 \times {10^{ - 5}}{T^2} - 2.2331268 \times {10^{ - 8}}{T^3}$

(1)

Figure 2. Properties for molten salt AlCl₃ at different temperatures.

DownLoad: Full-Size Img PowerPoint

where the units are p(g/cm³),T(K)in the range of 465-560 K.

$\mu = 3.2146 - 9.6606 \times {10^{ - 3}}T + 7.4554 \times {10^{ - 6}}{T^2}$

(2)

where the unit of viscosity is 10^-3(Pa s),and temperature is in K,in the range of 470-560K. The low viscosities may make AlCl₃ a very good component for a eutectic salt mixture as the low viscosity is important to a HTF.

${P_{vap}} = {10^{(7.42055 - 1948.55/T)}}$

(3)

where the unit of pressure is P_vap(mm*Hg = 133.32 Pa),and the temperature is in K,in the range of 455-530K.

The s pecific heat capacity of liquid AlCl₃ does not change greatly in a wide range of temperatures from466-1500K,which is around C_p=125.5 J/mol*K or equivalent to 941.2J/kg*K [67].

The vapor pressures of AlCl₃ are quite high which are not favorable as a heat transfer fluid. It is thus expected that ionic salts in a eutectic mixture with AlCl₃ can create inter-molecular bonding to suppress the vapor pressure. There are no thermal conductivity data found for liquid AlCl₃ in the current survey.

2.2. Zinc chloride(ZnCl₂)

Due to the low melting point and relatively low vapor pressure(compared to that of AlCl₃),ZnCl₂ is a very important species in halide salt family to contribute to a low melting point in a eutectic salt. The melting point of ZnCl₂ is T_m = 556 K(283 ℃)[68]. Other properties are summarized as follows.

The density,viscosity,and vapor pressure of molten ZnCl₂ versus temperatures are shown in Figure 3. The expressions for these properties as function of temperatures are given in Eqs.(4)-(6).

Figure 3. Properties for molten salt ZnCl₂.

DownLoad: Full-Size Img PowerPoint

$\rho = 2.424 - 0.00046 \times (T - 773)$

(4)

where the units are p(g/cm³),T(K),and the equations is applicable in a temperature range of 758.7-823.7 K.

$\ln (\mu ) = 0.2686 - 2665730.8 \times \ln (T)/{T^2} + 19840539/{T^2}$

(5)

where μ is in cp or 10^-3 Pa s,and T is in(K); the range of temperature is in 571-893K.

It is interesting to see from Figure 3(b)that the viscosity at low temperatures is high but it decreases dramatically when temperature increases. The low viscosity at high temperature is advantageous; while the high viscosity at low temperature should be modified by adding other components in a eutectic salt mixture.

The vapor pressure of liquid ZnCl₂ is relatively low compared to that of AlCl₃. As shown in Figure 3(c)the vapor pressure of ZnCl₂ is quite low at temperatures below 600 ℃,and at around 800 ℃ the vapor pressure reaches 2.0 atm. The correlation of vapor pressure and temperature reported by Keneshea and Cubicciotti [69] is in the form of

$\log p(mmHg) = ( - 8415.2/T) - 5.034\log T + 26.420$

(6)

where the unit of temperature is K,mm*Hg = 133.32 Pa. The relatively low vapor pressure of ZnCl₂ compared to that of AlCl₃ makes it a more favorable component to contribute to a relatively lower vapor pressure in a eutectic salt mixture.

The specific heat capacityin the temperature range of 591-973K is around 24.1 cal/mol*K[70],or 739.54 J/(kg K). The present review could not find thermal conductivities for molten ZnCl₂.

2.3. Iron(III)chloride(FeCl₃)

The melting point of salt FeCl₃ is T_m = 555 K(282 ℃)[68]. The specific heat capacity of molten FeCl₃ is Cp = 133.89 J/mol*K or 825.43 J/kg*K in the temperature range of 577-1500 K. This salt has a rather high vapor pressure as shown above in Figure 4. At about 340 ℃ the vapor pressure already reaches 2 atm [73]. For this salt to be a component in a eutectic salt mixture,its high vapor pressure can be a problem of concern. No other properties were found for molten FeCl₃ in this survey.

Figure 4. Vapor pressure of liquid FeCl₃ [73].

DownLoad: Full-Size Img PowerPoint

2.4. Sodium chloride(NaCl)

Well known as the table salt,NaCl has a melting point of T_m = 1075 K(802 ℃)[66]. NaCl also has an almost unlimited reserve in seawater and thus has relatively low cost. Properties of molten NaCl are surveyed and given in Figure 5.

Figure 5. Properties of molten NaCl against temperatures.

DownLoad: Full-Size Img PowerPoint

The correlations of the properties were given in Eqs.(7)to(9). For density there is

$\rho = 2.1393 - 5.430 \times {10^{ - 4}}T$

(7)

where the units are p(g/cm³) and T(K). The equation is applicable in the range of temperature of 1080-1290 K.

The expression of viscosity is

$\mu = 0.08931\exp (5248.1/RT)$ )

(8)

the unit of viscosity is μ(cp) and that of temperature is T(K),the range of temperatures for this equation is 1090-1200 K and the universal gas constant is R = 1.98716(cal/K*mol). The relatively low viscosity of molten NaCl is favorable for it being used as a component in a eutectic salt mixture,which needs to have low viscosities at high temperatures.

