A supplementary lighting and regulatory scheme using a multi-wavelength light emitting diode module for greenhouse application

Abstract

This study proposes a supplementary lighting and regulatory scheme for application in semi-closed crop production systems. This supplementary lighting control module is primarily based on fuzzy logic inference and an expert database. Under the hypothesis that the temperature and humidity conditions are consistent, the system will implement supplementary lighting of different wavelengths according to current solar spectral distribution conditions and plant growth patterns. Simulations demonstrate that under the condition of full sunlight, the system still provides supplementary light to compensate for insufficient light spectrum. Compared to traditional plant lamps that need to be lit for long periods of time, this system can randomly adjust light spectrum and intensity to achieve energy savings, and is therefore suitable for application in lighting control systems in future greenhouses.

1. Introduction

In response to global climate change and environmental pollution, many scholars and experts have begun to adopt necessary measures to protect the environment, save resources, and develop techniques to ensure food safety. In recent years, due to advances in energy-saving techniques, auto-control techniques, robotic arm techniques, light emitting diode (LED) manufacturing techniques, material recycling and reuse techniques, and internet communication techniques, the development and establishment of an automated vegetable and fruit production factory is no longer an impractical dream. During the establishment of a given factory, the environmental factors that affect crop growth are rather significant.^1,2 Among these environmental factors, light source selection and control is one of the indispensible aspects of the production system. As demonstrated by the literature, it is already well known that the light cycle and spectrum distribution are of considerable importance to plant photosynthesis and morphological growth.³ In 1937 and 1956, scholars found that red light (640 nm–690 nm) would boost the growth of lettuce seed sprouts, while far-red light (710 nm–750 nm) would suppress it.^4,5 In fact, the ratio of light density between the two wavelength bands in terms of red light and far-red light has a great impact on seedlings.⁶ In addition, related studies have specified how various red and blue LED combinations have different effects on plant growth and morphological growth^7–17 and that red light stimulates the phytochromes in plant cells.

The interchange between the presence of red light (bright) and the absence of red light (dark) makes phytochromes affect plants by causing the stems and leaves of the plants to grow.¹⁸ Blue light sources (460 nm–470 nm) are also crucial to plants. The growth of plants, the absorption of nutrients, and the growth of sturdy stems are all affected by blue light. In addition, the phototropism normally observed in plants is also caused by blue light. Blue light sources also play an important role in plant differentiation and stomatal regulation. If blue light is insufficient while red light is oversupplied (Blue/Red = 10%), a plant’s stem will grow excessively and cause the leaves to become yellow.³ During a plant’s growth process, only lights of certain wavelengths are needed and not all wavelengths of light are effective. The wavelength, illumination conditions, and duration of light sources affect the size of medium-sized stems and leaves during the growth process.^19–21 Previous studies have shown that light in the green wavelength band affects the formation of carotene and also intensifies the metabolism of carbohydrates and nitrates,²² even as green light has no significant effect on crop growth patterns,²³ Green light has been shown, on the other hand, to have an effect on the colouring of leaves.²⁴

At present, many scholars are conducting studies on the effects of LED multichrome lighting on plant growth. It has previously been found that in the seedling stage, the use of LED light sources and adjustments to light quality and photosynthetic photon flux density (PPFD) will not only decrease energy consumption and costs enormously but also shorten the cultivation cycle and improve the quality of plants to a certain degree. Many studies have proven that with high-brightness LEDs, the growth of crops such as lettuce, strawberries, and Phalaenopsis, among others, will be affected.^9,13,25,26

Sunlight is one of the vital elements for plant growth. However, the most efficient light spectral wavebands for photosynthesis in plants fall between 400 nm and 700 nm on the visible wavelength (see Figure 1).^7,27 It is also true that wave band intensities affect plant growth the most when red light and blue light are applied for plant photosynthesis. However, the three-dimensional cultivation techniques developed by the advanced greenhouse industry are insufficient, and the light intensities provided to plants by such methods are also insufficient.^28,29 Because the lower layer of a multi-layer cultivation frame is blanked by the upper layer, the average amount of sunlight of lower layer is less than that of upper layer. In fact, plants may be grown in vain as a result of various forms of sunlight radiation loss, such as reflection loss, transmission loss, and penetration loss, which can make plants unable to achieve maximum photosynthetic effects. Thus, effective supplemental lighting for plants is significant to produce high-quality crops.^30,31

Figure 1

Effect of different wavelengths on plant growth^7,27

In 2005, multiple combinations of LEDs of different wavelengths were used to design adjustable light sources to replace the current plant light source design, an advance which is well-suited to establishing manual supplementary lighting and regulatory systems in the three-dimensional agriculture of the future.³² However, Fryc et al.’s system still requires that an entire package of automatic adjustment mechanisms for the plant wavelengths be applied in the supplementary system.

The development of smart systems has been put into practice for many years. Commonly seen control methods currently rely on fuzzy logic inference and expert data to determine optimised parameters for a given system.^33–36 The expert system is composed of a knowledge database.³³ The user can revise the database to strengthen system performance. The expert system is adopted to solve complicated control problems. Fasanghari and Montazer³⁴ have demonstrated a fuzzy expert system for selecting superior stocks so as to meet the uncertainty of stock portfolio recommendations and to model the recommendation rules adopted by experts at the Tehran Stock Exchange for portfolio recommendations.

