Design and Optimization of a Hybrid Solar–Wind Power Generation System for Greenhouses

Abstract

The climate crisis and energy price increases make energy supply a crucial parameter in the design of greenhouses. One way to tackle both these issues is the local production of energy from renewable sources. Since the permitted photovoltaic power installation on a greenhouse roof is limited by the need for an adequate amount of photosynthetically active radiation at the crop level, the necessity of designing a hybrid production system combining different renewable sources, storage systems, and conventional sources arises. The present work addresses the multifactorial problem of the optimal design (in terms of energy production quality, produced electricity price and CO₂ emissions) of a hybrid power generation system (photovoltaics/wind turbine/accumulators/oil generating unit) to meet greenhouse needs. The design accounts for the needs of production (for tomato cultivation) for different combinations of production and energy equipment (for microclimate management). Extended parametric studies for available solar and wind potential and energy demand are used to generalize the conclusions. Special attention is given to the contribution of various wind turbine sizes. The effect of greenhouse orientation and of photovoltaic modules arrangement on arched roofs is also examined and the different greenhouse energy systems are assessed in terms of energy cost and environmental footprint.

1. Introduction

Due to the climate crisis and the increase of energy prices, the issue of energy supply becomes a crucial parameter in the design of greenhouses. One way to tackle both the problem of climate change and the cost of energy is the local production of energy from renewable sources. For the design of a hybrid system, it is necessary to know the daily and annual variation of the load that it must cover, due to the fact that the availability of renewable energy potential is characterized by temporal variability. The interest in energy consumption in greenhouses became intense in the last decade; according to [1], in 2011 energy costs were the third most important cost of production in greenhouses, rising from 5.9% of costs in 2003 to 10%. Heating represents 70–80% of energy demand, other activities that require electricity consumption represent 10–15% and the remaining energy demand concerns transport. Last decade, numerous works were published regarding the energy consumption dependence of greenhouse construction materials [2], calculation of greenhouse energy consumption [3,4], and mathematical models of greenhouse energy consumption for cooling [5]. However, even in the most detailed and recent works, theoretical energy demand in time series is used [6].

The renewable energy source that is favored for greenhouses is solar because it produces electricity with photovoltaics (PV). On the one hand, solar energy is available in a wide range of latitudes, and on the other hand, greenhouses have large areas where photovoltaics can be placed without depriving arable land. This is why there exists a relatively extended literature regarding the power production from PVs cladding in greenhouse roofing [7], the development of the transient system simulation tool (TRNSYS) code for life cycle cost analysis [8], and the use of hybrid and organic PVs [9]. In the use of PVs on greenhouse roofs, there is always a tradeoff between power production and shading, as crop growth is affected by reduced availability of photosynthetically active radiation (PAR) and increased need for heating during winter. This is why many researchers have tried to determine the optimum percentage of roof coverage, maximizing the PV production without affecting the crop growth. Of course, this depends on the type of PVs and on the kind of crop. In [10], a 20% roof coverage was suggested, with semitransparent PVs of 8.25% electrical efficiency that would cause shading of 35–40%. According to [11], 47% coverage of the roof with semitransparent PVs with an electrical efficiency of 3.15% could satisfy the requirements for available PAR from May to September for a tomato crop.

However, precisely because of the limitations placed on installing PVs on the roof, PVs are often unable to meet the needs of the greenhouse. For this reason, the connection of the greenhouse to the grid or the combination of PVs with power storage units and/or other available renewable energy sources (RES) and/or conventional power production units, i.e., as part of hybrid power generation system, is necessary. The design of such systems has a dual purpose: on the one hand, the use of PVs on greenhouse roof do not reduce crop production; on the other hand, achieving the lowest final cost of energy produced with the smallest possible environmental footprint.

A common option is to use a combination of a geothermal heat pump with photovoltaics. In [12], it was found that covering 50% of the roof with conventional opaque PVs, they achieved 50% crop growth. During the summer, they achieved covering 33.2–67.2% of the energy demand for cooling and lighting. However, on a yearly basis, PVs produce 95.7% to 104.5% (depending on the kind of crop) of the total energy demand for heating, cooling and lighting. Of course, this can lead to a net-zero greenhouse only in the case of grid-connected units in terms of net metering. This is exactly the problem with PVs as well as other RES. While it is possible on an annual basis to produce the energy needed by the greenhouse, in practice, this energy cannot be utilized due to the different time profile of production and consumption unless the system has some energy storage capability (net metering, batteries, fuel cells). The basic concept is to use the excess PV power production to create hydrogen through electrolysis [13,14,15].

Another option to cope with mismatch of PV power production with energy demand and to the limitations of roof-installed PV capacity, avoiding negatively affecting the crop, is the use of a hybrid system. A common option for hybrid power systems is the addition of a wind turbine. Nevertheless, this is not a popular choice for greenhouses, either in studies or in practical application. Is not easy to study the use of wind turbines (WT) in a general case, since its performance depends on the local wind potential. While the solar potential directly depends on the latitude, its daily and annual variation can be approximated when the average values are known; this is not the case for wind potential. Wind potential average values are determined by other parameters. Nevertheless, for some countries, there exist wind maps where this information (the yearly average wind velocity) is available. However, even in these cases, the knowledge of yearly average wind velocity gives absolutely no information about the annual and daily variation of wind velocity (data necessary for the design of a hybrid system). This makes research difficult; even in areas with good wind potential where the use of WT would be efficient in practice according to studies with hybrid optimization of multiple energy resources (HOMER) [16], no such system has been built [17].