The expression of thermal conductivity is

$k = 1.868 \times {10^{ - 3}} + 4.73 \times {10^{ - 7}}T$

(9)

where k has unit of(cal/cm*s*K = 418.4 W/m-K),and T is in(K),for the range of 1100-1200 K.

The expression of specific heat capacity is

${C_p} = - 42.4478 + 113.526 \times t - 43.6466 \times {t^2} + 5.89663 \times {t^3} + 39.1386/{t^2}$

(10)

where unit of Cp is J/(mol*K),t = T(K)/1000,and T is in(K). The equation is applicable in a range of temperature from 1074 K to 2500 K.

The vapor pressure in a narrow temperature range for molten NaCl is expressed as:

${P_{vap}} = {10^{(8.4459 - 9565/T)}}$

(11)

where P_vap is in(mm*Hg = 133.32Pa),T is in(K),and the range of the temperature for the equation is 1250-1530 K. The low vapor pressure of molten NaCl is very favorable for a heat transfer fluid.

2.5. Potassium chloride(KCl)

Potassium chloride has a melting point of T_m = 1043 K(770 ℃)[66]. The properties surveyed for molten salt KCl are given in Figure 6. The corresponding expressions of the properties against temperatures for the curves are given in Eqs.(12)to(15).

Figure 6. Properties of molten KCl against temperatures.

DownLoad: Full-Size Img PowerPoint

For density,there is

$\rho = 2.1359 - 5.831 \times {10^{ - 4}}T$

(12)

which is applicable in a temperature range of 1060-1200 K,the unit of density is p(g/cm³) and for T is(K). For viscosity there is

$\mu = 0.0732\exp (5601.7/RT)$

(13)

where the unit of viscosity is μ(cp or 10^-3 Pa s),T(K)is temperature,and R = 1.98716(cal/K*mol). The equation is applicable in a range of 1070-1170 K. The relatively low viscosities of this molten salt make it a favorable component for a eutectic salt mixture. The thermal conductivity is

$k = - 23.43 \times {10^{ - 4}} + 4.103 \times {10^{ - 6}}T$

(14)

where k has unit of cal/(cm*s*K = 418.4 W/m-K),temperature T has unit of(K),and the equations is applicable in the range of1050-1200 K. The thermal conductivities of molten salt KCl are slightly lower than that of molten NaCl.

${P_{vap}} = {10^{(8.2800 - 9032/T)}}$

(15)

where P_vap(mm*Hg = 133.32Pa)is vapor pressure,and T(K)is temperature which is in the range of 1180-1530 K. Obviously the low vapor pressure of molten KCl is helpful for it being used in a eutectic salt mixture.

The specific heat capacity of molten KCl is around C_p = 73.6 J/mol*K = 987.24 J/kg*K in the temperature range of 1044-2000 K.

3. Properties of Binary Molten Salt

This section surveyed properties of binary molten salts NaCl + AlCl₃,KCl + AlCl_3. The compositions mentioned in the discussion are all based on mole fractions. The covalent salts AlCl₃ and ZnCl₂ both have low melting point,while the ionic salts NaCl and KCl both have rather high boiling points or high temperature stability. It is thus promising that the mixture of covalent and ionic eutectic salts has the potential to possess both low melting point and good stability at high temperatures.

3.1. Mixture of NaCl + AlCl₃

Eutectic melting pointfor NaCl + AlCl₃ is generally low. For example,the mixture in a mole fraction of 36.8% NaCl + 63.2% AlCl₃ has a melting point of T_m = 378-381 K(105-108 ℃)[66].

The densities of salt mixture NaCl + AlCl₃ in three compositions under molten states at different temperatures are given in Table 2,which are also plotted in Figure 7(a). The more NaCl is included in the mixture,the high the densities are.

Table 2.Density of molten salt NaCl + AlCl₃(% by mole) [66].

T(K)	p(kg/m³)
T(K)	27% NaCl + 73% AlCl₃	38.2% NaCl + 61.8% AlCl₃	48% NaCl + 52% AlCl₃
27%NaCl + 73%AlCl₃: $\rho = 2011 - 0.92 \times T$ ,Temperature Range: 460-610K(16) 38.2%NaCl + 61.8%AlCl₃: $\rho = 2034 - 0.866 \times T$ ,Temperature Range: 440-540K(17) 48%NaCl + 52%AlCl₃: $\rho = 2068 - 0.838 \times T$ ,Temperature Range: 400-560K(18) where p(kg/m³)is density and T(K)is temperature.
400			1733
410			1724
420			1716
430			1708
440		1653	1699
450		1644	1691
460	1588	1635	1682
470	1579	1627	1674
480	1570	1618	1666
490	1561	1609	1657
500	1551	1601	1649
510	1542	1592	1641
520	1533	1583	1632
530	1524	1575	1624
540	1515	1566	1615
550	1505		1607
560	1496		1598
570	1487
580	1478
590	1469
600	1459
610	1450

| Show Table

DownLoad: CSV

Viscosities of molten salt mixture NaCl + AlCl₃ in seven compositions are obtained from literature [66] as given in Table 3 and illustrated in Figure 7(b). The low viscosities are favorable for the eutectic molten salts being used as heat transfer fluids.

Figure 7. Properties of molten salt mixture NaCl+AlCl₃.

DownLoad: Full-Size Img PowerPoint

Table 3.Viscosity of liquid NaCl + AlCl₃(% by mole) [66] .

