Optimizing Greenhouse Lighting for Advanced Agriculture ...downloads.hindawi.com/journals/mpe/2017/6862038.pdf4 MathematicalProblemsinEngineering 0 5 10 15 20 Time (hour) 0 5 10 15

Research ArticleOptimizing Greenhouse Lighting for Advanced AgricultureBased on Real Time Electricity Market Price

Mehdi Mahdavian and Naruemon Wattanapongsakorn

Department of Computer Engineering, King Mongkut’s University of Technology Thonburi, 126 Pracha-Utid Rd.,Bangmod, Toongkru, Bangkok 10140, Thailand

Correspondence should be addressed to Mehdi Mahdavian; meh [email protected]

Received 14 August 2016; Accepted 21 December 2016; Published 22 February 2017

Academic Editor: Alessandro Gasparetto

Copyright © 2017 Mehdi Mahdavian and Naruemon Wattanapongsakorn. This is an open access article distributed under theCreative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, providedthe original work is properly cited.

The world’s growing demand for food can bemet by agricultural technology. Use of artificial light to supplement natural sunlight ingreenhouse cultivation is one of the most common techniques to increase greenhouse production of food crops. However, artificiallight requires significant electrical energy, which increases the cost of greenhouse production and can reduce profit.This papermod-els the increments to greenhouse productivity aswell as the increases in cost from supplemental electric lighting, in a situationwherethe greenhouse is one of the elements of a smart grid, a system where the electric energy market is dynamic and prices vary overtime. We used our models to calculate the optimum values for supplemental light and the required electrical energy for HPS lampsin the greenhouse environment, using cherry tomato cultivation as a case study crop. We considered two optimization techniques:iterative search (IS) and genetic algorithm (GA). The two approaches produced similar results, although the GA method was muchfaster. Both approaches verify the advantages of using optimal supplemental light in terms of increasing production andhence profit.

1. Introduction

According to the UN Food and Agriculture Organization(FAO), providing adequate food for the increasingworld pop-ulation is one of the most serious challenges of the 21st cen-tury. Insufficient financial resources, limited freshwater, lim-ited agricultural land, incompatible climate, and inadequatelight for photosynthesis are some of the most important bar-riers to improving agricultural production. Use of industrialgreenhouses can help to solve or reduce the above-mentionedproblems. Industrial greenhouses allow certain crops to begrown in all seasons during the year. Greenhouses are veryimportant for food supply and food security in many regionsof the world, especially in the high latitude countries. Green-house climate control has led to significant improvements inthe efficiency and performance of agriculture. However,greenhouse climate is a complex, nonlinear, and uncertainsystem, involving environmental factors like temperature,humidity, light, CO

2concentration, and so forth [1]. Thus,

optimizing greenhouse climate is a key goal for today’sadvanced agriculture [2, 3].

The single most important factor in plant growth is light.Optimal control of greenhouse lighting is one of the key tech-niques in digital agriculture. Nowadays, it is very prevalentto use artificial and supplemental lighting in greenhouses.Electrical lamps have been used to grow plants for nearly 150years [4]. Optimizing light to maximize production, though,may result in higher costs and lower profit overall. In thetraditional power grid with constant or quasi-constant tariffs,computation of costs and benefits of using artificial lightingin greenhouse environment is not very difficult. Given thecurrent shift from traditional power grids to “smart grids,”however, the problem becomes far more complex.

The efficiency and performance improvement for green-house productions, for example, cherry tomato, can be con-sidered by optimization of effective components of the cropsproduction.TheRTP in the smart grid causesmore sensitivityof the greenhouse owners for electricity consumption ingreenhouse crops production.Use of evolutionary algorithmssuch as GA can determine the optimal values of electricityconsumption in production trend and it causes the efficiencyimprovement of the greenhouse production [5].

HindawiMathematical Problems in EngineeringVolume 2017, Article ID 6862038, 11 pageshttps://doi.org/10.1155/2017/6862038

https://doi.org/10.1155/2017/6862038

2 Mathematical Problems in Engineering

Today, the increased use of fossil fuels resources inelectrical energy production has been caused increasing ofthe CO

2footprint and this issue has been created by the

significant problems like the global warming, disruption ofthe wildlife ecosystems, drought increment, and, and soforth.The optimal electricity energy consumption in the lightsection of the greenhouse environment can cause reductionof CO

2footprint in the earth’s atmosphere.

Different types of the electrical lamps can be used forgreenhouse electrical lighting but the most commonly usedlamp types in tomato cultivation are High Pressure Sodium(HPS) and Light Emitting Diode (LED) due to their favorable(PAR/Watt) value and low failure rate, both at the beginningmoments of lighting and during the lamp’s life. Although theLED lamps technology is growing and developing, use ofHPSlamps in industrial greenhouses is substantially considered byfarmers due to some important reasons such as being of lowerprice, of higher available light power, and more appropriatelight spectrum [6].

The control of electrical lamps light in greenhouse envi-ronment can be done by turning on and off of the lamps orlamps dimming.The on-off switch controlmethod is a simpleand affordable approach in comparison with using industrialProportional Integral Derivative (PID) controllers which areimplemented by microcontrollers. But the controllability ofthe PID controllers is much better than controllability ofthe on-off switch controllers. Moreover, use of the lampsdimming method provides the smoother and more uniformlight for greenhouse environment in comparison to the on-offmethod [7].