Kolhe et al.³⁵ proposed an integrated intelligent multimedia interface based on fuzzy logic for crop disease management. The system can provide a significant aid in crop management, giving quick solutions for plant pathologies. A multistaged fuzzy logic controller has been adopted to regulate plant growth.³⁶ This strategy can create the optimal growth conditions for plants at different growth stages so as to enhance the quality of plants. In this study, a fuzzy controller was used under fixed conditions of humidity and temperature and a constant cycle of light and darkness. The proposed system gives plant lighting in terms of red light, green light, and blue light at seedling stage based on leaf area size, leaf colour, and stem height, respectively. Meanwhile, supplementary lighting is performed given an insufficient amount of sunshine. In short, the supplementary lighting system was able to shorten the duration of growth by keeping the crop growing efficiently. Moreover, this efficiency allows energy savings to be achieved.

2. Materials and methods

2.1. An overview of fuzzy logic

A fuzzy logic control system is a digital control methodology that simulates human thinking by incorporating the imprecision inherent in all physical systems.^37,38 The main feature of the system is that expert knowledge can be transferred via a linguistic control strategy into an automatic digital control strategy while system control output can be obtained via a fuzzy logic inference.³⁹ In the control system, there are four principle components, namely, the fuzzification interface (FI), the decision making logic, the knowledge database, and the de-fuzzification interface (DFI).

The functions of the FI are three-fold: first, to input the variables for pre-handling (i.e. normalisation); second, to conduct scaling mapping, a measure to transfer the range of values for the input variables into corresponding universes of discourse; and third, to input data for the transformation of suitable linguistic values which could be seen as labels for fuzzy sets. The second of the aforementioned components, decision-making logic, is formed according to the fuzzy behaviour concept of humans in making decisions, and this decision-making is determined by employing fuzzy implications and the rules of inference in fuzzy logic. The third component is the expert knowledge database, which is incorporated with fuzzy database and linguistic control rules. The database is applied to supply definitions of fuzzy subsets of variables and membership function (MF) when planning. A linguistic control rule database is used to describe the control purpose and strategy through the standard rules established by experts in the field. The last of the four components is the DFI. It serves to transfer the values of the output variables into the corresponding universes of discourse, and then the fuzzy control actions inferred can be transformed into actual control actions. In the following section, the application of the fuzzy control strategy used in the automatic supplementary lighting system for plant cultivation will be introduced.

2.2. LED supplementary and regulatory system

Figure 2 presents a block diagram of the whole LED supplementary lighting system. It contains the fuzzy logic controller, the knowledge database, image processing and LED spectrum model establishment, and photometric unit transformation. The fuzzy logic controller was mainly used to determine the amount of red, green, and blue light necessary. The knowledge database contains plant growth pattern data and related data on the amounts of red, green, and blue light required. An LED spectrum model was used to transfer the driving current of the light source outputs from the fuzzy logic controller into their respective spectrum distribution functions. Irradiance unit transformation, on the other hand, was used to transform spectral radiant intensity into PPFD for the purpose of being in accordance with the light intensity standard for agriculture.^40,41 With image processing – specifically, through an image capture process used during the cultivation of the plants⁴² – the plant leaf area indexes (LAIs) and stem heights are calculated, while leaf colours are also judged according to the knowledge database, in order to determine the most suitable light spectrum and intensity. In the following section, how the fuzzy logic system for the supplementary lighting works will be introduced step by step.

Figure 2

The LED supplementary lighting system LED: light emitting diode; LAI: leaf area index; DF: de-fuzzification.

2.2.1. Selection of input/output variables

The seedling stage growth pattern database established in this study was based on previous data and related literature.^42,43 The light spectrum was determined by the fuzzy system, and lettuce was the experimental crop. The lettuce in the seedling stage was supplemented with different wavelength bands of light quality and light quantity and was then assessed by the system for the resulting effects. The input variables for the system were LAI, leaf hue, and stem height. It is known that red light and blue light affect the leaf area and stem height in the seedling stage of the plant, while green light stimulates the metabolism of nitrates and carbohydrates and causes the leaves to exhibit more colours. As a result, the output variables in the entire system apply three types of wavelength (red light, blue light, and green light) as the driving current values. The selections of the wavelengths can be determined by the required components of the plants that users wish to grow. An extra wavelength can also be added in the system. However, the addition of an extra wavelength will lead to more consumption of LED drive current, which results in more power consumption by the proposed system. Meanwhile, the total system complexity is enhanced due to the increase in the number of output variables. In the system, three LED wavelength bands, such as red, blue and green, are selected. The choice of colour is determined by the user. Afterwards, the LED driving current values of each wavelength should be substituted into equation (1), and the light intensity distribution function of the multi-wavelength LED is thereby formed. Since the single wavelength LED output is a narrow waveband, the bandwidth (full width at half magnitude) usually falls between 20 nm and 50 nm. According to the physical characteristics of the LED, under ideal circumstances, the physical characteristics of the spectrum can be obtained approximately by using a Gaussian distribution function.^44,45 As a result, it can be hypothesised that the three types of light quality function are all Gaussian distribution functions, and the functions can be merged by computing the three functions respectively and summing them. The equation involved is as follows:

ρ_{λ} = ρ_{0} {α_{1} I_{1} \exp [\frac{- 2 \ln 4}{η_{1}^{2}} (λ - λ_{p 1}) 2] + α_{2} I_{2} \exp [\frac{- 2 \ln 4}{η_{2}^{2}} (λ - λ_{p 2}) 2] + α_{3} I_{3} \exp [\frac{- 2 \ln 4}{η_{3}^{2}} (λ - λ_{p 3}) 2]}

(1)

where

ρ_{λ}

represents the spectral light intensity, while

α_{1}

α_{2}

, and

α_{3}

represent the light power output transformed coefficients of red, green, and blue light, respectively. Each coefficient can be deemed a constant, and it is assumed that when the driving currents are presented as I₁, I₂, and I₃_, respectively, of the LED wavelengths, the corresponding peak values