T(K)	μ(kg/m*s)
	50%NaCl	45%NaCl	40.01%NaCl	35.04%NaCl	30.08%NaCl	25.20%NaCl	20.28%NaCl
	50%NaCl + 50%AlCl₃ : $\mu = 7.2702 \times {10^{ - 6}}\exp (3285.3/RT)$ ,Range : 460-570 K(19) 45%NaCl + 55%AlCl₃ : $\mu = 6.8398 \times {10^{ - 6}}\exp (3413.5/RT)$ ,Range : 460-570 K(20) 40.01%NaCl + 59.99%AlCl₃ : $\mu = 5.7828 \times {10^{ - 6}}\exp (3661.1/RT)$ ,Range : 450-570 K(21) 35.04%NaCl + 64.96%AlCl₃ : $\mu = 4.9477 \times {10^{ - 6}}\exp (3850.2/RT)$ ,Range : 450-570 K(22) 30.08%NaCl + 69.92%AlCl₃ : $\mu = 4.2341 \times {10^{ - 6}}\exp (3966.7/RT)$ ,Range : 460-570 K(23) 25.20%NaCl + 74.80%AlCl₃ : $\mu = 3.6622 \times {10^{ - 6}}\exp (3977.5/RT)$ ,Range : 470-570 K(24) 20.28%NaCl + 79.72%AlCl₃ : $\mu = 2.8309 \times {10^{ - 6}}\exp (3985.8/RT)$ ,Range : 480-570 K(25) where μ(kg/ms)is viscosity,T(K)is temperature,and R = 1.98716(cal/Kmol).
50%AlCl₃	55%AlCl₃	59.99%AlCl₃	64.96%AlCl₃	69.92%AlCl₃	74.80%AlCl₃	79.72%AlCl₃
450			0.0003469	0.0003667
460	0.0002645	0.000286	0.0003174	0.0003339	0.0003246
470	0.000245	0.000264	0.0002914	0.0003053	0.000296	0.000259
480	0.0002277	0.000245	0.0002686	0.0002802	0.0002709	0.000237	0.0001848
490	0.0002123	0.000228	0.0002483	0.000258	0.0002489	0.0002177	0.0001697
500	0.0001984	0.000212	0.0002304	0.0002384	0.0002294	0.0002006	0.0001564
510	0.0001859	0.000199	0.0002143	0.000221	0.0002121	0.0001854	0.0001445
520	0.0001747	0.000186	0.0001999	0.0002054	0.0001967	0.000172	0.000134
530	0.0001645	0.000175	0.000187	0.0001915	0.000183	0.0001599	0.0001246
540	0.0001553	0.000165	0.0001753	0.0001789	0.0001707	0.0001491	0.0001162
550	0.0001469	0.000155	0.0001648	0.0001676	0.0001596	0.0001394	0.0001086
560	0.0001392	0.000147	0.0001552	0.0001574	0.0001496	0.0001306	0.0001017
570	0.0001322	0.000139	0.0001465	0.0001481	0.0001405	0.0001227	0.0000955

| Show Table

DownLoad: CSV

There is a very limited number of data for the thermal conductivity of NaCl + AlCl₃ [66]. For a mixture in mole fraction of 27%NaCl + 73%AlCl₃(with a melting point of 460 K)the liquid thermal conductivity at 467 K is 0.2217 W/(m K).

V apor pressures of NaCl + AlCl₃ in a dozen of different compositions are shown in Table 4 and Figure 7(c). The corresponding equations of the vapor pressures are also provided in the table.

Table 4.Vapor pressure of NaCl + AlCl₃(% by mole) [66] .