A so-called smart grid delivers electricity from suppliersto consumers by predicting and responding intelligently tothe behavior and actions of all electric power users.The smartgrid concept incorporates many issues such as power systemstability, power system reliability, self-healing, renewableenergies, privacy and security, energy storage, flexibility innetwork topology, efficiency, sustainability, marketing, peakcurtailment, price leveling, and dynamic pricing [8]. Thelast item poses the greatest challenge to greenhouse lightingoptimization.

The electricity market in a smart grid will be highlyactive. Electricity tariffs can be different from one momentto the next [9]. Tariffs in the smart grid will be a function ofmarket rules as well as internal and external grid conditions.The best performance and efficiency in greenhouse lightingsystems can be achieved by proper control of electric energyin bulbs [10], but if varying electricity prices in the smart gridare ignored, employing the artificial lighting in greenhouseenvironment can be associated with financial losses [11].On the other hand, if prices are the only consideration,greenhouse performance may drop to a level lower than thatachieved where artificial lighting is not used at all [12].

In this paper, we consider greenhouse lighting optimiza-tion for cherry tomato cultivation based onReal Time Pricing(RTP) of electricity in the smart grid. Maximum possiblecrop productivity with minimum electrical energy cost is theultimate target. We derive a set of mathematical models todescribe the use of artificial light in maintaining an industrialgreenhouse environment. Our models incorporate both the

productivity benefits to be gained from artificial lighting andthe varying costs associated with electricity derived from asmart grid.We then explore two optimizationmethods, itera-tive search (IS) and genetic algorithm (GA), to determine theenergy use and timing that result in the best overall outcomes.We use tomatoes as a test crop since this vegetable has beengrown in greenhouses for nearly 100 years [13] and has beenstudied extensively.

Other research in this area has generally used hypotheti-cal or stem models [14, 15]. Our research emphasizes the useof updated practical models in the simulation process, basedon real world information about crop productivity, lightsource energy, and smart grid pricing. Hence we expect thatour work could be immediately applied in an industrial agri-culture setting.

The implementation of the proposed method can bedone by use of the architecture of a versatile networkingand control platform for greenhouse electrical lighting whichcan be based on embedded wireless and wired networkingtechnologies. In this regard, dynamic greenhouse electri-cal lighting can be applied and optimized by consideringobjective functions like cost, the required wiring installationexpenses, maintenance complexity, and so forth [16].

2. Materials and Methods

2.1. Modeling of Tomato Greenhouse Cultivation. Using arti-ficial lighting for tomato cultivation has a big influence ontomato growth process [19]. Tomato is one of the most com-mon global greenhouse crops.The growth of tomato is relatedto many factors and parameters. One of the most importantfactors is the light. Over the past century, several models havebeen presented for representing the relationship betweenphotosynthesis and light [20]. Some of these models aresimple, while others are more sophisticated. Equations (1)–(6) express some of these models. In these equations, 𝑃(𝐼),𝑃𝑚, 𝐼, and 𝐼

𝑠are the values of photosynthesis, maximum pho-

tosynthesis, light and light saturation constant, respectively.

𝑃 (𝐼) = {{{{{𝑃𝑚( 𝐼𝐼𝑠

) 𝐼 ≤ 𝐼𝑠,

𝑃𝑚𝐼 > 𝐼𝑠,

Blackman (1905) , (1)

𝑃 (𝐼) = 𝑃𝑚 (𝐼/𝐼𝑠)1 + (𝐼/𝐼𝑠) , Baly (1935) , (2)

𝑃 (𝐼) = 𝑃𝑚 (𝐼/𝐼𝑠)

√(1 + (𝐼/𝐼𝑠)2) ,Smith (1936) , Talling (1957) ,

(3)

𝑃 (𝐼) = 𝑃𝑚 ( 𝐼𝐼𝑠

) exp(1 − 𝐼𝐼𝑠

) , Steele (1962) , (4)

𝑃 (𝐼) = 𝑃𝑚 [1 − exp( 𝐼𝐼𝑠

)] , web et al. (1974) , (5)

𝑃 (𝐼) = 𝑃𝑚 tanh( 𝐼𝐼𝑠

) , Jessby and Platt (1976) . (6)

Mathematical Problems in Engineering 3

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 20

0.2

0.4

0.6

0.8

1

(P/P

m)

(I/Iopt)

P (I) = Pm ( I

Iopt) exp(1 − I

Iopt)

Figure 1: Steele photosynthesis model graph.

Table 1: Tomato production per plant based on the amount of lightduring 120 days.

Light (PAR)(mol/m2 day)

Fresh tomato(Kg/Plant)

20.4 4.2710.2 1.78

Among the above-mentioned models, model of (1) is alinear function. Equations (2), (3), (5), and (6) are nonlinearmodels but they are monotonic functions. These models incomparison to Steele model (4) are not eligible for the casestudy of this paper due to at least twomain reasons. Firstly, thephotosynthesis process of cherry tomato is a nonlinear phe-nomenon [21]. Secondly, the indiscriminate increase of lightwould weaken the photosynthesis process of cherry tomatoand would be harmful for the plant [22].

Therefore, the Steele model is the best model which canbe selected among the above-mentioned models consideringtomato cultivation as the case study in this paper. This modelhas more flexibility and is better adapted for the case studyof this research in comparison with the other models above.Equation (4) can be modified as (7). The dynamics of thismodel are shown in Figure 1. It is worthy to notice that 𝐼optis the value of optimal light.