HOMER is the most commonly used software for the design and optimization of full hybrid systems (WT + PV + diesel generator + batteries) for various applications, such as a milking parlor and telecommunication station [18], an island [19], or a village [20]. For simpler systems, such as PV + WT only, simpler codes in MatLab [6,21] or genetic algorithms were developed [22] for the design of a hybrid system with economic and environmental criteria. Therefore, the majority of existing works concern theoretical mathematical models without insight into the physics of greenhouses and hybrid system units.

Therefore, the optimal design of a hybrid power generation system to meet greenhouse needs is a multifactorial problem. This means that, on the one hand, many factors must be taken into account in the design, and on the other hand, many elements must be determined in the optimization of the design with the aim to serve many objectives. The factors that must be taken into account are related to: (i) the load that must be met by the greenhouse (average values of consumed energy and their annual and daily variation patterns), which in turn depends on the type and technology used in the greenhouse; (ii) available renewable sources (also average values, annual and daily variation); (iii) net electricity and conventional fuel prices, as well as purchase, operation, and maintenance costs of hybrid system components. Regarding the design and optimization, the power and the operating characteristics of the components of the hybrid system should be determined, as well as its operation strategy. Finally, all this should simultaneously lead to reduced energy production costs and reduced environmental footprint, while ensuring the required level of energy supply security in the greenhouse. This work addresses this multifactorial problem, as detailed next.

In this work, a hybrid system with PV + WT + diesel generator (DG) + batteries is optimized for two technology levels of existing greenhouses (where the use of geothermal would have a high cost) with different time profiles of consumption, aiming to reduce energy costs and environmental footprint. In one greenhouse technology, the heating will be done by combustion, and in the other case, by electricity with a heat pump. Calculations will be made for different solar and wind potentials and daily load demand. Optimum greenhouse orientation will be examined, as well as the feasibility of using the greenhouse arched roof for PV installation in terms of energy production. The initial design of the systems will be done with semi-empirical methods, and the optimization will be done with HOMER. Elements of innovation in relation to existing jobs are: (a) Detailed determination of the time profile of energy demand with a combination of analytical models, existing literature and EnergyPlus; (b) Study of different wind potentials with identification of typical wind patterns at a European level; (c) Detailed calculation of PV energy production in arched roof; (d) The margin of failure of the system is taken into account, from zero to the average of the difference in demand for heating, cooling, ventilation and other electricity consumption; (e) The energy demand will also take into account other consumption besides heating, cooling and lighting; (f) In no work so far has the self-consumption of the hybrid system been taken into account—these are important and concern the self-consumption of the energy conversion equipment and, mainly, energy consumption for ventilation and cooling of the area where the batteries are installed. The ventilation of the battery room is required by international standards, while temperature regulation is necessary for the efficient operation of batteries and for the removal of the hydrogen produced. Finally, the two greenhouse technologies combined with hybrid systems are assessed in terms of total energy cost and environmental footprint.

2. Materials and Methods

For the optimization of a hybrid power system involving RES, two time series should be known: (a) the time series of the power demand and (b) the time series of the available RES potential (in our case solar and wind potential). First, the time series of the energy demand is assessed for two greenhouses technologies. Then, the time series of the available RES potential are formed. Following this, the characteristics of the components of the hybrid system are presented, along with the assumptions and the considered economics.

2.1. Energy Demand Assessment

Two greenhouse technologies are considered: (a) greenhouse using a conventional fossil-fuel burner for heating and evaporative pads for cooling, referred to as load type A, and (b) greenhouse using a heat pump for heating and cooling, referred to as load type B, hereafter. They are both low-tech greenhouses which are, however, representative Mediterranean constructions that are called upon to face the energy crisis without the possibility of radical modifications. In the calculations of hybrid systems, there is a general tendency to ignore the self-consumptions of the hybrid system. When the hybrid system includes batteries, then the energy consumption for the preservation of the battery room at the desired temperature is important (otherwise the battery efficiency and lifespan will be seriously affected); this work analytically accounts for this.

In both cases the energy demand for heating, cooling, irrigation, cold storage, fertigation, sorting and packaging and lighting are calculated according to [3] and [4] for a 200 m² greenhouse, sited in central Greece, with a volume of 1000 m³, a gothic roof, width 9.6 m, covered with double polyethylene with thermal transmittance, and U = 4 W/m²K for hydroponic tomato cultivation.

According to [3], the yearly energy demand for heating is 220 kWh/m² and 12 kWh/m² for cooling, with design temperatures of 14 °C during the night and 20 °C during the day. The temporal distribution of the demand during the year and during the day, considering the 15th of each month as a representative day, results from the corresponding change in the outside temperature. For irrigation, it is considered that 3 kWh per 10 m³ of water are required. The water consumption per day according to the month is presented in

Table 1. Installed power.

Greenhouse without Heat Pump		Greenhouse with Heat Pump		Hybrid System Lodge
Device	Power [W]	Device	Power [W]	Device	Power [W]
Cooling system (pump for evaporative pads and fans for air intake)	1470	Heat Pump	10,830	Lighting	108
Mixing fans	300	Mixing fans	300	Energy maintenance	26 W/4 W
Irrigation–Fertigation	645	Irrigation-Fertigation	645	Ventilation fan	2 × 30
Sorting–packaging	1000	Sorting-packaging	1000	Heat pump	1050
Storage	100	Storage	100
Lighting	250	Lighting	250
Total [W]	3865		14,695		1278

in [3]. The energy demand for fertigation is calculated at 45 kWh per year, considering one 0.1 kW sulphur evaporator operating 3 h every night for 150 nights. The energy demand for short-period storage is considered to be 1.2 kWh per day for 250 days during the year. The collection and packaging are considered to be 75 kWh/year, distributed over 150 days during the year. For the ventilation of the greenhouse, there are fans with installed power of 0.3 kW, with a yearly consumption considered at 820 kWh. Finally, the energy consumption for lighting in the greenhouse is considered, for an installed 0.25 kW lighting system operating during the night.