T(K)	P_vap(kPa)
T(K)	46.21% NaCl 53.79% AlCl₃	45.75% NaCl 54.25% AlCl₃	44.49% NaCl 55.51% AlCl₃	41.94% NaCl 58.06% AlCl₃	39.02% NaCl 60.98% AlCl₃	37.32% NaCl 62.68% AlCl₃	36.94% NaCl 63.06% AlCl₃	34.10% NaCl 65.90% AlCl₃	33.96% NaCl 66.04% AlCl₃	30.72% NaCl 69.28% AlCl₃	29.75% NaCl 70.25% AlCl₃	26.07% NaCl 73.93% AlCl₃
46.21%NaCl + 53.79%AlCl₃ : ${P_{vap}} = {10^{(4.71496 - 1771/T)}}/760 \times 101.325$ ,Range : 440-490K(26) 45.75%NaCl + 54.25%AlCl₃ : ${P_{vap}} = {10^{(5.94281 - 2304.5/T)}}/760 \times 101.325$ ,Range : 420-520K(27) 44.487%NaCl + 55.513%AlCl₃ : ${P_{vap}} = {10^{(5.77583 - 2177.5/T)}}/760 \times 101.325$ ,Range : 420-470K(28) 41.938%NaCl + 58.062%AlCl₃ : ${P_{vap}} = {10^{(6.72729 - 2416.6/T)}}/760 \times 101.325$ ,Range : 380-520K(29) 39.023%NaCl + 60.977%AlCl₃： ${P_{vap}} = {10^{(7.34912 - 2542.8/T)}}/760 \times 101.325$ ,Range : 410-520K(30) 37.323%NaCl + 62.677%AlCl₃ : ${P_{vap}} = {10^{(7.27205 - 2427.5/T)}}/760 \times 101.325$ ,Range : 410-520K(31) 36.954%NaCl + 63.046%AlCl₃ : ${P_{vap}} = {10^{(7.09260 - 2304.9/T)}}/760 \times 101.325$ ,Range : 410-520K(32) 34.096%NaCl + 65.904%AlCl₃ : ${P_{vap}} = {10^{(6.66296 - 1952.0/T)}}/760 \times 101.325$ ,Range : 430-520K(33) 33.964%NaCl + 66.036%AlCl₃ : ${P_{vap}} = {10^{(7.20848 - 2208.4/T)}}/760 \times 101.325$ ,Range : 430-520K(34) 30.723%NaCl + 69.277%AlCl₃ : ${P_{vap}} = {10^{(6.96901 - 1951.6/T)}}/760 \times 101.325$ ,Range : 440-520K(35) 29.745%NaCl + 70.255%AlCl₃ : ${P_{vap}} = {10^{(7.04376 - 1956.6/T)}}/760 \times 101.325$ ,Range : 450-480K(36) 26.071%NaCl + 73.929%AlCl₃ : ${P_{vap}} = {10^{(7.14703 - 1894.8/T)}}/760 \times 101.325$ ,Range : 460-520K(37) where P_vap(kPa)is vapor pressure and T(K)is temperature.
380				0.311
390				0.453
400				0.647
410				0.908	1.871	2.994	3.942
420		0.381	0.52	1.255	2.629	4.142	5.366
430		0.511	0.687	1.707	3.636	5.645	7.199	17.715	15.76
440	0.653	0.676	0.895	2.29	4.954	7.585	9.53	22.465	20.608	45.547
450	0.803	0.884	1.153	3.033	6.658	10.058	12.46	28.19	26.658	57.151	66.171
460	0.977	1.143	1.469	3.969	8.835	13.176	16.101	35.026	34.081	71.009	82.26	142.171
470	1.18	1.461	1.852	5.134	11.583	17.064	20.581	43.12	43.119	87.414	101.318	173.96
480	1.413	1.848		6.57	15.015	21.861	26.039	52.626	54.02	106.681	123.714	211.076
490	1.681	2.316		8.325	19.26	27.726	32.631	63.708	67.858	129.14		254.098
500		2.876		10.447	24.458	34.83	40.523	76.536	82.525	155.139		303.627
510		3.541		12.995	30.771	43.367	49.9	91.288	100.738	185.034		360.285
520		4.325		16.028	38.373	53.541	60.955	108.15	122.03	219.201		424.712

| Show Table

DownLoad: CSV

3.2. Mixture of KCl + AlCl₃

The melting point of this binary salt system is relatively low. For the mole composition of 33%KCl + 67%AlCl₃ there is T_m = 401 K(128 ℃)[66]. The densities for mixtures in four different compositions at different temperatures are shown in Table 5 and drawn in Figure 8(a)[66]. Vapor pressures of the mixtures at four different compositions are shown in Table 6 and Figure 8(b)[66].

Table .5Density of liquid KCl + AlCl₃(% by mole) [66] .

T(K)	p(kg/m³)
T(K)	20%KCl 80%AlCl₃	33.33%KCl 66.67%AlCl₃	50.03%KCl 49.97%AlCl₃	52.78%KCl 47.22%AlCl₃
20%KCl + 80%AlCl₃ : $\rho = 2025.2 - 1.0038 \times T$ ,Range : 480-540K(38) 33.33%KCl + 66.67%AlCl₃ : $\rho = 1988.9 - 0.7901 \times T$ ,Range : 500-780K(39) 50.03%KCl + 49.97%AlCl₃ : $\rho = 1955.6 - 0.6622 \times T$ ,Range : 740-1040K(40) 66.66%KCl + 33.34%AlCl₃ : $\rho = 1973.4 - 0.6101 \times T$ ,Range : 960-1040K(41) where $\rho$ (kg/m³)is density and T(K)is temperature.
480	1543
500	1523	1594
520	1503	1578
540	1483	1562
560		1547
580		1531
600		1515
620		1499
640		1483
660		1468
680		1452
700		1436
720		1420
740		1404	1466
760		1389	1452
780		1373	1439
800			1426
820			1413
840			1399
860			1386
880			1373
900			1360
920			1346
940			1333
960			1320	1388
980			1307	1376
1000			1293	1363
1020			1280	1351
1040			1267	1339

| Show Table

DownLoad: CSV

Table .6Vapor pressure of liquid KCl + AlCl₃ [66].

T(K)	P_vap(kPa)
T(K)	49.9%KCl + 50.01%AlCl₃	51.5%KCl + 48.5%AlCl₃	57.6%KCl + 42.4%AlCl₃	63.8%KCl + 36.2%AlCl₃
49.9%KCl + 50.01%AlCl₃: ${P_{vap}} = ({10^{7.395 - 5860/T}})/760 \times 101.325$ ,Range: 870-1070K(42) 51.5%KCl + 48.5%AlCl₃: ${P_{vap}} = ({10^{9.2386 - 7846/T}})/760 \times 101.325$ ,Range: 910-1070K(43) 57.6%KCl + 42.4%AlCl₃: ${P_{vap}} = ({10^{8.943 - 7634/T}})/760 \times 101.325$ ,Range: 920-1030K(44) 63.8%KCl + 36.2%AlCl₃: ${P_{vap}} = ({10^{8.4231 - 7212/T}})/760 \times 101.325$ ,Range: 950-1030K(45) where P_vap is in kPa,unit of temperature is in K.
870	0.613
880	0.72
890	0.867
900	1.027
910	1.2	0.547
920	1.413	0.68	0.587
930	1.653	0.84	0.72
940	1.933	1.04	0.88
950	2.24	1.267	1.08	0.907
960	2.6	1.547	1.307	1.08
970	3.013	1.88	1.573	1.293
980	3.466	2.28	1.893	1.547
990	3.986	2.746	2.28	1.827
1000	4.573	3.293	2.72	2.173
1010	5.226	3.933	3.226	2.56
1020	5.96	4.693	3.84	3
1030	6.773	5.573	4.533	3.52
1040	7.679	6.599
1050	8.693	7.786
1060	9.813	9.159
1070	11.052	10.732

| Show Table

DownLoad: CSV

Figure 8. Properties of molten salt mixture KCl+AlCl₃.