𝑃 (𝐼) = 𝑃𝑚 ( 𝐼𝐼opt) exp(1 − 𝐼𝐼opt) , Steele (1962) . (7)

In a practical research study [21], cherry tomato busheswere cultivated in the greenhouse environment from Febru-ary 4 to June 4. The mean day length for place of research(Germany) was considered approximately from 8 AM to 4PM. In this research, the mean temperature was approxi-mately equal to 19.6∘C, the relative humidity was approx-imately equal to 74.8%, and the CO

2concentration was

approximately equal to 391 𝜇mol. Also the plants were cul-tivated and fed based on the commercial best practices.Table 1 summarizes the results of this research, in terms ofthe relationship between light and yield.

Photosynthesis Active Radiation (PAR, measured interms ofmicromole/sec where 1molemeans 6∗1023 photons)

Table 2: Sunlight at different time of the day.

Time(Hour)

Sunlight (PAR)(mol/m2 day)

0 06 08 1.4212 31.216 1.4218 024 0

is essential for plant growth. Also the plant density isapproximately about two plants per square meter.

Cherry tomato plant density has a big influence on finalyield and quality. The number of tomato plants per unit area(square meter) tends to be normally set between 2 to 4 plantsper squaremeter [17]. Some studies reportmore than 4 plantsper square meter [23]. In this paper, we use the lower bound(2 plants per square meter) to maximize of reliability marginof numerical calculations. In fact, the calculations of thispaperwere done based onminimumpractical plant density (2plants per square meter) in greenhouse environment. In thisregard, if usage of electrical lighting for the minimum plantdensity is affordable, thus it would be more affordable forhigher plant density (more than 2 plants per square meter).

We substituted the values from Table 2 into (7) to obtaina tomato production model based on 𝐼 as the emitted light.The model is illustrated by (8) and (9). In these equations,PAR𝑡, PAR

𝑛, and PAR

𝑠are the total values of light, natural

light (sunlight), and supplemental light, respectively.

𝑃 (𝐼) = 0.222𝐼 exp (1 − 𝐼56.1) (Kg/m2) , (8)

𝑃 (PAR𝑡) = 0.222PAR

𝑡exp (1 − PAR

𝑡56.1 ) (Kg/m2) , (9)

PAR𝑡= PAR

𝑛+ PAR

𝑠(mol/m2 day) . (10)

2.2. Modeling of the Light Resources. Sunlight is the bestand cheapest light resource. It provides a continuous perfectspectrum for the photosynthesis. After passing through theearth atmosphere, it reaches earth’s surface in the frequencyrange of 250 nm to 2500 nm, which includes the visible lightdomain (380 nm to 720 nm) plus the invisible light domain.The photosynthesis frequency domain of plants is between250 nm and 750 nm. Sunlight radiation modeling is a trendbased on accuratemeasurement.Measurement andmodelingof solar radiation is important for many systems and applica-tions [24, 25]. Our sunlight model was formulated by extract-ing the sunlight radiation values in the paper discussed above[21]. We then used MATLAB software to fit a Gaussian equa-tion to these values. The peak of sunlight radiation occurs atnoon when the sun is at its greatest height above the horizon.


0 5 10 15 20

Time (hour)

0

5

10

15

20

25

30

Sunl

ight

(PA

R) (m

ol/m

2da

y)

Figure 2: Sunlight radiation model.

Table 3: Technical data for GE HPS lamps [17].

Lamp type Power(Watt)

PAR𝑠

(𝜇mol/m2s)PAR𝑠

(mol/m2 day)LU250W/PSL 250 430 37.15LU400W/PSL 420 710 61.34LU600W/PSL 615 1080 93.31LU750W/PSL 755 1320 114.05LU400V/600W/PSL 620 1120 96.77LU400V/750W/PSL 765 1390 120.1

The extracted values, results of modeling, and model equa-tion are shown in Table 2, Figure 2, and Gaussian equation(11), respectively.

PAR𝑛

= 31.2 exp(−((Time − 12)2.273 )2) (mol/m2 day) . (11)

Based on (10), (12) can be written as follows:

PAR𝑠= PAR

𝑡− PAR

𝑛(mol/m2 day) . (12)

By combining (10) and (12), the sunlight radiation modelcan be expressed

PAR𝑠

= PAR𝑡

− 31.2 exp(−((Time − 12)2.273 )2) (mol/m2 day) .(13)

HPS lamps produced by some companies such as GeneralElectric (GE) have been optimized specially for greenhouseuse by improving the red portion in their output spectrum[26].Thismeans that HPS lamps designed for the greenhousemarket can use a lower light level (lumen) in the visiblespectral range compared to HPS lamps designed for streetlighting. Some technical data for HPS lamps which aremanufactured by GE are presented in Table 3.

GE HPS lamp modeling

200 300 400 500 600 700 800

Power (watts)

200

400

600

800

1000

1200

1400

1600

1800

Data 1LinearQuadratic

Cubic4th degree

y = 1.8 ∗ x − 41y = 0.00017 ∗ x2 + 1.7 ∗ x − 1.7y = −2.4e − 006 ∗ x3 + 0.0039 ∗ x2 − 0.14 ∗ x + 2.6e + 002

y = 1.9e − 007 ∗ x4 − 0.00038 ∗ x3 + x −0.28 ∗ 2

86 ∗ x + 9.4e + 003

PAR

(𝜇m

ol/m

2s)

Figure 3: Several models for GE HPS lamps.