λ_{p 1}

λ_{p 2}

, and

λ_{p 1}

will remain the same without any shift, where

η_{1}

η_{2}

, and

η_{3}

represent the full width at half magnitude of light quantity intensity distribution functions of red, green and blue light, respectively. The function simulation figure and the actual experimental test data differentiation are shown in Figure 3, where the light intensity is normalised. Observation indicates that the outcome approximates the simulated shapes and values drawn by the corresponding colour-matched function in the International Commission on Illumination’s 1931RGB. In Figure 3, each light quality band is a little broader and can be adjusted and narrowed by adjusting

η_{1}

η_{2}

and

η_{3}

. In this paper,

η_{1} = 35 nm

η_{2} = 45 nm

and

η_{3} = 40 nm

, respectively.

Figure 3

Differential comparison of CIE 1931 colour-matching functions and theoretical model values in simulations.

Consequently, a LED multi-wavelength spectrum model is established based on the output of each corresponding light quality from the fuzzy system. Achieving this kind of LED light spectrum distribution is no longer a difficult task, thanks to the advances in LED manufacturing techniques.

2.2.2. Fuzzification interface

In fuzzification, the corresponding fuzzy subset amount of each single input variable and MF type must be decided. Since the plant growth pattern is curved instead of linear, Gaussian MFs and sigmoidal MFs are applicable for the system input MFs, and the equation is presented with vector x acting as function $μ_{A i} (x)$ as follows⁴⁶:

μ_{A i} (x) = {\exp (- \frac{1}{2} (\frac{(c_{i} - x) 2}{χ_{i}^{2}})) for Gaussian MF \frac{1}{1 + exp (- a_{i} (x - b_{i}))} for Sigmoidally - shaped MF

(2)

where c_i and

χ_{i}

are the centre and MF standard deviation, respectively, of the

i th

fuzzy set

A i

_; a_i and b_i are two parameters of sigmoid MFs. It has been taken into account that

\exp (- a_{i} (x - b_{i}))

in equation (3) tends to infinity or 0 when the absolute value of the exponent is higher than 5. Parameters are defined by the boundary conditions:

x {\leq | d_{max} |, μ_{A i} (x) ≅ 0 \geq δ | d_{max} |, μ_{A i} (x) ≅ 1

(3)

where δ is the coefficient defining the maximum increase of d_i. Thus,

a_{i} \geq 2 \exp (- a_{i} (x - b_{i})) / (d_{i} (δ - 1))

, and

b_{i} \geq d_{i} (δ + 1) / 2

. In

x = b_{i}

, the sigmoid MF has an inflection point.

The three input variables, $X_{R}^{n}$ , $X_{C}^{n}$ , and $X_{B}^{n}$ represent the LAI, leaf colour, and stem height, respectively, of the nth growing stage of the plant. The driving current value of red light $Z_{R}^{n}$ , the green light driving current value $Z_{C}^{n}$ , and the blue light driving current value $Z_{B}^{n}$ are the three system output variables. Each variable mentioned above corresponds to the three subsets in the fuzzy set, respectively. The set is formed by a Gaussian MF and two sigmoid MFs. The set label is defined as follows: Pale (P), Small (S), Low (L), Medium (M), Big (B), High (H), and Dark (D). See Figure 4 for each input or output MF set range and subset parameter.

Figure 4

Input and output membership functions of fuzzy logic controller: (a) membership function of LAI fuzzification; (b) membership function of leaf colouration fuzzification; (c) membership function of stem height fuzzification; (d) membership function of fuzzy logic controller output LAI: leaf area index.

2.2.3. Decision-making logic and knowledge database

After the selection of input/output variables and the setup of the MF parameters, the fuzzy rule database can be established. The rule database was completed with the aid of the expert discourse and rules of thumb. The entire rule chart is shown in Table 1. The inference rule is applicable to the seedling stage of the lettuce. The fuzzy inference discourse is based on the Mamdani model,⁴⁶ as shown below:

Linguistic control rule: If $X_{R}$ is $G_{R}^{i}$ and $X_{B}$ is $G_{B}^{i}$ and $X_{C}$ is $G_{C}^{i}$ then $Z_{R}$ is $G_{μ, R}^{i}$ and $Z_{B}$ is $G_{μ, B}^{i}$ and $Z_{C}$ is $G_{μ, C}^{i}$

Table 1

The fuzzy inference and rules

Stem height		LAI
			S			M			B
			Leaf colouration
			D	M	P	D	M	P	D	M	P
H	(Red)	H	H	H	M	M	M	L	L	L
	(Green)	L	M	H	L	M	H	L	M	H
	(Blue)	L	L	L	L	L	L	L	L	L
M	(Red)	M	H	H	M	M	M	L	L	L
	(Green)	L	M	H	L	H	H	L	M	H
	(Blue)	M	M	M	M	M	M	M	M	M
L	(Red)	L	H	H	M	M	M	L	L	L
	(Green)	L	M	H	L	M	H	L	M	H
	(Blue)	H	H	H	H	H	H	H	H	H

where

X

Z

, and Π are the linguistic variables and

G i

and

G_{μ}^{i}

are the corresponding linguistic labels characterised by MFs. The criteria of rule depend on the database provided by the expert. The relationship of plant growth with different LED light spectra is established in Table 1.^3,18,24 The setup standard of the rule database is according to the principles that leaf area size is related to red light, leaf colour is connected to green light, and stem height is associated with blue light. It is known from the above that the knowledge database represents the linguistic control rule database of the fuzzy system (as shown in Table 1), which could also be called the decision-making logic. As long as the responses to different situations are properly stated, the system can judge by itself which outcome should be presented.