DownLoad: Full-Size Img PowerPoint

4. Properties of Ternary Molten Salt Systems

This section surveyed properties of two ternary molten salts,NaCl + KCl + AlCl₃ and NaCl + KCl + ZnCl₂. All the compositions referred to are based on mole fraction.

4.1. NaCl+KCl+AlCl₃

There are two eutectic points for NaCl + KCl + AlCl₃ [66]. The one with a composition of 20%NaCl + 16.5%KCl + 63.5%AlCl₃ has a melting point ofT_m = 361.9 K(88.9 ℃). The densities of the ternary salt with different compositions at different temperatures are given in Table 7 and Figure 9. There is no vapor pressure data found for this ternary system. However,the high vapor pressure of AlCl₃ could make this system to have high vapor pressures and thus not suitable as a HTF.

Table 7. Density of liquid NaCl + KCl + AlCl₃(% by mole) [66] .

T(K)	p(kg/m³)
T(K)	50%NaCl +40%KCl +10%AlCl₃	55%NaCl +25%KCl +20%AlCl₃	60%NaCl +30%KCl +10%AlCl₃	60%NaCl +20%KCl +20%AlCl₃	60%NaCl +10%KCl +30%AlCl₃	65%NaCl +25%KCl +10%AlCl₃	65%NaCl +10%KCl +25%AlCl₃
50%NaCl+40%KCl+10%AlCl₃: $\rho = 2136 - 0.923 \times T$ ,for T(K)in 500-540K(46) 55%NaCl+25%KCl+20%AlCl₃: $\rho = 2105 - 0.903 \times T$ ,for T(K)in 440-480K(47) 60%NaCl+30%KCl+10%AlCl₃: $\rho = 2096 - 0.882 \times T$ ,for T(K)in 430-480K(48) 60%NaCl+20%KCl+20%AlCl₃: $\rho = 2117 - 0.97 \times T$ ,for T(K)in 430-480K(49) 60%NaCl+10%KCl+30%AlCl₃: $\rho = 2059 - 0.822 \times T$ ,for T(K)in 470-520K(50) 65%NaCl+25%KCl+10%AlCl₃: $\rho = 2059 - 0.877 \times T$ ,for T(K)in 430-480K(51) 65%NaCl+10%KCl+25%AlCl₃: $\rho = 2115 - 1.02 \times T$ ,for T(K)in 960-1170K(52) where unit of density is p(kg/m³) and T is in(K).
430			1716.7	1699.9		1681.9	1676.4
440		1707.7	1707.9	1690.2		1673.1	1666.2
450		1698.7	1699.1	1680.5		1664.4	1656.0
460		1689.6	1690.3	1670.8		1655.6	1645.8
470		1680.6	1681.5	1661.1	1672.7	1646.8	1635.6
480		1671.6	1672.6	1651.4	1664.4	1638.0	1625.4
490					1656.2
500	1674.5				1648.0
510	1665.3				1639.8
520	1656.0				1631.6
530	1646.8
540	1637.6

| Show Table

DownLoad: CSV

Figure 9. Densities of ternary mixture NaCl + KCl + AlCl₃ [66] at various compositions.

DownLoad: Full-Size Img PowerPoint

4.2. NaCl + KCl + ZnCl₂

There are three eutectic mixtures for this ternary system that may have melting temperatures below 250 ℃. One particular composition is 20% NaCl + 20% KCl + 60% ZnCl₂ that has a melting point of T_m = 476 K(203 ℃). Because of the relatively low vapor pressure of ZnCl₂,this eutectic salt may also have a low vapor pressure and thus is very promising to be a HTF. The densities of this eutectic mixture are given in Table 8 and Figure 10(a). Viscosities of the salt are given in Table 9 and Figure 10(b).

Table 8. Densities of the molten salt 20%NaCl + 20%KCl + 60%ZnCl₂ (% by mole) [74]

T(K)	p(g/cm³)	p(kg/m³)
$\rho = 2.59 - 6.36 \times {10^{ - 4}}(T - 273)$ ; where the units are p(kg/m³),and T(K),in the range of 473-573K.(53)
473	2.46	2462.8
483	2.46	2456.44
493	2.45	2450.08
503	2.44	2443.72
513	2.44	2437.36
523	2.43	2431
533	2.42	2424.64
543	2.42	2418.28
553	2.41	2411.92
563	2.41	2405.56
573	2.4	2399.2

| Show Table

DownLoad: CSV

Table 9. Viscosity of the molten salt 20%NaCl + 20%KCl + 60%ZnCl₂ (% by mole) [74]

T(K)	μ(cp)	μ(kg/m*s)
$\mu = 1.23 \times {10^{ - 2}}\sqrt T \exp \left( {\frac{{1.21 \times {{10}^3}}}{{T - 283}}} \right)$ ; where the units areμ(cp),and T(K),within the range of 473-573K.(54)
473	155.9932	0.15599
483	114.6464	0.11465
493	86.8344	0.08683
503	67.5008	0.0675
513	53.6699	0.05367
523	43.5235	0.04352
533	35.9132	0.03591
543	30.0916	0.03009
553	25.5594	0.02556
563	21.9752	0.02198
573	19.1002	0.0191

| Show Table

DownLoad: CSV

Figure 10. Properties of salt(20%NaCl + 20%KCl + 60%ZnCl₂(% by mole)).