Data 1LinearQuadratic

Cubic4th degree

0 50 100 150

Pow

er (w

atts)

GE HPS lamp modeling

−200

0

200

400

600

800

1000

y = 6.4 ∗ x + 9.2

y = −0.0048 ∗ x2 + 7 ∗ x − 0.41

y = 6.6e − 006 ∗ x3 − 0.006 ∗ x2 + 7.1 ∗ x − 0.6

y = 1.6e − 006 ∗ x4 − 0.00041 ∗ x3 + 0.027 ∗ x2 +6.3 ∗ x − 0.27

PAR (mol/m2 day)

Figure 4: Different models for GE HPS lamps.

The relationship between PAR and lamp power can bemodeled by fitting a curve the data in Table 3. We fit severaldifferent models using MATLAB software (version 7.12.0).The simulation results are shown in Figures 3 and 4. InFigure 3, the 𝑌 and𝑋 axes are PAR and power, respectively.

The models do not differ much, and all have relativelylow error. Goodness of fit values for these models is shown inTable 4. We chose to use the quadratic model which is shownin (14) and (15) where the power is total wattage amount ofHPS lamps.

PAR𝑠= 0.14688 [(Power)210000 + (Power)

− 1] (mol/m2 day) ,(14)


Table 4:The goodness of fit values for the above-mentionedmodels(obtained models in Figures 3 and 4).

Index Linear Quadratic Cubic 4th degreeSSE 724.3 641 595.1 274𝑅-Square 0.9964 0.9968 0.997 0.9986Adjusted 𝑅-square 0.9955 0.9947 0.9926 0.9932RMSE 13.46 14.62 17.25 16.57

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour)

0

5,000

10,000

15,000

20,000

25,000

30,000

Dem

and

(meg

a wat

t)

Figure 5: RTDs for electricity in a smart grid [18].

Power = [(−48104 (PAR𝑠)2) + 7 (PAR

𝑠)

− 0.41] (Watt/m2) .(15)

2.3. Modeling of Electricity Pricing in the Smart Grid. Oneof the important issues in the smart grid is dynamic pricingof electricity. Various methods of electricity pricing likeCritical Peak Pricing (CPP), Time-of-Use Pricing (TUP),Real Time Pricing (RTP), and so forth have been presented.RTP is reportedly one of the best methods [11]. Real timeelectricity pricing creates a closed loop feedback betweenthe physical layer and the market layer in the system. Infact, by dynamically changing their usage of electricity, thecustomers such as greenhouses can adapt their consumptionto the market conditions [27].

Modeling of electricity price in smart grid is one of the keyfeatures in the smart grid. Figure 5 shows the demand levels(load) at different times of a day for a real case [21]. In fact, thisgraph presents the Real Time Demands (RTDs) in the smartgrid. Based on Figure 5, Figure 6 displays Real Time Prices(RTPs) of electricity during a period of 24 hours [18].

Although the RTP of electricity is related tomany param-eters, RTD is one of themost critical. As illustrated by Figures5 and 6, the price of electricity increases with increasingelectricity demand and vice versa. Moreover, although thepeak demand for electricity is approximately 60%higher than

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour)

0

50

100

150

200

250

Pric

e (U

$/m

ega w

att h

our)

Figure 6: RTPs for electricity in a smart grid [18].

40

60

80

100

120

140

160

180

200

220

240

Pric

e (U

$/m

ega w

att h

our)

2 4 6 8 10 12 14 16 18 20 22

Time (hour)Price versus timeFit model

Figure 7: Curve fitting for 21 hot spots using MATLAB.

Table 5: The goodness of fit values for the proposed model.

Index ValueSum square error (SSE) 142.6𝑅-Square 0.9984Adjusted 𝑅-square 0.9893Root mean square error (RMSE) 6.893

the lowest electricity demand, the peak price of electricity isapproximately 600% higher than the lowest price of electric-ity. In other words, the RTP variations are very sensitive to theRTD variations. We extracted the key spots from the pricecurve above and used MATLAB to derive a mathematicalmodel. This model shows the relationship between price ofelectricity and time. The result is shown in Figure 7.

We use the Fourier equation (16) to represent the RTPmodel for electricity in the smart grid. Goodness of fit valuesfor this model is shown in Table 5.


Price = 𝑎0+ 8∑𝑛=1

(𝑎𝑛cos 𝑛 ∗ Time3.879 + 𝑏

𝑛sin 𝑛 ∗ Time3.879 ) (US$/Mega Watt Hour) , (16)

where 𝑎0= 114.3, 𝑎

1= −65.25, 𝑏

1= −48.16, 𝑎

2= 26.08, 𝑏

2=−0.341, 𝑎

3= −12.33, 𝑏

3= 2.064, 𝑎

4= −4.652, 𝑏

4= −1.629,𝑎

5= 4.869, 𝑏

5= 12.6, 𝑎

6= −7.953, 𝑏

6= −1.68, 𝑎

7= 16.71,𝑏

7= 11.76, 𝑎

8= −4.902, and 𝑏

8= −9.571.

2.4. Greenhouse Lighting Optimization. The ultimate goal inthis research is to maximize tomato cultivation profit by con-sidering several factors such as tomato price, electricity pricein themarket, and photosynthesis dynamics. In particular, wewant to find the optimal values of total light and supplementallight. The objective function can be written in the generalform of

Objective Function

= (Income − Cost) (US$/m2Hour) . (17)

The Income function can be written as (18)–(20).

Income = (Amount of Tomato)∗ (Price of Tomato) (US$/m2Hour) , (18)

where the amount of Tomato is specified in (9).