2.2.4. De-fuzzification interface

After performing the decision-making logic, the system will receive three output fuzzy sets. The three different wavelength current output values will therefore be obtained by using the DFI. The de-fuzzification methods applied in this study were the centre of gravity and mean of maxima methods. The current output value of each light quality set can be computed by Equation (4), as shown in the following:

Ψ_{p} = \frac{\sum_{j = 1}^{k} μ_{A i} (x_{j}) \cdot x_{j}}{\sum_{j = 1}^{k} μ_{A i} (x_{j})}

(4)

where

x_{j}

is the numerical value of the

j th

output MF,

k

is the number of controller output sets, and

μ_{A i} (x_{j})

denotes its membership value at the

j th

quantisation level. The current value

Ψ_{p}

at the

p th

time step is computed using Equation (4), and then the current value

Ψ_{p + 1}

for the next time can be obtained.

2.2.5. LED irradiance unit transformation

After finishing the de-fuzzification step, the system will yield each light quality current value output. The current values should be then substituted into Equation (1) to obtain the spectral intensity of each light. However, the growth rate of the plant is affected by photosynthetically active radiation, specifically, by the PPFD (the customary unit of which is $μ mol \cdot sec - 1 \cdot m - 2$ ), which consists of photons falling between 400 nm and 700 nm on the spectrum band. As a result, a transformation formula is required to transfer the radiometric into photon flux density. The solar constant defined by World Meteorological Organization denotes 1367 $\pm 7 W \cdot m - 2$ . In full sunshine, the solar constant approximates 500 $W \cdot m - 2$ near the ground, which equals 2200 $μ mol \cdot s - 1 \cdot m - 2$ ⁴¹. When transferred to quantum flux density 38.4% of 2200 $μ mol \cdot s - 1 \cdot m - 2$ is distributed over the wavelength band 400 nm–700 nm where photosynthesis is conducted.⁴⁷ Converted into photosynthetic photon flux, the amount is approximately 845 $μ mol \cdot s - 1 \cdot m - 2$ . Photon flux density and irradiance ${\bar{ρ}}_{λ}$ can be shown in an equation as follows:

{\bar{ρ}}_{λ} ≅ 6.022 \times 1017 (R κ φ) \frac{\sum_{k} M_{λ_{k}} Δ λ}{\sum_{k} λ_{k} M_{λ_{k}}} (W \cdot m - 2) ≅ 119.8 (R) \frac{\sum_{k} M_{λ_{k}}}{\sum_{k} λ_{k} M_{λ_{k}}} ≅ α R

(5)

where

1 μ mol = 6.022 \times 1017

photons,

φ = 3 \times 108 m / sec

denotes the speed of light, κ represents Planck’s constant,

R

indicates the photon flux density measured with a quantum sensor (Field Scout® 3414F/3415F; Spectrum Technologies, Inc.), the constant α is a transformation coefficient, and the standard radiation output

M_{λ_{k}}

represents the number k midpoint wavelength λ between the wavelength bands 400 nm and 700 nm, where

Δ λ = 1 nm

. In this study, the transformation coefficient of red light was 5.31, the transformation coefficient of blue light was 7.22, and the transformation coefficient of green light was 5.83. Assume the red light irradiance to be 20

W \cdot m - 2

, for example: the PPFD value can be computed, and its result is 106.2

μ mol \cdot s - 1 \cdot m - 2

. With a consistent unit, the PPFD total of the natural light sources and supplementary light sources can be computed. Note that the method in this study does not limit the maximum total photon density value since only the required supplementary lighting quantity that the growth pattern of the plants requires was considered.

2.3. Image processing techniques

This system utilised two sets of video cameras to capture photos of the plant side views and leaf areas. The main purpose of this was to observe the growth height and to calculate the LAI. The technique used here was to transfer the RGB pixels of the photos into a gray scale image, and then conduct binary image processing.⁴⁸ Afterwards, the component labeling method was used to locate eight neighbouring objects and calculate each plant’s leaf area and stem height. Meanwhile, the noise was also filtered.⁴⁹ Finally, the leaf area of the each unit area (i.e. the LAI) was calculated; this value can also be obtained through manual measurement and it is closely related to the final volume of production.⁵⁰ The LAI can be calculated with the following equation:

LAI = 0.75 (ψ / q) \sum_{v = 1}^{V} \sum_{u = 1}^{U} (Λ_{uv} \times Ω_{uv})

(6)

where

Λ_{uv}

denotes the leaf length of each leaf of each plant and

Ω_{uv}

indicates the broadest width of the leaf. In Equation (6), U represents the total leaf number in the

v - th

plant,

q

is the measured number of plants, and ψ represents the planting density. The LAI is an appropriate dynamic index to reflect the crop group size. In a certain range, the volume of production is raised as the LAI increases. However, when the LAI increases over a certain range, the illumination becomes insufficient, the plants in the farming field become too crowded, and the photosynthetic efficiency becomes weaker, such that the volume of production decreases. Usually the maximum LAI for vegetables and fruits is around 5, and it is ideal when the LAI remains between 3 and 4.^51,52

As for determining the leaf colours, there are various methods, including direct visual inspections, the colorimetric method⁵³ the chlorophyll metre method,⁵⁴ and the spectrophotometry method,⁵⁵ among others. The most convenient and fastest method is direct visual inspection, and its correlation with the chlorophyll content of plants is the highest. Chlorophyll content can be assessed by a linear equation; therefore, for leaf colour determination, the direct visual inspection method was adopted. The colours of leaves can be divided into brownish-yellow (level 1), light green (or greyish-green; level 2), green (level 3), dark olive green (level 4) and dark green (level 5). The leaf colour of a given leaf is categorised into one of these five levels (1–5).^56,57 Usually if the leaf colour is dark green, it indicates that the leaf is rich in nitrogen fertiliser, or it may imply that the leaf absorbs too much green light.⁵³ As a result, in leaf colour determination, it can be said that a green or dark olive green leaf represents an ideal condition.