DownLoad: Full-Size Img PowerPoint

5. Conclusion and Outlook

This paper gave a brief introduction to the currently available high temperature HTFs and the necessity of developing new-generation materials for even higher working temperatures(at least 800 ℃)for the goal of high thermal efficiency in concentrated solar thermal power plant.

The authors particularly gave attention to possible mixtures of ionic and covalent halide salts for the potential of creating a eutectic salts to meet the basic criteria for a HTF. Low cost and almost unlimited reserve of halide salts is also important due to the large dem and in industry. The reviewing has found that some binary and ternary halide salts could have eutectic mixtures with low melting points,which therefore are recommendable for the components of a HTF.

For those halide salts recommended as the components of mixture eutectic molten salts,available properties of the individual c and idate and binary and ternary mixtures are surveyed and presented for reference of further studying and investigation of all the properties that remaining unknown so far. Table 10 summaries the availability of results obtained from literature. Some key conclusions from the survey are drawn:

(1)It is understood that the three covalent halide salts AlCl₃,ZnCl₂,and FeCl₃ have relatively low melting points which may contribute to the low eutectic melting point for a mixture of ionic and covalent halide salts. However,the relatively very high vapor pressures of AlCl₃ and FeCl₃ are of concerns to the possible high vapor pressures of eutectic salt mixtures.

(2)Two binary eutectics salts NaCl-AlCl₃ and KCl-AlCl₃ show low eutectic melting points below 250 ℃. The system of NaCl-AlCl₃ still has rather high vapor pressure which may come from the high vapor pressure of AlCl₃. However,the eutectic of KCl-AlCl₃ show relatively lower vapor pressure compared to that of NaCl-AlCl₃. This is an indication that KCl is an important component to suppress the vapor pressure of eutectic mixtures. The viscosities of these two binary systems are low and very favorable.

(3)Two ternary eutectic salts from the system of NaCl-KCl-AlCl₃ and NaCl-KCl-ZnCl₂ showed low viscosities that are very favorable for heat transfer fluids. The densities of the two eutectic salts are also acceptable. However,there are no other thermal and transport properties,which need research and investigation.

Last but not least,the chemical corrosion of the surveyed salts to metals of pipes and containers is a very important issue. A detailed survey is yet to be carried out. However,one recent article [75] showed results of corrosion of Hastelloys in 13.4NaCl-33.7KCl-52.9ZnCl₂(mol%)eutectic system. Evaluated from both electrochemical method and immersion test,the type N Hastelloy showed a higher corrosion rate of > 150 µm per year at 500 ℃,but the Hastelloy C-276 exhibited the lowest corrosion rate of 40 µm per year at 500 ℃. More results on corrosions of these salts to metals are expected to see in the future.

Referring to the available properties for these halide salts,the researchers teamed-up from the University of Arizona,Georgia Institute of Technology,and Arizona State University have been investigating details of properties of ternary salt systems(KCl-NaCl-AlCl3),(KCl-NaCl-ZnCl₂),and (KCl-NaCl-FeCl₃),as well as quaternary systems. Simulations and fast screening method are applied for searching eutectic salt compositions; while experimental tests are followed up to measure all the properties,including corrosion to metals,for evaluation of the suitability for a HTF.

Table 10. Summary of the availability of physical properties of AlCl₃,ZnCl₂,FeCl₃,NaCl,KCl,and their mixtures.

	Single Molten Salt					Binary Molten Salt			Ternary Molten Salt
	AlCl₃	ZnCl₂	FeCl₃	NaCl	KCl	NaCl+AlCl₃	KCl+AlCl₃	NaCl+KCl	NaCl+KCl +AlCl₃	NaCl+KCl +ZnCl₂
√ : Data obtained from literature;× : No data available from literature
Density	√	√	×	√	√	√	√	√	√	√
Viscosity	√	√	×	√	√	√	×	√	×	√
Thermal Conductivity	×	×	×	√	√	×	×	×	×	×
Specific Heat Capacity	√	√	√	√	√	×	×	×	×	×
Vapor Pressure	√	√	√	√	√	√	√	√	×	×
Melting Point	√	√	√	√	√	√	√	√	√	√

| Show Table

DownLoad: CSV

Acknowledgement

The support from the U.S. Office of DOE under the Contract DE-EE0005942 is gratefully acknowledged.

Conflict of Interest

All authors declare no conflicts of interest in this paper.