Income = [ 𝑃 (PAR𝑡)

(120 ∗ 24)] ∗ 5 (US$/m2Hour) . (19)

We used a value of 5 US$ per kg for the mean price ofcherry tomatoes. This price is the average price of cherrytomato in United States market in 2013 as reported by theUnited States Department of Agriculture [28]. We assumed atypical duration of tomato cultivation, namely, 120 days, andmodeled the full 24 hours of each day. Finally, by combining(9) and (19), the Income function can be written as

Income

= [0.222PAR𝑡exp (1 − PAR

𝑡/56.1)

576 ] (US$/m2Hour) . (20)

Considering electricity cost as themain factor in total costin tomato cultivation, the Cost function can be written as

Cost = (Consumed Power)∗ (Price of electricity) (US$/m2Hour) . (21)

The consumed power is specified in (15) and price ofelectricity is determined in (16). The variable (PAR

𝑠) in (15)

is specified in (13). Based on (9), (10), (13), and (14), the Costfunction can be written as in

Cost

= [(10−6(([(− 48104 (PAR𝑡− (31.2 exp(−((Time − 12)2.273 )2)))2) + 7(PAR

𝑡− (31.2 exp(−((Time − 12)2.273 )2))) − 0.41])))

⋅ (𝑎0+ 8∑𝑛=1

((𝑎𝑛cos 𝑛 ∗ Time3.879 + 𝑏

𝑛sin 𝑛 ∗ Time3.879 )))] (US$/m2Hour) .

(22)

Finally, based on (20) and (22), the objective function ispresented in

Objective Function = [0.222PAR𝑡exp (1 − PAR

𝑡/56.1)

576 ]

− [(10−6(([(− 48104 (PAR𝑡− (31.2 exp(−((Time − 12)2.273 )2)))2) + 7(PAR

𝑡− (31.2 exp(−((Time − 12)2.273 )2))) − 0.41])))

⋅ (𝑎0+ 8∑𝑛=1

((𝑎𝑛cos 𝑛 ∗ Time3.879 + 𝑏

𝑛sin 𝑛 ∗ Time3.879 )))] (US$/m2Hour) .

(23)


STARTFor Time = 1 to 24 step 1

For PAR𝑡= 0 to 151.3 step 0.01

Calculate 𝐹objCompare New 𝐹obj with Old 𝐹objIf (New 𝐹obj > Old 𝐹obj)

Keep New 𝐹objElse

Keep Old 𝐹objNext PAR

𝑡𝐹obj (Time) = New 𝐹objNext Time

END

Pseudocode 1: Pseudo code for iterative search.

Our research attempts tomaximize the profit from cherrytomato cultivation in the greenhouse environment by varyingthe supplemental light usage. The best solution will be theone that maximizes the objective function. In this study, weusedMATLAB software version 7.12.0 (R2011a).Wewrote twoMATLAB programs, one implementing iterative search (IS)and one implementing a genetic algorithm (GA). Both pro-grams were designed to find the best values for supplementallight in the greenhouse environment. Pseudocode 1 shows thelogic for the iterative search technique. Basically the methodsearches exhaustively, examining all possible combinationsof hourly intervals and amount of supplemental lighting andcomparing the values of the objective function obtained witheach combination.

In contrast to the exhaustive search employed by IS, GA isa heuristic search method which does not necessarily exam-ine all possible solutions. It is one of the most common usedapproaches for optimization [15]. A genetic algorithm con-sists of four important steps including evaluation, selection,crossover, andmutation. In the evaluation step, GAmeasuresthe fitness of each individual solution in the population andassigns a score to it. In the selection step, the GA randomlyselects individuals of the current population for developmentof the next generation. In the crossover step, GA combinestwo selected individuals at a crossover point, thereby creatingtwo new individuals. And finally in the mutation step, GArandomlymodifies the genes of an individual that is subjectedto a smallmutation factor and introduces further randomnessinto the population. This iterative process continues untilone of the possible termination criteria is accepted. Figure 8shows us a proposed GA flowchart.

Table 6 shows the parameters we used for the GA. TheNumber of Generations parameter indicates the stoppingcondition. In fact these are the default parameters for theMATLABGA toolbox.We experimentedwith changing someof these parameters, but this did not produce any better solu-tions. According to the proposed settings, the program runsfor at least 50 generations. After reaching the 50th generation,if the difference value of 10 consecutive generations is lessthan 10−6, the program will stop iterating. Otherwise theprogram runs for 100 generations and then stops.

Table 6: Setting parameters of GA in MATLAB toolbox.

Setting parameters of GA toolboxParameter SettingPopulation type Double vectorPopulation size 20 (5–100)Parent creation function Constraint dependentScaling function RankSelection function Stochastic uniformCrossover function Two pointsCrossover probability (𝑃

𝑐) 0.8 (0.1–0.9)

Mutation function Constraint dependentMutation probability (𝑃

𝑚) 0.2 (0.1–0.9)

Number of generations 100 (50–200)

3. Results and Discussion

As noted previously, supplemental artificial lighting in green-house environments can be used to increase production butfinding the optimum value of required supplemental light isvery important, especially when electricity pricing is done byRTPmethod.We calculated values for the values of optimumsupplemental light and wattage of HPS lamps in the green-house environment using two different approaches, namely,IS and GA. Figure 9 shows the optimum supplemental lightcalculated by means of IS.