3. Numerical example

3.1. Establishing the plant photo-physiology database

To determine the correctness of the flux density adjustment of this system, a database concerning the relations between the spectrum and the growth patterns of plants needed to be set up first before running the system simulation. Therefore, a 13-day crop cultivation experiment was conducted with a combination of natural light and LED light in the mini greenhouse in the Department of Biomechatronics Engineering building at National Pingtung University of Science and Technology. The target plant is garden lettuce, where the seed source is provided by the Know-you Seed Co. Ltd. Hydroponic nutrient solution include 1.8 g ammonium nitrogen, 8.7 g carbon-nitrogen nitrate, 9 g water-soluble phosphoric acid and 28.5 g water-soluble potassium hydroxide, which totally equals 48 g. One litre of water is added for dilution. Each planting bed is 90 cm in length, and 45 cm in width. The cycle time of the pump in the circulating system is 15/45 (on/off) minutes. The CO₂ concentration is between 600 ppm and 800 ppm, temperature is between 21 $\circ C$ and 24 $\circ C$ , and humidity is between 60% and 75% upon the establishment of the database. The experimental plants were divided into six groups. One group was given only natural light. The other five groups were given artificial supplementary lighting with various ratios of light quality and intensity in three wavelength bands. The desired light spectrum depends upon the amount and spectrum of sunshine. The system will supplement insufficient light quality of that wavelength band, which results in a different light spectrum. The nutrient solutions used for all the crops were the same. The spectrum distribution of the natural light and the artificial light for all the groups was recorded every 30 min daily. The growth pattern of the plants was filmed with a video camera. Image analysis was used to calculate the LAI, leaf colour, and stem height of each plant throughout the growing process. For example, Figure 5 illustrates the steps of calculating leaf area of lettuce using the image identification method; step 1: image capture; step 2: remove the noise of image; step 3: use the RGB colour space method to remove the areas of non-specific colour of plant; step 4: remove the RB colour space; step 5: find the possible area of plant leaf; step 6: perform the erosion and expansion; and step 7: calculate the leaf area of plant. The results of the image process are shown in Figure 5.

Figure 5

(a) Original image; (b) remove the noise of image; (c) use the RGB colour space method to remove the areas of non-specific colour of plant; (d) remove the RB colour space; (e) find the possible area of plant leaf; and (f) perform the erosion and expansion

Figure 6

Light supplementation of the system under three different sunlight intensities: (a) full sunlight; (b) shaded by dark clouds; and (c) mostly shaded or when sunlight intensity is low

Figure 7

Light supplementation at different times within one day

All the growth pattern analyses were calculated over the average of 30 plants, excluding plants with extreme growth pattern variations. To enable comparison and verification, a combination of three types of LED light spectrums was also tested, where the wavelength of the red light was $λ_{p 1} = 625 nm$ , the wavelength of the blue light was $λ_{p 2} = 470 nm$ , and the wavelength of the green light was $λ_{p 2} = 525 nm$ . The main purpose of this experiment was to set up a database for the verification of the fuzzy supplementary lighting system simulation results, so that the rule base of the fuzzy system can be adjusted whenever necessary and the system performance can thus be optimised.

3.2. Simulation process

The main purpose of this part of the study was to determine the feasibility of the proposed mechanism and to analyse and evaluate the effects of the said mechanism. The focus here was on the light distribution achieved once the system had started to emit supplementary lighting. When the system was offline, the steps below were followed in sequence for internal system verification, with the light–dark cycle, temperature, humidity, and uniformity of the distribution of light quality and intensity remaining consistent.

Step 1: First, Lactuca sativa Linn. (garden lettuce) is chosen as the subject, and a fuzzy MF set up has begun, as shown in Figure 4.

Step 2: Then, a database is established on the relations between the growth pattern of plants and various light qualities and intensities (Table 1). At the same time, the spectrum database of plant growth is also established.

Step 3: The current growing day of the lettuce is recorded and compared with the corresponding ideal plant growth pattern data (including LAI, stem height, and leaf colour) in the database. The spectrum distribution and the growth pattern of the plant growing in a human-controlled environment is then compared with the data in the database, calculating the variations and inputting the growth pattern variation data and the spectrum distribution variation data into the fuzzy system. Based on these variation data, the fuzzy system can then decide whether or not supplementary lighting is required and, if so, what type of supplementary lighting is required.

Step 4: The parameters of the ideal spectral function are determined, including spectrum wavelength band and full width at half magnitude, as well as the transformation coefficient for the current and optical power output (as in equations (1) and (5)). Based on these functions, the system can then adjust the current for the spectrum wavelength band to achieve the ideal spectrum distribution.

Step 5: The three output values of the system are then transformed into a photosynthetic photon flux value (these transformation steps have already been explained above). After the spectrum of the natural light source is compiled, the complete spectrum distribution of the light received by the plant can be drafted out and the total PPFD value can be calculated.

Step 6: Step 3 is then returned to, and the correctness of the supplementary lighting of the system is tested.