If the heating is performed with conventional fossil fuel burning, then the only burden on electricity consumption comes from the operation of the burner and the circulators and can be considered negligible. Since the cooling is done with evaporative pads, the energy consumption for the operation of a 0.37 kW water pump and fans with an installed power of 1.1 kW with a total yearly consumption of 2300 kWh [3], distributed throughout the year and the day according to external temperature. The installed electrical power is given in Table 1. The daily energy consumption for the 15th day of four representative months (January, April, July and October) is given in Figure 1. From now on, this consumption pattern will be referred to as ‘load type A (LTA)’.

To cover the energy demand for heating and cooling, a heat pump of 10.83 kW with seasonal coefficient of performance (SCOP) 2.4 is considered. In this case, the installed electrical power is presented in Table 1; the daily time profile of electrical energy consumption for the representative days and months is presented in Figure 2. From now on this consumption pattern will be referred to as ‘load type Β (LTB)’.

The hybrid system will consist of PVs, WT, diesel generator, batteries and equipment for electrical energy maintenance (conversion of direct current, DC, to alternating current, AC, battery charger, etc.). Batteries, diesel generator and energy maintenance equipment should be housed. Special requirements exist for the room that houses the batteries since it requires ventilation and temperature control, since high temperatures lead to reduction of batteries efficiency and of their life [23]. The considerations for equipment housing were a lodge 2 m × 4 m × 2.7 m, with a room for the diesel generator and the fuel tank, and a room for the batteries and the energy maintenance equipment constructed of polyurethane panels of 8 cm for the roof and floor, and of 5 cm for the walls. The lodge should be equipped with lighting, ventilation fan, and a heat pump with 2.5 kW cooling power and seasonal energy efficiency ratio (SEER) 2.5 for cooling the battery room. In Table 1, the installed capacity of each device is given. For the calculation of energy consumption for cooling the lodge, energy performance was simulated with EnergyPlus [24]. The daily energy consumption for the operation of the hybrid system is given in Figure 3 for the representative days from May to September.

2.2. Examined Solar Potential

The characteristics of solar potential for areas with low solar potential (for Greece) are based on data from [25] and [26] and calculated according to [27]. The solar data required for the design and optimization of a hybrid system are summarized in

Table 2. Solar potential.

Month	Monthly Total Radiation on Horizontal Surface [kWh/m2]	Monthly Diffusive Radiation on Horizontal Surface [kWh/m2]	Monthly Average Clearness Index [-]	Monthly Average Daily Temperature [C]	Cloudy Days [-]	Consecutive Number of Cloudy Days [-]
Jan	54.76	22.33	0.42	6.34	9.75	5.38
Feb	70.09	29.65	0.44	7.73	7.54	3.41
Mar	108.75	48.03	0.46	10.61	6.84	4.91
Apr	142.59	64.41	0.5	15.03	4.93	3.61
May	184.13	82.08	0.54	20.28	4.93	2.15
June	205.07	86.53	0.59	25.13	2.94	1.1
July	211.88	85.98	0.61	27.5	0.39	0.2
Aug	190.31	73.27	0.61	26.98	2.36	1.98
Sep	141.78	53.9	0.57	22.91	5.06	4.05
Oct	97.1	37.53	0.51	17.34	11.18	9.53
Nov	60.08	23.68	0.45	11.73	11.68	7.34
Dec	47.04	19.14	0.41	7.58	7.69	7.55

.

2.3. Examined Wind Potential

The determination of typical wind potential is much more difficult than the determination of the typical solar potential of a location, as it does not depend on latitude and it is not possible to calculate the daily and annual variation with analytical models. The first element in the determination of wind potential is the average wind velocity, which should be more than 5 m/s at a height of 10 m, as in the smallest wind turbines the cut-in velocity is around 3 m/s, and the rated velocity is around 11 m/s. For small WTs, the wind rose is not of interest, as they have a wind-direction monitoring system. The maximum wind velocity is also not of interest, since the destruction velocity for WTs is very high. The turbulence intensity can be considered to correspond to an open plain with roughness length z₀ = 0.03 [m]. The next required information for calculating the yearly energy production from a WT is the Weibull scale parameter, C, and the Weibull shape parameter, k. However, when it comes to an autonomous hybrid system, then the knowledge of the daily and annual variation of the wind is also important. In the European area, two basic patterns of annual [28] and daily wind variation are recognized: the Mediterranean model, where local winds dominate, affected by the land–sea temperature difference and resulting in high monthly and daily wind velocity variations; and the central and northern European model, where the main component of the winds are the geostrophic winds, characterized by a small annual and daily variability. Heraklion (Greece) is chosen for a monthly variation pattern of wind velocity for the Mediterranean area, and for the central and northern Europe, Texel (Netherlands) is chosen [28]. Despite the fact that the considered greenhouse is located in the Mediterranean, both of these basic patterns will be examined, as well as a case in which the wind potential is characterized as unexploitable. According to the above, the examined wind potentials are summarized in

Table 3. Examined wind potential characteristics.