Numerical values for optimum supplemental light andHPS lamps wattage during the hours of day and night aredisplayed in Table 8. In this table, results with IS and with GAare compared and evaluated. Moreover, in this table severalfactors related to the use of electrical lighting in greenhouseenvironment were calculated. Among them, profit of usingelectrical lighting is very important and should be considered.For this purpose, cost of electricity usage for supplementallighting is considered based on RTP of electricity in the smartgrid. Also enhancement of greenhouse revenue is calculatedbased on increased production of tomatoes. In fact, electricallighting increases tomato photosynthesis and, thus, tomatoproduction will significantly increase within the intendedtime period. The zeroes in the Table 7 show that using elec-trical lighting inside the greenhouse environment for tomatocultivation is not always practical, froman economic perspec-tive. This depends on the price of electricity in every period.Table 8 presents the improvement of production indices ingreenhouse environment considering use of electrical light-ing as the supplemental lighting for tomato cultivation. Thistable is calculated based on an industrial greenhouse witharea of 200,000 square meters and 120 days of cultivationperiod.

Table 8 illustrates that using the optimum electrical light-ing in the greenhouse environment under study has a greatinfluence on greenhouse production. The results, indeed,show that use of an optimal electrical lighting which is calcu-lated based on RTP of electricity in the smart grid system willimprove financial profit of the greenhouse significantly. Forinstance, in an industrial greenhouse with an area equalto 200,000 square meters, net profit of the greenhouse can


Tabl

e7:

Opt

imum

values

ofsu

pplem

entallight

,lam

pwattage

,inc

ome,

cost,

andfit

ness

valuef

orch

erry

tom

atocu

ltiva

tion.

Tim

e(H

our)

Supp

lem

ental

light

(mol/m2da

y)

HPS

lam

pswattage

(watt/m2)

Inco

me

(US$

/m2ho

ur)

Cos

t(U

S$/m2ho

ur)

Fitn

essv

alue

(US$

/m2ho

ur)

With IS

With GA

With IS

With GA

With IS

With GA

With IS

With GA

With IS

With GA

0-1

21.62

2214

8.68

64151

0.01

540.01

560.00

960.00

970.00

580.00

591-2

27.13

26.8

185.96

7018

40.01

750.01

740.00

950.00

940.00

800.00

802-

332

.51

32.6

222.08

6922

30.01

910.01

910.00

840.00

840.01

070.01

073-

431.32

31.7

214.12

15217

0.01

880.01

890.00

840.00

850.01

040.01

044-

531.27

31.6

213.78

3521

60.01

880.01

880.00

850.00

860.01

030.01

025-

633

.1932

.822

6.63

2422

40.01

930.01

920.00

840.00

830.01

090.01

096-

724

.1323

.25

165.70

5216

00.01

650.01

620.00

990.00

960.00

660.00

667-8

17.43

17.5

120.14

1712

10.01

410.01

410.00

880.00

890.00

530.00

528-

99.5

49.6

65.9131

670.01

200.01

210.00

560.00

560.00

640.00

659-

100.02

60.01

50

00

00

00

010

-11

00

00

00

00

00

11-12

00

00

00

00

00

12-13

00

00

00

00

00

13-14

0.02

70.01

50

00

00

00

014

-15

0.04

60.03

50

00

00

00

015-16

0.113

0.09

00

00

00

00

16-17

0.05

70.05

00

00

00

00

17-18

0.06

40.07

00

00

00

00

18-19

8.98

9.262

.062

964

0.00

800.00

820.00

670.00

690.00

130.00

1319

-20

4.91

4.8

33.843

333

0.00

470.00

460.00

430.00

420.00

040.00

0420

-21

1.02

1.16.72

507

0.00

100.00

110.00

080.00

090.00

020.00

0221-2

212

.07

11.9

83.380

782

0.01

020.00

950.00

820.00

750.00

200.00

2022

-23

22.68

22.4

155.88

1015

40.01

590.01

570.00

990.00

980.00

600.00

5923

-24

20.97

21.4

144.26

9214

70.01

510.01

530.00

950.00

970.00

560.00

56Av

erag

e0.00

860.00

8575

0.00

4854

20.00

4833

30.00

3745

80.00

3741

7


Start

Generate a population:x1, x2, . . . , xn

Calculation of the fitness function:f (x1) , f (x2) , . . . , f (xn)

YesIs terminationcriterion(s)

satisfied?

No

Selection the chromosomes for mating

Exchanging of parts of two selected chromosomes and offspringCreation using crossover probability

(Pm)

Insertion of the results in

Is the size of newpopulation equal to n?

Yes

No

Chromosome populationReplacing with new population

Stop

the new population

Changing of gene values in the offspring chromosomesusing mutation probability

(Pc)

Figure 8: Flowchart of genetic algorithm procedure.

Table 8: Comparison of production indices with and without electrical lighting for an industrial greenhouse with an area equal to 200,000square meters and 120 days of cherry tomato cultivation.

Natural light Natural light + Electrical Light Change percentageAverage light amount(mol/m2 day) 20.4 20.4 + 12.46263 = 32.86263 +61.1%

Tomato production amount(Tons) 1,708 1, 708 + 990.72 = 2698.72 +58%

Income of greenhouse(USD) 8,540,000 8,540,000 + 4,953,600 = 13,493,600 +58%

Electricity consumption forlighting (mega watt hour) — 49180.798 —

Cost of electricity forlighting (USD) — 2,796,019.2 —

Profit of greenhouse basedon using electrical lighting(USD)

— 2,157,580.8 —


0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Time (hour)

0

5

10

15

20

25

30

35

Optimal supplemental light

PAR

(mol

/m2

day)

Figure 9: Optimum supplemental light with IS.

Table 9: Comparison of computation times for optimization meth-ods.