As seen in Figure 6, it was determined that when there are changes in the natural light spectrum, the system can promptly provide supplementary lights of different intensities to achieve the ideal spectrum distribution. Figure 6(c) shows that the supplementation of green light is relatively low because the plant requires a low amount of green light during photosynthesis. Once the offline simulation test was verified, the online simulation could then be conducted.

4. Experiment

4.1. Method

During the online system simulation process, data collected from a one-day experiment and a one-week experiment were analysed. The main purpose of the one-day experiment was to explore the light supplementation of the fuzzy system at different time points in the day. The purpose of the one-week experiment was to explore the growth pattern variations of plants with and without light supplementation. In the simulation system, the light intensity was adjusted every 30 min. The air temperature, relative humidity, and CO₂ levels for all treatment were maintained at 22.3 $\pm 1.2 \circ C$ , 70% $\pm 7.5 %$ , and 680 $\pm 80 μ mol \cdot mol - 1$ , respectively. A data collector and handheld LED spectrometer are employed real time to record the above environment data and light spectrum, respectively (WatchDog 1650 Micro Station, Spectrum Technologies, Inc; Leaf Wetness Sensor, Spectrum Technologies, Inc; CO₂ Sensors, Spectrum Technologies, Inc; Handheld LED spectrometer, ConvergEver, Inc.).

4.2. Results

Figure 7 shows the one-day light supplementation. The real-time simulation of the system involved activity at 10:00 a.m., 2:00 p.m. and 6:00 p.m. At these time points, light supplementation was observed. From the spectrum distribution shown in Figure 7, it can be seen that the intensity of the light in the red wavelength band was relatively weak in the morning, the intensity of the light in the blue wavelength band was relatively weak in the afternoon, and the intensity of all the lights in the entire spectrum was weak in the evening. Although the intensity of all the lights in the entire spectrum was weak, the system was able to supplement all the light in the various wavelength bands.

This system was designed, in fact, to supplement different light qualities and intensities at different time points of the day for the three wavelength bands which are vital for the photosynthesis of plants. Compared with the traditional system, which is unable to adjust light intensity, this system is more efficient and requires less energy consumption. Figure 8 shows the spectrum distribution of the supplementary lighting at the three different time points of the day. It clearly shows that the amount of supplementary light required in the evening is higher than at other time points. Figure 9 shows the variations between the total amount of light supplementation provided by the proposed system and by the traditional system over one day. Findings show that the system proposed in this study provides the right amount of light supplementation at different time points of the day and is able to reduce energy consumption. A separate system performance test was also conducted, in which the growth period of the plant was extended to 11 days (the actual simulation time of the system is 5 min). The growth pattern of each plant was analysed. The data in Figure 10 shows the changes in the daily growth pattern of the plants. Recordings of the findings were begun from day 3, when the seeds germinated. The growth pattern information for the plants without supplementary lighting was derived from the results of the experiments conducted in the mini greenhouse earlier. As a complete database for the relations between the growth pattern and the light quality had already been established, the output value of the fuzzy system could be compared with the growth pattern of the corresponding plants in the database. The plant growth information in the database can serve as a reference for the supplementary lighting quality required by the fuzzy system (see Figure 10).

Figure 10

Variations in plant growth patterns (day 3 to day 13): (a) leaf area index; (b) leaf colour; and (c) height of stem

Figure 8

Supplementary light spectrum at different time ps within one day

Figure 9

Total light supplementation (PPF) at different times within one day PPF: photosynthetic photon flux.

Figure 10 shows that without light supplementation, the LAI was relatively low and the hypocotyl length of leggy seedlings was significantly shorter. There were no significant variations in leaf colour with or without light supplementation as the photosynthesis of a plant during the seedling phase requires a minimum amount of light in the green wavelength band. Therefore, this wavelength band has low impact on chlorophyll production and will not cause rapid changes in leaf colour. Table 2 demonstrates the results of the physiological measurements from all treatments. Plant tissue samples were dried in a drying oven for 3 hours at 80℃ before weighing. As can be seen, the use of the proposed method had the greatest fresh shoots and dry weight. Meanwhile, the plants with the greatest height of stem were grown with the proposed method. Although the mechanisms of changes in phytochemicals under different supplemental light quality are not well known, the results shown that supplemental light quality could be employed to enhance growth of lettuce grown under red, green, and blue light.

Table 2

Influence of light quality on plant height, root length, number of leaves, shoot dry weight, and shoot fresh weight at 13th day after sowing (DAS)

Parameters	Without (W/o) proposed scheme	With proposed scheme
Lettuce treatments (13 DAS)
Plant height (cm)	5.16 ± 0.18	5.89 ± 0.12
Root length (cm)	2.12 ± 0.21	2.45 ± 0.16
Number of leaves	5 ± 1	5 ± 1
Shoot dry weight (g/plant)	4.95 ± 0.24	5.18 ± 0.21
Shoot fresh weight (g/plant)	11.4 ± 0.6	11.9 ± 0.5
Canopy leaf temperature ( $\circ C$ )	22.3 ± 1.2
CO₂ ( $μ mol \cdot mol - 1$ )	680 ± 80

5. Conclusion

This paper proposes an intelligent light supplementation and regulatory system which can be used in modern greenhouse crop cultivation. The results indicate that this system can, with a fuzzy logic inference and the guidance of experts, effectively overcome light adjustment inefficiency in a semiclosed crop production systems in a greenhouse. Optimum light quality and the right amount of light can be supplied with this system. Our simulations showed that even under the condition of full sunlight, the system can still provide supplementary light to make up for insufficient light of certain wavelength bands within the sunlight. This improves the growth patterns of the plants and reduces waste of power resources. In the future, carbon dioxide control technology and an automatic LED lamp lift system may be integrated into this system for optimum agricultural growth.