Parameter	Mediterranean Pattern	Central/North Europe Pattern
Mean wind velocity at 10 m height, Ū [m/s]	5	5
Ground roughness length, z0 [m]	0.03	0.03
Turbulence Intensity [%]	17.2	17.2
Maximum expected wind velocity in 50 years, Umax [m/s]	23.59	23.59
Weibull scale parameter, C [m/s]	5.57	5.64
Weibull shape parameter, k [-]	1.8	2
1-h autocorrelation factor, r1 [-]	0.9	0.8
Diurnal pattern strength, δ [-]	0.3	0.04
Time of maximum wind velocity [h]	13	5

.

2.4. Hybrid System Characteristics

In order to calculate the cost of produced energy in a hybrid system, the reduction of CO₂ with respect to the zero solution (grid connected or use of diesel generator) and to optimize the design of the hybrid system according to the above criteria, the operational characteristics of the hybrid system components and the relative costs should be determined. For the determination of hybrid component characteristics, semi-empirical methods will be used [18]. Most of these methods give upper limits of component required power because they are designed for the calculation of each component in the event that it should cover the load alone. The initial design will also determine the space search. Then, for the optimization of the design and the simulation of the system performance, the HOMER software will be used [16]. All the components operate to charge the batteries with 40% initial charge level and a charging time of 2–5 h. The maximum acceptable lack of energy coverage is determined as the percentage of electricity demand for purposes other than microclimate control. In

Table 4. Hybrid system basic characteristics.

Component	Properties	Costs
Diesel generator	electrical efficiency 29% minimum load 48% life span 1500 h	Fuel price: 1.5 €/l Initial installation cost: 250 €/kW Replacement cost: 250 €/kW Maintenance cost: 0.005 €/h
Wind Turbine	Uci = 3.5 m/s, Ur = 12 m/s, Uco = 20 m/s Life span 25 years Hub height 10 m	Initial installation cost: 3000 €/kW Replacement cost: 2000 €/kW Maintenance cost: 300 €/y
Photovoltaics	Semi-transparent, Module nominal power 4 W, Efficiency 3.25% Derating factor 0.9 No temperature efficiency dependence	Initial installation cost: 1.4 €/W Replacement cost: 0.8 €/W Maintenance cost: 0.01 €/W
Batteries	C10h = 900 Ah, C120h = 1220 Ah, C240h = 1365 Ah, Vb = 4 V	Initial installation cost: 400 €/item Replacement cost: 400 €/item Maintenance cost: 350 €/year
Energy maintenance unit	Life span 20 years Efficiency 95%	Initial installation cost: 3700 € Replacement cost: 3700 € Maintenance cost: 80 €/year
System	Life span 25 years Maximum annual capacity shortage 1% for the case of heat pump, 12% for the case of fossil fuel	Nominal discount rate 7%

, the basic component characteristics and costs are presented.

The maximum annual capacity shortage refers to loads whose service can be shifted in time without affecting production. Those loads could correspond to energy consumption for cold storage, fertigation, sorting and packaging, and lighting. In the case of load type A, these consumptions correspond to 12% of the annual energy demand. In the case of load type B, they represent 1% of the annual electrical energy consumption.

3. Results Discussion

3.1. Initial Design According to Semi-Empirical Methods

There are at least three proposed methods for the sizing of the diesel generator. The first method [29] suggests that the DG should be able to simultaneously cover the load and the battery charging.

$$N_E=N_L+CR$$

(1)

where N_E is the active DG rated power, N_L the maximum electrical energy demand and CR the maximum charging rate calculated as

$$CR=\frac{C_{120}·V_b}{120 h}·n$$

(2)

where C₁₂₀ is the 120 h battery capacity, V_b the battery voltage and n the number of batteries. A second approach proposes that it is enough to cover the maximum load [30]

$$N_E=N_L$$

(3)

In both cases, the reactive power should also be taken into account in order to finally calculate the apparent power of the DG and N_E. However, as it is proven in [18], the best DG power prediction was achieved assuming that the DG will be used only for battery charging, since in this case there is no reactive power [30].

$$N_E=\frac{C_{10}·V_b}{120h}·n$$

(4)

where C₁₀ is the 10-h battery capacity. In our work, the second approach was chosen for the determination of the search-space lower limit and the third approach for the determination of the search-space upper limit.

For the initial design of the PV power, two semi-empirical methods were used: the Amber-hour (A-h) method and the method of calculating the maximum and minimum number of PV modules required. According to [18], for constant and intensive daily and seasonal load variation, the best method for the initial design is the A-h method, initially introduced by Sandia National Laboratories [31], and its modification, the W-h method introduced in [18]. According to the A-h method, initially the daily load current is determined in Ah/day for the used voltage, and the PV performance characteristics are calculated for the examined environmental air temperature. Then the number of series-connected PV panels is calculated from the inverter (or charger) input voltage; the number of parallel-connected PV panels results from the daily load. Modern inverters offer the ability of connecting different numbers of panels in different DC inputs; therefore, the A-h was modified. In this modified A-h (also referred to as watt-hour (W-h) method), the daily load is determined in Wh and the total number of required PV panels is determined according to this load. In the initial version of the A-h method, which was developed for autonomous photovoltaic systems at a time when the photovoltaic panel cost was much higher than battery cost, the battery capacity was determined as the maximum for the cases ‘load not covered by PV’ and ‘load of successively cloudy days’. In this work, since except from PVs and batteries also exist DG and WT, it was chosen to consider the load for one autonomy day.