Simulation subject ApproachComputation

time(seconds)

Supplemental light & HPSlamps wattage

With IS 6817.24With GA 98.4

be improved 2,157,580.8 USD using 49180.798MW elec-tricity within 120 days of cultivation. We assume that thecosts of greenhouse operation other than the electricity areapproximately constant with and without using of electricallighting. In the above case, the additional cost of electricity forsupplemental lighting is 2,796,019.2 USD but the profit fromthis electricity usage is 4,953,600 USD.

Note that the cost of electrical lighting equipment otherthan the lamps themselves is incurred only once. Then thisequipment can be used during the lifetime of the greenhouse.Also the lifetime of mentioned HPS lamps in this paper isabout 30,000 hours (1250 days). Thus, the cultivation time ofthe case study in this paper is lesser than 10%ofHPS lamps lifetime. Accordingly, the other costs involved in using electricallighting for cherry tomato cultivation in a greenhouse envi-ronment are not the significant fraction of total cost in thiscase study and can be ignored.

Computation times for the two approaches are displayedand compared in Table 9. Obviously, the GAmethod is muchmore efficient from a computational perspective. It requiresabout 100 seconds to find an optimal solution. In contrast,the IS method takes almost two hours.

4. Conclusions

In this study, optimization of greenhouse supplemental light-ing based onRTP in the electric energymarketwas investigat-ed. One of the key goals in advanced agriculture is increasedgreenhouse efficiency. Controlling the greenhouse climate isone of themost important issues in the industrial agriculture.By controlling and tuning greenhouse parameters such asartificial light, it is possible to significantly increase the green-house efficiency and crop cultivation performance. Deter-mining optimal supplemental light and then controlling the

greenhouse lighting based on this result can increase theeconomic efficiency of greenhouses.

We derived a mathematical model for cherry tomato pro-duction based on the literature. We modeled sunlight radi-ation based on the measured data in the greenhouse understudy, HPS lamps according to the manufacture’s data (GE),and RTP of electricity in the smart grid. The objective func-tion was obtained by combining the above models withempirical data.

Using the above models, we created an objective functionthat expressed the trade-off between the benefits to be derivedfrom artificial greenhouse lighting, in terms of increasedproduction and hence greater Income, and the costs, in termsof higher energy use. We then used two methods, iterativesearch and genetic algorithms, to determine the best amountand timing of artificial lighting.

The results verify that the profit in the case study in thispaper can be approximately doubled by employing electricallighting with the appropriate timing. Also the simulationresults show that each of the above-mentioned approacheshas some advantages and disadvantages. The main advantageof the iterative search approach is potentially higher accuracy,since it examines every possible solution. Its main disad-vantage is lower computational speed. In contrast, the mainadvantage of theGA approach is higher computational speed,which is obtained with the possible loss of accuracy, since GAis heuristic and may miss some solutions. On the other hand,the simulation results show that, in this problem, the resultsof the two approaches are very similar. Hence using GA forthe calculation method is justified in this situation.

Disclosure

This research article is an extended version of our conferencepaper published in the 10th International Conference onElectrical Engineering/Electronics, Computer, Telecommu-nications and Information Technology, 2013.

Competing Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper.

Acknowledgments

This study was supported by the King Mongkut’s Universityof Technology Thonburi (KMUTT). The authors thank SallyE. Goldin for reading the manuscript and corrections.

References

[1] H. Hu, L. Xu, R. Wei, and B. Zhu, “Multi-objective controloptimization for greenhouse environment using evolutionaryalgorithms,” Sensors, vol. 11, no. 6, pp. 5792–5807, 2011.

[2] K. P. Ferentinos and L. D. Albright, “Optimal design of plantlighting system by genetic algorithms,” Engineering Applicationsof Artificial Intelligence, vol. 18, no. 4, pp. 473–484, 2005.

[3] G. D. Pasgianos, K. G. Arvanitis, P. Polycarpou, andN. Sigrimis,“A nonlinear feedback technique for greenhouse environmental


control,”Computers and Electronics in Agriculture, vol. 40, no. 1–3, pp. 153–177, 2003.

[4] A. Brazaityte, P. Duchovskis, A. Urbonaviciute et al., “Theeffect of light-emitting diodes lighting on the growth of tomatotransplant,” Zemdirbyte—Agriculture Journal, vol. 97, no. 2, pp.89–98, 2010.

[5] M. Mahdavian, M. Bayati Poudeh, and N. Wattanapongsakorn,“Greenhouse lighting optimization for tomato cultivation con-sidering Real-Time Pricing (RTP) of electricity in the smartgrid,” in Proceedings of the 10th International Conference on Elec-trical Engineering/Electronics, Computer, Telecommunicationsand Information Technology (ECTI-CON ’13), Krabi, Thailand,May 2013.

[6] J. A. Nelson and B. Bugbee, “Economic analysis of greenhouselighting: light emitting diodes vs. high intensity dischargefixtures,” PLoS ONE, vol. 9, no. 6, Article ID e99010, 2014.

[7] M. Saad, A. Farij, A. Salah, and A. Abdaljalil, “Automatic streetlight control systemusingmicrocontroller.Mathematicalmeth-ods and optimization techniques in engineering,” inProceedingsof the 1st International Conference onOptimization Techniques inEngineering, pp. 92–96, Antalya, Turkey, 2013.

[8] M. Biabani, M. A. Golkar, A. H. Z. Kasiry, and M. Akbari,“Smart grid in Iran: driving factors, evolution, challenges andpossible solutions,” in Proceedings of the 10th InternationalConference on Environment and Electrical Engineering, pp. 1–4,IEEE, Rome, Italy, May 2011.