Footnotes

Funding

This work was supported by the National Science Council, Taiwan, ROC, under grant nos. NSC 101-2622-E-020-002-CC2 and NSC 102-2622-E-020-006-CC3.

References

Asian Vegetable Research and Development Center. Vegetable Production Training Manual 1990: Shanhua, Tainan: AVRDC, pp. 442–442. [Reprinted 1992].

Adams

Rosenzweig

Peart

Ritchie

Mccarl

Glyer

Curry

Jones

Boote

Allen

Jr . Global climate change and US agriculture. Nature 1990; 345: 219–224.

Kendrick RE, Kronenborg GHM, editors. Photomorphogenesis in Plants, 2nd Edition. Dordrecht, Netherlands: Kluwer Academic Publishers, 1986.

Lewis

McAlister

. Wavelengths of Radiation in the Visible Spectrum Promoting the Germination of Light Sensitive Lettuce Seed, Paper no. 3414. Washington, DC: Smithsonian Miscellaneous Collections, 1937, pp. 96–96.

Toole

Hendricks

Borthwick

Toole

. Physiology of seed germination. Annual Review of Plant Physiology 1956; 7: 299–324.

Hendricks

Borthwick

Downs

. Pigment conversion in the formative responses of plants to radiation. Proceedings of the National Academy of Sciences 1956; 42: 19–26.

Balegh

Biddulph

. The photosynthetic action spectrum of the bean plant. Plant Physiology 1970; 46: 1–5.

Akoyunoglou

Anni

Blue light effect on chloroplast development in higher plants. In: Senger

(ed.) Blue Light Effects in Biological Systems, Berlin: Springer-Verlag, 1984, pp. 397–406.

Okamoto

Yanagi

Takita

. Development of plant growth apparatus using blue and red LED as artificial light source. Acta Horticulturae 1996; 440: 111–116.

10.

Goins

G.D

Yorio

Sanwo

Brown

. Photomorphogenesis, photosynthesis, and seed yield of wheat plants grown under red light-emitting diodes (LEDs) with and without supplemental blue lighting. Journal of Experimental Botany 1997; 48: 1407–1413.

11.

Yorio

Wheeler

Goins

Sanwo-Lewandowski

Mackowiak

Brown

Sager

Stutte

. Blue light requirement for crop plants used in bioregenerative life support system. Life Support Biosphere Science 1998; 5: 119–128.

12.

Yorio

Goins

Kagie

Wheeler

Sager

. Improving spinach, radish and lettuce growth under red light emitting diodes (LEDs) with blue light supplementation. HortScience 2001; 36: 380–383.

13.

Nhut

Takamura

Watanabe

Tanaka

. Artificial light source using light-emitting diodes (LEDs) in the efficient micropropagation of spathiphyllum plantlets. Acta Horticulturae 2005; 692: 137–142.

14.

Jao

Fang

. Effects of frequency and duty ratio on the growth of potato plantlets in vitro using light-emitting diodes. HortScience 2004; 39: 375–379.

15.

Takemiya

Takahashi

Shimazaki

. Leaf temperature reduction by blue light-dependent stomatal opening. Cryobiology and Cryotechnology 2007; 53: 1–5.

16.

Avercheva

Berkovich

Erokhin

Zhigalova

Pogosyan

Smolyanina

. Growth and photosynthesis of Chinese cabbage plants grown under light-emitting diode-based light source. Russian Journal of Plant Physiology 2009; 56: 14–21.

17.

Johkan

Shoji

Goto

Hashida

Yoshihara

. Blue light-emitting diode light irradiation of seedling improves seedling quality and growth after transplanting in red leaf lettuce. HortScience 2010; 45: 1809–1814.

18.

Pons

Seed responses to light. In: Fenner

(ed.) Seeds: The Ecology of Regeneration in Plant Communities, 2nd. Wallingford: CAB International, 1992, pp. 259–284.

19.

Bula RJ, Tennessen DJ, Morrow RC, Tibbitts TW. Light emitting diode as a plant lighting source: International Lighting Controlled Environments Workshop, Florida, USA, 27–30 March 1994, 255–267.

20.

Miyashita

Kitaya

Kozai

. Effects of red and far-red light on the growth and morphology plantlets in vitro: using light emitting diode as a light source for micropropagation. Acta Horticulturae 1995; 393: 189–194.

21.

Matsuda

Ohashi-Kaneko

Fujiwara

Goto

Kurata

. Photosynthetic characteristics of rice leaves grown under red light with or without supplemental blue light. Plant Cell Physiology 2004; 45: 1870–1874.

22.

Urbonavičiūte

. Effect of short-wavelength light on lettuce growth and nutritional quality. Sodininkyste ir Daržininkyste 2007; 26: 157–165.

23.

Folta

Maruhnich

. Green light: a signal to slow down or stop. Journal of Experimental Botany 2007; 58: 3099–3111.

24.

Kim

Goins

Wheeler

Sager

. Green-light supplementation enhances lettuce growth under red- and blue-light-emitting diodes. Hortscience 2004; 39: 1617–1622.

25.

Bula

Morrow

Tibbitts

Barta

Ignatius

Martin

. Light-emitting diodes as a radiation source for plants. Hortscience 1991; 26: 203–205.

26.

Hoenecke

Bula

Tibbitts

. Importance of “Blue” photon levels for lettuce seedlings grown under red-light-emitting diodes. Hortscience 1992; 27: 427–430.

27.