References

[1]	W. J. Yan, D. H. Chen, L. Liu, Research progress of hyperspectral image classification, Opt. Precis. Eng., 27 (2019), 680-693.
[2]	S. Y. Chen, H. Z. Lin, X. Zhao, G. Wang, Deep learning-based classification of hyperspectral data, IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., 7 (2014), 2094-2107.
[3]	F. W. Fu, B. W. Zou, Review of remote sensing image classification based on deep learning, Appl. Res. Comput., 35 (2018), 3521-3525.
[4]	J. Zhao, Y. Zhong, H. Shu, L. P. Zhang, High-resolution image classification integrating spectral-spatial-location cues by conditional random fields, IEEE. Trans. Image Process. Publ. IEEE Signal Process. Soc., 25 (2016), 4033-4045.
[5]	A. B. Salberg, Detection of seals in remote sensing images using features extracted from deep convolutional neural network, Geosci. Remote Sens. Symp. IEEE, (2015), 1893-1896.
[6]	C. Cortes, V. Vapnik, Support-vector networks, Mach. Learn., 3 (1995), 273-297.
[7]	D. C. Feng, G. Chen, W. Y. Du, X. Y. Wu, K. L. Xiao, Remote sensing image classification based on minimum distance method, J. Beihua. Inst. Aerosp. Technol., 22 (2012), 1-3.
[8]	A. Ahmad, S. Hashmi, K-Harmonic means type clustering algorithm for mixed datasets, Appl. Soft Comput., (2016), 39-49.
[9]	Y. Zhang, H. T. Yang, C. H. Yuan, A survey of remote sensing image classification methods, J. Weapon Equip. Eng., 39 (2018), 108-112.
[10]	M. Volpi, D. Tuia, Dense semantic labeling of subdecimeter resolution images with convolutional neural networks, IEEE Trans. Geosci. Remote Sens., (2017), 1-13.
[11]	G. E. Hinton, R. R. Salakhutdinov, Reducing the dimensionality of data with neural networks, Sci., 313 (2006), 504-507.
[12]	W. Y. Pang, L. M. Sun, H. X. Jiang, L. X. Li, Convolution in convolution for network in network, IEEE Trans. Neural Networks. Learn. Syst., 29 (2018), 1587-1597.
[13]	A. S. Razavian, H. Azizpour, J. Sullivan, S. Carlsson, CNN features off-the-shelf: An astounding baseline for recognition, Comput. Vision Pattern Recognit. Workshops. IEEE, (2014), 512-519.
[14]	J. P. Zhao, W. W. Guo, S. Y. Cui, Z. H. Zhang, Convolutional neural network for SAR image classification at patch level, Geosci. Remote Sens. Symp. IEEE, (2016), 945-948.
[15]	H. B. Lyu, H. Lu, L. C. Mou, Learning a transferable change rule from a recurrent neural network for land cover change detection, Remote Sens., 8 (2016).
[16]	N. Kussul, M. Lavreniuk, S. V. Skakun, A. Y. Shelestov, Deep learning classification of land cover and crop types using remote sensing data, IEEE Geosci. Remote Sens. Lett., 14(2017), 778-782.
[17]	R. Marc, K. Marco, Multi-temporal land cover classification with sequential recurrent encoders, ISPRS Int. J. Geo-Inf., 7 (2018).
[18]	Y. Liu, Z. S. Liao, Multiple kernel learning with the generalized error bound of support vector machine, J. Wuhan Univ. (Nat. Sci. Ed.), 58 (2012), 149-156.
[19]	Q. J. Zhang, X. J. Zhang. Z. Zhao, Y. J. Wang, Classification of polarimetric SAR images based on multiscale segmentation and radial basis function neural network, Geomat. Spat. Inf. Technol., (2019), 67-71.
[20]	M. Zhou, Dimension reduction and classification of hyperspectral remote sensing image based on RBF neural network, Territ. Nat. Resour. Study, (2016), 14-16.
[21]	P. V. Arun, K. M. Buddhiraju, A. Porwal, CNN based sub-pixel mapping for hyperspectral images, Neurocomput., 311 (2018), 51-64.
[22]	A. Krizhevsky, I. Sutskever, G. Hinton, Image net classification with deep convolutional neural networks, NIPS, 60 (2017), 84-90.
[23]	C. Szegedy, V. Vanhoucke, S. Loffe, J. Shlens, Z. Wojna, Rethinking the inception architecture for computer vision, IEEE Conf. Comput. Vision Pattern Recognit., 2016.
[24]	Krizhevsky, Alex, Sutskever, Ilya, Hinton, E. Geoffrey, Image net classification with deep convolutional neural networks, Commun. ACM, 60 (2012), 84-90.
[25]	Q. Z. Gong, P. Zhong, Y. Yu, D. W. Hu, Diversity-promoting deep structural metric learning for remote sensing scene classification, IEEE Trans. Geosci. Remote Sens., (2017), 1-20.
[26]	X. Glorot, A. Bordes, Y. Bengio, Deep sparse rectifier neural networks, J. Mach. Learn. Res., 2011.
[27]	N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, R. Salakhutdinov, Dropout: A simple way to prevent neural networks from overfitting, J. Mach. Learn. Res., (2014), 1929-1958.