[9] M. Roozbehani, M. A. Dahleh, and S. K. Mitter, “Volatilityof power grids under real-time pricing,” IEEE Transactions onPower Systems, vol. 27, no. 4, pp. 1926–1940, 2012.

[10] H. Zareipour, A. Janjani, H. Leung, A. Motamedi, and A.Schellenberg, “Classification of future electricitymarket prices,”IEEE Transactions on Power Systems, vol. 26, no. 1, pp. 165–173,2011.

[11] A. Motamedi, H. Zareipour, and W. D. Rosehart, “Electricitymarket price forecasting in a price-responsive smart gridenvironment,” in Proceedings of the IEEE Power and EnergySociety General Meeting, Minneapolis, Minn, USA, July 2010.

[12] P. Vijayapriya, G. Bapna, and D. P. Kothari, “Smart tariff forsmartmeters in smart grid,” International Journal of Engineeringand Technology, vol. 2, no. 5, pp. 310–315, 2010.

[13] R. G. Anderson, Greenhouse Tomato Production Practices,Horticulture Opportunities Journal, University of Kentucky,College of Agriculture, 2002.

[14] P. H. B. De Visser, G. H. Buck-Sorlin, and G. W. A. M. van derHeijden, “Optimizing illumination in the greenhouse using a3D model of tomato and a ray tracer,” Frontiers in Plant Science,vol. 5, article 48, 2014.

[15] V. Kachitvichyanukul, “Comparison of three evolutionary algo-rithms: GA, PSO, and DE,” Industrial Engineering & Manage-ment Systems, vol. 11, no. 3, pp. 215–223, 2012.

[16] D. D. Piromalis, K. G. Arvanitis, P. G. Papageorgas, D. I. Tseles,and C. S. Psomopoulos, “LEDWIRE: a versatile networkingplatform for smart LED lighting applications using LIN-bus andWSNs,” Sensors & Transducers, vol. 200, no. 5, pp. 50–59, 2016.

[17] S. Barkley, Commercial Greenhouse Tomato Production: TomatoPlant Propagation, Agricultural and Rural Development ofAlberta Ministry, Edmonton, Canada, 2011.

[18] P. L. Joskow, “Creating a smarter U.S. electricity grid,” Journal ofEconomic Perspectives, vol. 26, no. 1, pp. 29–48, 2012.

[19] M. Dorais, “The use of supplemental lighting for vegetablecrop production: light intensity, crop response, nutrition, crop

management, cultural practices,” in Proceedings of the CanadianGreenhouse Conference, Toronto, Canada, 2003.

[20] R. H. Aalderink and R. Jovin, “Estimation of the photosynthe-sis/irradiance (P/I) curve parameters from light and dark bottleexperiments,” Journal of Plankton Research, vol. 19, no. 11, pp.1713–1742, 1997.

[21] A. Krumbein and D. Schwarz, “Grafting: a possibility toenhance health-promoting and flavour compounds in tomatofruits of shaded plants?” Scientia Horticulturae, vol. 149, pp. 97–107, 2013.

[22] B. Jones, Tomato Plant Culture, CRC Press, Boca Raton, Fla,USA, 1999.

[23] S. Leoni, B. Pisanu, and R. Grudina, “A new system oftomato greenhouse cultivation: high density aeroponic system(HDAS),”Acta Horticulturae Journal, vol. 361, pp. 210–217, 1994.

[24] T. Sung, S. Y. Yoon, and K. C. Kim, “A mathematical modelof hourly solar radiation in varying weather conditions for adynamic simulation of the solar organic rankine cycle,”Energies,vol. 8, no. 7, pp. 7058–7069, 2015.

[25] D. R. Myers, “Solar radiation modeling and measurements forrenewable energy applications: data andmodel quality,” Energy,vol. 30, no. 9, pp. 1517–1531, 2005.

[26] General Electric Company, Lamps and Lighting for Horticulture,Lighting for Growth Catalog, GEC, 2010.

[27] R. Kumar, P. D. Ray, and C. Reed, “Smart grid: an electricitymarket perspective,” in Proceedings of the IEEE PES Conferenceon Innovative Smart Grid Technologies (ISGT ’11), Anaheim,Calif, USA, January 2011.

[28] Agricultural Marketing Services, National Fruit and VegetableRetail Report, United States Department of Agriculture, 2013.

Submit your manuscripts athttps://www.hindawi.com

Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

MathematicsJournal of


Mathematical Problems in Engineering

Hindawi Publishing Corporationhttp://www.hindawi.com

Differential EquationsInternational Journal of

Volume 2014

Applied MathematicsJournal of


Probability and StatisticsHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

Journal of


Mathematical PhysicsAdvances in

Complex AnalysisJournal of


OptimizationJournal of


CombinatoricsHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

International Journal of


Operations ResearchAdvances in

Journal of


Function Spaces

Abstract and Applied AnalysisHindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

International Journal of Mathematics and Mathematical Sciences


The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014


Algebra

Discrete Dynamics in Nature and Society



Decision SciencesAdvances in

Discrete MathematicsJournal of

Hindawi Publishing Corporationhttp://www.hindawi.com

Volume 2014 Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014

Stochastic AnalysisInternational Journal of

Documents

Optimizing Greenhouse Lighting for Advanced Agriculture ...downloads.hindawi.com/journals/mpe/2017/6862038.pdf4 MathematicalProblemsinEngineering 0 5 10 15 20 Time (hour) 0 5 10 15