Plant Factory Laboratory. Light source for plant cultivation. Retrieved 20 March 2012, from http://www.sasrc.jp/kougen.htm [in Japanese].

28.

Lopez

Runkle

. Photosynthetic daily light integral during propagation influences rooting and growth of cuttings and subsequent development of New Guinea impatiens and petunia. HortScience 2008; 43: 2052–2059.

29.

Runkle

Warner

. Timing and duration of supplemental lighting during the seedling stage influence quality and flowering in petunia and pansy. HortScience 2010; 45: 1332–1337.

30.

Torres

Lopez

. Photosynthetic daily light integral during propagation of Tecoma stans influences seedling rooting and growth. HortScience 2011; 46: 282–286.

31.

Currey

Hutchinson

Lopez

. Growth, morphology, and quality of rooted cuttings of several herbaceous annual bedding plants is influenced by photosynthetic daily light integral during root development. HortScience 2012; 47: 25–30.

32.

Fryc I, Brown SW, Ohno Y. Spectral matching with an LED-based spectrally tunable light source: Proceedings of the Fifth International Conference on Solid State Lighting, San Diego, CA, USA, July 31: 2005: paper no. 59411I-9.

33.

Waterman

. A Guide to Expert Systems, Boston: Addison-Wesley Longman Publishing, 1986.

34.

Fasanghari

Montazer

. Design and implementation of fuzzy expert system for Tehran Stock Exchange portfolio recommendation. Expert System Applications 2010; 37: 6138–6147.

35.

Kolhe

Kamal

Saini

Gupta

. An intelligent multimedia interface for fuzzy-logic based inference in crops. Expert System Applications 2011; 38: 14592–14601.

36.

Chang

Sie

. A multistaged fuzzy logic scheme in a biobotanic growth regulation system. HortScience 2012; 47: 762–770.

37.

Zadeh

. Fuzzy sets. Information and Control 1965; 8: 338–353.

38.

Zadeh

. A rationale for fuzzy control. Journal of Dynamic Systems-Transactions of the American Society of Mechanical Engineers 1972; 94: 3–4.

39.

Marks

II RJ

. Fuzzy logic technology and applications. IEEE Technology Update Series, IEEE, 1994, pp. 19–24.

40.

Sager

Edwards

Klein

. Light energy utilization efficiency for photosynthesis. Transactions of the American Society of Agricultural Engineers 1982; 25: 1737–1746.

41.

Thimijan

Heins

. Photometric, radiometric and quantum light units of measure: a review of procedures for interconversion. HortScience 1983; 18: 818–822.

42.

Chang CL, Huang YC, Hong GF, Chen JF, Shie JL, Li LY. Automatic LED lamps lift system with image/light quantum sensors for multi-layer cultivation frame: International Symposium on Machinery and Mechatronics for Agriculture and Bio-systems Engineering (ISMAB), Jeonju, Korea, June 18–20: 2012: P-BR4, 564–569.

43.

Weeplian T, Ho YS, Chang CL, Lin CF. Effect of red and blue LED lighting on vegetable seedling production under multilayer cultivation: 7th Taiwan-Thailand Bilateral Conference, Pingtung County, Taiwan, 18–19 October 2012.

44.

Jones

. The calculation of the luminous efficacy of a semiconductor lamp. Journal of Physics D: Applied Physics 1973; 7: L80–L83.

45.

. Calculation of luminous efficacy, and both power and quantum efficiencies of LED by direct integration of Gaussian functions. Acta Physica Sinica 1978; 27: 691–699.

46.

Lee

. Fuzzy logic in control system: fuzzy logic controller, Part I and II. IEEE Transactions on System, Man and Cybernetics 1990; 20: 404–435.

47.

Nobel

. Physicochemical and Environmental Plant Physiology, London: Academic Press, 1991.

48.

Otsu

. A threshold selection method from gray level histogram. IEEE Transactions on Systems, Man and Cybernetics 1978; 9: 62–66.

49.

Gonzalez

Woods

. Digital Image Processing, New York: Addison Wesley, 1992.

50.

Chen

Black

. Defining leaf area index for non-flat leaves. Agricultural and Forest Meteorology 1992; 57: 1–12.

51.

Tei

Scaife

Aikman

. Growth of lettuce, onion, and red beet. 1. Growth analysis, light interception, and radiation use efficiency. Annals of Botany 1996; 78: 633–643.

52.

Andriolo

da Luz

Witter

Godoi1

RDS

Barros

Bortolotto

. Growth and yield of lettuce plants under salinity. Horticultura Brasileira 2005; 23: 931–934.

53.

Housecroft

Constable

. Chemistry: An Introduction to Organic, Inorganic, and Physical Chemistry, London: Pearson Education, 2006.

54.

Netto

Campostrini

Gonc

de Oliveira

Bressan-Smith

. Photosynthetic pigments, nitrogen, chlorophyll a fluorescence and SPAD-502 readings in coffee leaves. Scientia Horticulturae 2005; 104: 199–209.

55.

Gazula

Kleinhenz

Scheerens

Ling

. Anthocyanin levels in nine lettuce (Lactuca sativa) cultivars: influence of planting date and relations among analytic, instrumented, and visual assessments of color. Hortscience 2007; 42: 232–238.

56.

International Rice Research Institute (IRRI). Use of leaf colour chart (LCC) for N management in rice. Retrieved 27 September 2012, from http://irri.org/index.php?option=com_ k2&view=item&id=10980&Itemid=100331.

57.

Yang

Peng

Huang

Sanico

Buresh

Witt

. Using leaf color charts to estimate leaf nitrogen status of rice. Agronomy Journal 2003; 95: 212–217.