In parallel with the A-h, an additional algorithm [32] was used to determine the upper and lower limits of the number of PV panels used.

$$z_{min}=\frac{\frac{E_0}{\left(n_{cc}·n_{inv}·n_{wire}\right)}}{\left(H_T·n_{pv}\right)·S_{panel}·n_{temp}·n_{pol}·n_{ag}}$$

(5)

$$z_{max}=\frac{z_{min}}{n^{*}}$$

(6)

where Ε₀ is the monthly load [kWh/mo], H_T the monthly available solar energy [kWh/(mo.m²)], S_panel the PV panel area [m²], n_cc the monthly average charger efficiency, n_inv the monthly average inverter efficiency, n_wire the wiring losses, n_pv the PV panel nominal efficiency, n_temp the temperature PV coefficient, n_pol the pollution factor due to deposits, n_ag the ageing factor and n* the storage system efficiency.

In the same context, a relative algorithm [33] was also used for approaching the maximum and minimum required battery capacity alongside the modified A-h method.

$$Q_{max}=h_0·\frac{E_y}{8760·n_{dc}}·\frac{1}{DO{D}_L·V_b}$$

(7)

$$Q_{min}=\left(1-DO{D}_l\right)·Q_{max}$$

(8)

where h₀ is the yearly autonomous operation hours, E_y the yearly energy demand [kWh], η_dc the storage system efficiency, DOD_L the maximum discharge depth, and V_b the battery’ operating voltage.

The proposed rules of thumb suggested for autonomous systems [29,34] restrict the installed WT nominal power between 20–30% of the nominal DG load. As far it concerns the WT power curve, it is stated that the cut-in wind speed (U_ci) should be higher than 0.6 of the yearly average wind speed (U_av); the rated wind speed (U_r) should be lower than 2U_av; and the cut-off wind speed (U_co) should be lower than 3Uav; furthermore, it should guarantee that for the 90% of the time the local wind speed is lower than U_co.

In the case of heating with fossil fuel (load type A), if the diesel generator alone was used to cover the demand, then the installed power of 5 kWe would be chosen for the lower limit; otherwise, if it was used for battery charge, the upper limit would be 24 kWe. Therefore, the search space will be between 5 and 24 kW.

According to the modified A-h method, the required power of PV would be 7.5 kW; according to the method of minimum required area, it would be 18–42 kW; while according to the method of minimum and maximum required number of modules, the power would be between 1.1 to 4.8 kW. Nevertheless, in the case of a greenhouse, there are the constraints on the permitted roof coverage so as not to affect the crop. According to [11], this is 47%, so the maximum installed power in the greenhouse roof is 3.2 kW, which can be increased by 1.2 kW if the lodge roof is also used. Thus, the search space will be between 1.1 and 4.4 kW.

According to the modified A-h method for one day of autonomy, 8 to 42 batteries would be required (actually, since we require 48 [V] at the exit of the battery string, the research space is 12-24-36-48 batteries). Finally, the optimum wind turbine rated power will be investigated between 0.5 and 6 kW.

In the case of heating with a heat pump (load type B), if the diesel generator was used alone to cover the demand, the installed power of 16 kWe would be chosen for the lower limit; otherwise, if it was used for battery charge, an upper limit would be 33 kWe. Therefore, the search space will be between 16 and 33 kW.

According to the modified A-h method, the required power of PV would be 13.3 kW; according to the method of minimum required area, it would be 32–120 kW; while according to the method of minimum and maximum required number of modules, the power would be between 1.6 to 5.8 kW. Again, the maximum permitted installed power in the greenhouse and in the lodge roof is 4.4 kW. Thus, the search space will be between 1.6 and 4.4 kW.

According to the modified A-h method for one day of autonomy, 15 to 74 batteries would be required, in practice the research space is 12-24-48-60-72 batteries. Again, the optimum wind turbine rated power will be investigated between 0.5 and 6 kW.

3.2. Optimization with HOMER

In the

Table 5. Optimization results of base cases configurations.

a/a	Case	DG [kW]	OPV [kW]	Bat [no]	WT [kW]	COE €/kWh	Ren.Fr. [%]
1	Load type A	4	4.4	12	3	0.199	37.7
2	Load type A no WT	5	4.4	12	0	0.254	14
3	Load type B heat pump	9	4.4	12	3	0.324	23.4
4	Load type B no WT	10	4.4	12	0	0.37	8.03
5	Load type A-WP2	4	4.4	12	3	0.188	39.6
7	Load type B-WP2	8	4.4	24	3	0.322	23.5

, the configurations of the examined versions of basic cases resulting from the HOMER optimization are presented along with the achieved cost of energy (COE) and the load percentage covered by the renewable components (Ren.Fr.). In all the cases, the best COE were achieved for the highest examined capacity shortage. The two basic load profiles were examined for operation in the Mediterranean wind-potential pattern (WP1), the central and northern Europe wind-potential pattern (WP2), and the case in which systems do not include a WT. Since the available roof area limits the PV to 4.4 kW, the excess required energy is covered mainly by the diesel generator. Thus, the higher energy demand of type B load (125 kWh/day average demand vs. 73.28 kWh/d energy demand of type A load) leads to bigger diesel generator systems. Furthermore, in the case of load type B, the diesel generator operation hours are increased resulting in lower penetration of renewables in the covering of energy demand. Finally, it should be noted that in the case of load type A, the permitted shortage capacity is 12%; in load type B, the permitted shortage capacity is only 1%. All these facts have as result higher cost of produced energy for the hybrid systems using heat pumps. Obviously, this could be addressed by using extra surfaces to accommodate more PV panels. The existence of wind turbines affects positively and significantly both the cost of energy and the exploitation of renewable sources. Finally, the lower daily and yearly variation of wind potential improves slightly but clearly the system performance for both load types.