This article has been cited by:

1.	Ringetani Matlou, Yonas T. Bahta, Enoch Owusu-Sekyere, Henry Jordaan, Impact of Agricultural Drought Resilience on the Welfare of Smallholder Livestock Farming Households in the Northern Cape Province of South Africa, 2021, 10, 2073-445X, 562, 10.3390/land10060562
2.	Sheunesu Ruwanza, Gladman Thondhlana, Menelisi Falayi, Research Progress and Conceptual Insights on Drought Impacts and Responses among Smallholder Farmers in South Africa: A Review, 2022, 11, 2073-445X, 159, 10.3390/land11020159
3.	Yonas T. Bahta, Nexus between Coping Strategies and Households’ Agricultural Drought Resilience to Food Insecurity in South Africa, 2022, 11, 2073-445X, 893, 10.3390/land11060893
4.	Bethwell Moyo, Khuliso E. Ravhuhali, Dry season feeding strategies and winter forage production by communal area sheep farmers of the Eastern Cape province in South Africa, 2023, 9, 2331-1932, 10.1080/23311932.2022.2161775
5.	Leocadia Zhou, Mhlangabezi Slayi, Saul Ngarava, Ishmael Festus Jaja, Lovemore Musemwa, A Systematic Review of Climate Change Risks to Communal Livestock Production and Response Strategies in South Africa, 2022, 3, 2673-6225, 10.3389/fanim.2022.868468
6.	Lindie V. Maltitz, Yonas T. Bahta, Empowerment of smallholder female livestock farmers and its potential impacts to their resilience to agricultural drought, 2021, 6, 2471-2086, 603, 10.3934/agrfood.2021036
7.	Yonas T. Bahta, Vuyiseka A. Myeki, Adaptation, coping strategies and resilience of agricultural drought in South Africa: implication for the sustainability of livestock sector, 2021, 7, 24058440, e08280, 10.1016/j.heliyon.2021.e08280
8.	Yonas T. Bahta, Social vulnerability to agricultural drought: Insights from Northern Cape, South Africa, 2022, 17, 24682276, e01324, 10.1016/j.sciaf.2022.e01324
9.	Silindile Sibiya, Jai Kumar Clifford-Holmes, James Gambiza, Drivers of Degradation of Croplands and Abandoned Lands: A Case Study of Macubeni Communal Land in the Eastern Cape, South Africa, 2023, 12, 2073-445X, 606, 10.3390/land12030606
10.	Yonas T. Bahta, Vuyiseka A. Myeki, The Impact of Agricultural Drought on Smallholder Livestock Farmers: Empirical Evidence Insights from Northern Cape, South Africa, 2022, 12, 2077-0472, 442, 10.3390/agriculture12040442
11.	Chenaimoyo L.F. Katiyatiya, Jabulani Majaha, Obert C. Chikwanha, Kennedy Dzama, Nkgaugelo Kgasago, Cletos Mapiye, Drought's implications on agricultural skills in South Africa, 2022, 51, 0030-7270, 293, 10.1177/00307270221103289
12.	Zahra Khoshnodifar, Hamid Karimi, Pouria Ataei, Mechanisms to change farmers' drought adaptation behaviors in Sistan and Baluchistan Province, Iran, 2023, 7, 2571-581X, 10.3389/fsufs.2023.1121254
13.	Yonas T. Bahta, Joseph P. Musara, Diversity of Food Insecurity Coping Strategies among Livestock Farmers in Northern Cape Province of South Africa, 2023, 11, 2225-1154, 82, 10.3390/cli11040082
14.	Thabo Ndlovu, Sipho Mamba, Contextualizing the Seasonal Livelihoods Programming Tool in Drought Resilience Planning Settings: Experiences from Southern Zimbabwe, 2022, 1556-5068, 10.2139/ssrn.4110994
15.	Marco d’Errico, Jeanne Pinay, Ellestina Jumbe, Anh Hong Luu, Drivers and stressors of resilience to food insecurity: evidence from 35 countries, 2023, 15, 1876-4517, 1161, 10.1007/s12571-023-01373-5
16.	Yonas T. Bahta, Willem A. Lombard, Nexus between Social Vulnerability and Resilience to Agricultural Drought amongst South African Smallholder Livestock Households, 2023, 14, 2073-4433, 900, 10.3390/atmos14050900
17.	Ashish Sharma, Rainfall deficiency, drought and economic growth in the Bundelkhand region of India, 2023, 9, 2363-5037, 10.1007/s40899-023-00851-0
18.	Prince Nketiah, Herbert Ntuli, Empirical analysis of drought-induced cattle destocking in South Africa, 2024, 16, 1996-1421, 10.4102/jamba.v16i1.1557
19.	Dumisani Shoko Kori, Clare Kelso, Walter Musakwa, Climate change adaptation by smallholder farmers in Southern Africa: a bibliometric analysis and systematic review, 2024, 6, 2515-7620, 032002, 10.1088/2515-7620/ad3127
20.	L. Mdiya, M. Aliber, S. Ngarava, N.V. Bontsa, L. Zhou, Impact of Extension Services on the Use of Climate Change Coping Strategies for Smallholder Ruminant Livestock Farmers in Raymond Local Municipality, Eastern Cape Province, South Africa, 2023, 51, 2413-3221, 150, 10.17159/2413-3221/2023/v51n2a15725
21.	Sennan D. Mattar, Neil J. W. Crawford, 2023, Chapter 4, 978-3-031-29528-7, 51, 10.1007/978-3-031-29529-4_4
22.	Bouchra El Amiri, Reda Azmi, Said Laaribya, Brahim Dani, Mounia Sibaoueih, Berkat Omar, 2024, Chapter 7, 978-3-031-59602-5, 99, 10.1007/978-3-031-59603-2_7
23.	Thabo Ndlovu, Sipho Felix Mamba, Contextualizing the seasonal livelihoods programming tool in drought resilience planning settings: Experiences from southern Zimbabwe, 2023, 95, 22124209, 103908, 10.1016/j.ijdrr.2023.103908

Reader Comments

Your name:*

Email:*
© 2020 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)