Since the increase of PV installed power is limited, the influence of WTs rated power is investigated. Thus, the two basic case configurations were optimized for the other two WT systems: for a WT with rated power 1.5 kW and for a WT with rated power 6 kW. The optimization results are presented in

Table 6. Optimization results of base cases using different WTs.

a/a	Case	DG [kW]	OPV [kW]	Bat [no]	WT [kW]	COE [€/kWh]	Ren.Fr. [%]
1	Load type A	4	4.4	12	1.5	0.248	25.1
2	Load type B	9	4.4	12	1.5	0.37	8.03
3	Load type A	4	4.4	12	6	0.235	37.3
4	Load type B	9	4.4	12	6	0.331	25.7

.

From Table 6, it is clear that wind turbines smaller than 3 kW are not able to exploit the available wind potential, burdening the hybrid system performance both in terms of cost of energy and renewable penetration. However, even the largest wind turbine does not manage to improve the cost of energy production, even though in the case of load type B it increases the participation of renewables in covering the load. This increase in wind potential utilization in this case is not enough to justify the extra cost. Therefore, the initial sizing of the wind turbine for the considered loads was successful.

3.3. Parametric Study

Then the six cases were examined for different wind and solar regimes, as well as for different loads. The examined solar potential is characterized by a yearly average value of incident radiation of 4.14 kWh/m²/d. A parametric study for the same daily and yearly profile, but with yearly average values of 3, 3.5, 4.5 and 5 kWh/m², is presented. The examined wind potential corresponds to a yearly average wind velocity of 5 m/s, a parametric value for the same yearly and daily profile but with average wind velocities of 4, 6 and 7 m/s was performed. Finally, the examined load profiles corresponded to yearly average daily consumptions of 73 kWh/d for the ‘load type A’ configuration, and of 125 kWh/d for the case with heat pump. For the ‘load type A’ cases the yearly averaged daily loads of 50, 90, 110 and 130 kWh/d were examined while in the ‘load type B’ consumption case the loads of 100, 150, 175 and 200 kWh/d were investigated. For each of the new parametric studies, different optimum configurations were proposed, but in all cases PVs of 4.4 kW were required. In Figure 4, the results of parametric studies for different solar potentials are presented. For the examined cases, Figure 4a presents the variation of COE vs. the solar potential, and Figure 4b presents the variation of renewable fraction vs. solar potential. In Figure 4, as well as Figure 5 and Figure 6, the following notation is used: Case 1 (load type A) black continuous line, Case 2 (load type A without use of WT) dashed red line, Case 3 (load type B) dotted green line, Case 4 (load type B without use of WT) blue hidden line, Case 5 (load type A for the 2nd examined wind-pattern type) brown thin continuous line with circles, and Case 6 (load type B for the 2nd examined wind pattern type) orange thin line with triangles.

In all the examined cases, the cost of energy decreases as the solar potential increases, while the percentage of load covered by renewables increases at a faster rate. The best performance is achieved for the load type A profile, with the minimum energy cost corresponding to the second examined wind pattern type (when there is low yearly and daily variation of wind velocity). On the contrary, in the case of load type B, the type of wind pattern does not significantly affect the cost of the energy produced. In this case, the absence of WT restricts the percentage of demand covered by renewables to very low values. The appearance of WT in the system for the load type A obtains a COE that is competitive with electricity supplied to the consumer from conventional sources. In the load type B, surfaces beyond the greenhouse roof should be utilized in order to install more PV power so as to increase the percentage of the load covered by RES and make the hybrid system competitive with the purchase price of energy from the grid, even for high solar potentials. In Figure 5a,b, the cost of produced energy and the renewable fraction respectively are given vs. the available wind potential.

Increasing the available wind potential seems to be a more important influence on the cost of energy and the percentage of load covered by renewables than the solar potential. For important wind potential (yearly average wind speed higher than 6 m/s) in the case of load type A, the achieved cost of energy corresponds to a very profitable hybrid system. However, even areas with low wind potential (wind speed less than 4 m/s) offer an attractive cost of energy. Further improvement in the case of load type B can be achieved with the addition of more PV power. In the case of load type B, there is no significant deviation between the two types of wind regime. In the Figure 6a,b, the dependence of hybrid system performance on the yearly average daily load for the two examined load profiles (with and without heat pump) is investigated in terms of cost of energy and renewable fraction.

Increasing the energy demand in both load profiles increases the cost of energy and decreases the load percentage covered by renewables, but at a decreasing rate. The energy profile demand corresponding to operation with a heat pump does not significantly differentiate the performance of the hybrid system, since for the same daily demand the two types of consumption result in similar energy-cost values and almost the same percentages of load coverage by renewables. This is because in the Mediterranean, the heat pump is equally used for heating and cooling. The effect of not using wind turbine becomes less significant as demand increases, since the selected wind turbine is sized for lower demand and appears to exhaust its contribution margin after 125 kWh/d.

3.3.1. Optimization of Wind Turbines Power

The initial sizing of the wind turbines is optimized by simulating the system performance with HOMER software for two wind turbines with lower and higher rated power and analogous power curves. The following wind turbines were examined: (a) wind turbine of 1.5 kW rated power with U_ci = 3.5 m/s, U_r = 14 m/s and U_co = 20 m/s, and (b) wind turbine with rated power 6 kW with U_ci = 3.5 m/s, U_r = 14 m/s and U_co = 20 m/s. The above-described parametric studies for different wind and solar potentials, as well as different annual and daily average loads, were repeated for the two new wind turbines. In Figure 7, the wind potential parametric study results are presented with respect to the achieved cost of energy and load fraction covered by renewable systems for the two examined load types and for the default wind-potential pattern. The annotations that will be used for the next figures is summarized as follows: (a) Wind turbine of rated power 1.5 kW for load type A (WT-1.5-LTA) with black continuous line, (b) Wind turbine of rated power 3 kW for load type A (WT-3-LTA) with red dashed line, (c) Wind turbine of rated power 6 kW for load type A (WT-6-LTA) with green dashed dotted line, (d) Wind turbine of rated power 1.5 kW for load type B (WT-1.5-LTB) with dotted blue line, (e) Wind turbine of rated power of 3 kW for load type B (WT-3-LTB) with continues purple line with purple squares, and (f) Wind turbine of 6 kW rated power for load type B (WT-6-LTB) with continuous brown line with brown triangles.

In the case of load type, A, for low wind velocities, the bigger wind turbine cannot offer any energy profit that could justify its cost. In fact, for wind velocities lower than 5 m/s, even the small WT of 1.5 kW provides a lower cost of energy. In general, it turns out that the initial sizing of 3 kW was the best choice for the wind turbine for all the examined wind potentials. For the load type B, the small wind turbine proves to be completely insufficient to increase its contribution to the system, even at higher wind velocities. Nevertheless, for wind velocities higher than 5.5 m/s, the bigger WT of 6 kW proves to be more adequate for the designed system, since it offers lower cost of energy.

While the considered 6 kW WT for load type A cannot reduce the cost of the produced energy, it can reduce the production of pollutants, since for wind speeds above 5.5 m/s, it achieves a percentage coverage from renewables greater than that achieved by the 3 kW WT. For the load type B, the ‘greener’ contribution of the 6 kW WT is expanded in the whole examined range of wind potential.

Figure 8a,b presents the cost of energy and the load percentage covered by renewables versus the wind velocity for the wind pattern type 2.

For the small wind turbine and for low wind velocities, the achieved cost of energy is lower for the default wind pattern. In all the other cases, the wind pattern 2 (with lower yearly and daily variation) gives better performance in terms of produced energy cost and renewable energy contributions. In fact, even the performance of the 1.5 kW WT seems to be favored for the load type 2 when wind velocity increases. The qualitative conclusions are the same with the conclusions derived for the default wind potential concerning the chosen wind turbine power. However, for the load type A, the superiority of the bigger 6 kW WT begins at lower wind velocities, while for the load type B, it is delayed until higher velocities.

In Figure 9, the solar parametric study is presented for the three examined wind turbines and the two types of loads for the default wind pattern. Results show that for the examined wind potential, the initial choice of a wind turbine with rated power 3 kW can give a lower cost of energy, no matter the variation of solar potential for both load types. This confirms the correct selection of the initial sizing of the wind turbine. As expected, the increase of available solar radiation has a higher impact in the load type A for the small 1.5 kW wind turbine; in this case, the margin of improvement of the cost of energy is important, since the wind turbine has limited potential to contribute to covering the load. The same conclusions emerge from studying the percentage of the load covered by renewables for load type A. For the load type B, the higher renewable penetration and, consequently, the higher reduction of CO₂ emissions, is achieved with the wind turbine of 6 kW.

Figure 10 presents the solar parametric study for the three different wind turbines for wind pattern 2; it shows that the basic conclusions obtained from the study of the default wind pattern remain the same.

Figure 11 presents the load parametric study for the different wind turbines for the default wind pattern for both load types. For the load type A, it is obvious that the wind turbine is well sized for the whole examined load-demand range, with smaller wind turbines leading to a much higher cost of energy, while bigger wind turbines used is not justified with regard to produced energy cost. For the load type B both smaller and bigger wind turbines present minimum in the initially concerned load due to the whole system sizing for this specific load. Nevertheless, the small wind turbine is not able to provide adequate energy while the big wind turbine seems to prevail, in terms of energy cost, only around this load demand. Analogous are the conclusions drawn from the renewable fraction evolution. For the load type A the small wind turbine is not big enough while the big wind turbine cannot offer higher renewable penetration. For the load type B the bigger wind turbine provides a higher renewable fraction contribution to the load leading to lower CO₂ emissions, especially for loads around the initially concerned demand.

In Figure 12, the same parametric study is presented for the wind pattern 2. The conclusions about hybrid systems performance with respect to the load demand variation is the same as for the default wind pattern, although slightly improved results are achieved.

3.3.2. Effect of Roof Geometry in the PV Potential

In all the above calculations, the photovoltaics were considered to operate at an inclination angle of 0° in order for the results to have a more general validity regardless of the type of ceiling on which they will be placed. With this inclination angle, the yearly solar produced energy in the specific site was 5676 kWh. Since the examined greenhouse has an arched roof, the PVs will in fact be installed in arrays with different inclinations. Thus, the effect of different inclination angles of each PV array was also examined. Two orientations of greenhouse were examined: East–West and North–South. In

Table 7. PV’s array inclination and azimuth angles.

Orientation East-West
Array	Inclination Angle [deg]	Azimuth [deg]	Installed Power [kW]	Produced Energy [kWh]
A.1	48.53	0	0.517	715
A.2	26.61	0	0.517	739
A.3	4.64	0	0.451	600
A.4	4.64	180	0.451	562
A.5	26.61	180	0.517	502
A.6	48.53	80	0.517	362
A.7	30	0	1.42	2034
Summary East–West [kWh]				5514
Orientation North-South
B.1	48.53	−90	0.517	588
B.2	26.61	−90	0.517	645
B.3	4.64	−90	0.451	583
B.4	4.64	90	0.451	580
B.5	26.61	90	0.517	627
B.6	48.53	90	0.517	564
B.7	30	0	1.42	2034
Summary North–South [kWh]				5621

, the considered arrays with the inclination angles and solar azimuths, along with the yearly produced energy, are presented. It should be noted that an array of 1.42 kW would be installed on the roof of the hybrid system lodge in order to avoid overshadowing and where the optimum inclination angle can be used.

From the calculations it is shown that the North–South orientation gives 1.9% better production in comparison with the East–West orientation. Nevertheless, the North–South orientation gives almost the same production as the zero-inclination angle considered in the above parametric studies, so the conclusions drawn are not affected. The result is merely attributed to the high performance of the 7th array installed on the hybrid lodge roof in the optimum orientation and inclination. Comparing the performance of only the PVs installed on the greenhouse arched roof, the zero inclination would give 6.6% better results in comparison with the North–South orientation and 9.4% better than the East–West orientation. Finally, if the performance of North–South orientation is compared with the performance of an array sited in land with inclination angle 30° and south orientation, then it will appear that its performance is reduced by 42%. However, since the installation on the roof does not take up additional space and also does not require the use of bases, it can be considered that the installation of PV on roofing is economically competitive.

3.4. Total Cost of Energy and Environmental Footprint for the Two Types of Loads

The whole analysis up to now concerns the dependence of hybrid system performance for different types of greenhouse loads and not actually comparing two different technologies of greenhouses in terms of energy cost and environmental footprint. In the load type B, all the energy demand (for all the greenhouse energy needs) concerns electrical energy, which is provided either by conventional diesel generators or by renewables. In the load type A, the energy needs for heating are not included. Considering the same energy demand for heating of 226 kWh/m² [4], combustion efficiency of a diesel boiler of 95%, distribution efficiency 92%, and emission efficiency of 93%, the yearly energy consumption for heating a 200 m² greenhouse is 55,600 kWh. For the calculation of the additional cost of this energy, the net calorific value for diesel was considered to be 11.92 kWh/kg, density of 850 kg/m³ and cost of fuel of 1.5 €/l (as in Table 4). Thus, the total cost of energy for the two greenhouse technologies is configured as shown in

Table 8. Comparison between two greenhouse technologies.

Case	COE [€/kWh]	Energy Cost per Greenhouse Area [€/m2]	Primary Energy Consumption per Year [kWh/y/m2]
Load Type A	0.167	70.37	391.46
Load Type A–no Wind Turbine	0.187	80.35	417.66
Load Type B	0.324	76.88	189.84
Load Type B–no wind Turbine	0.370	93.80	229.58
Load Type A–WP2	0.163	68.00	387.22
Load Type B–WP2	0.322	76.65	191.17

. Nevertheless, reading the cost of kilowatt hour alone can be misleading in estimating the energy cost of a greenhouse. This is because greenhouses using a conventional boiler consume much more energy, but at a lower price per kilowatt hour; in Table 8, this other information is included: the cost of energy per greenhouse area and the yearly primary energy consumption. Concerning the cost of produced kWh, the greenhouses with technology type A achieves a lower cost of kWh, since this energy is used for heating the cheaper conventional fuel source. However, this led to consumption of higher amounts of energy. Thus, the final cost to cover the energy demand in the two technology types are comparable, but when the primary energy consumption is considered, the superiority of technology type B is obvious.

Since primary energy consumption is also related with the environmental performance of the greenhouse, this aspect is further investigated in

Table 9. Environmental footprint of the two greenhouse technologies.

Case	CO2 per Energy Production [kg/kWh]	CO2 per Greenhouse Area [kg/m2]
Load Type A	0.22	93.95
Load Type A—no Wind Turbine	0.23	100.24
Load Type B	0.19	45.56
Load Type B—no wind Turbine	0.22	55.10
Load Type A—WP2	0.22	92.93
Load Type B—WP2	0.19	45.88

, where the CO₂ emissions from the two technology types are presented. In the first column, the CO₂ emissions per total energy production is calculated where the technology using heat pumps achieve slightly lower CO₂ emissions. However, calculating the emitted CO₂ per greenhouse area, it turns out that greenhouses type A burdens the atmosphere with twice as many pollutants.

4. Conclusions

The knowledge not only of the annual consumption but also the daily and annual distribution of energy demand in a greenhouse is essential information for the correct design of a hybrid system. In greenhouses in which electricity is not used for heating, the use of photovoltaics is preferred, since the maximum demand for electricity coincides with the period of maximum available solar radiation. In all cases considered, hybrid systems designed based on the maximum power of semi-transparent photovoltaics, that can be installed on roofing without affecting the growth of the crop, can contribute significantly to the reduction of CO₂ emissions as the percentage of demand covered by renewable sources varies from 8–40%. From the parametric studies it is shown that the initial sizing was quite successful.

In the considered cases, due to the limited possibility of installing PV on the roof, the available solar potential does not dramatically affect energy costs (reduction of the order of €0.025 per kWh/m²/d of solar radiation), but offers a higher penetration of renewable energy sources in the energy mix.

Even with low wind potential, the wind turbine in a hybrid system can significantly help the economic viability of a hybrid system in both wind patterns typical of Europe (reduction by 34% of produced energy cost). Nevertheless, when the daily energy demand increases beyond the ranges of wind turbine initial sizing, then the contribution of WT is restricted. In these cases, and for higher available wind potential, it is worth investigating the use of bigger wind turbines. Finally, lower yearly and daily wind variations provide slightly but clearly better hybrid system performance (decrease of energy cost by an order of 7%).

The use of an arched greenhouse roof for the installation of photovoltaics proves to be a competitive solution compared with installation in a field, where the choice of appropriate inclination and orientation is possible. The North–South orientation of the greenhouse proved to be a slightly more efficient choice for the installation of PVs on greenhouse roofing.

Properly dimensioned hybrid systems can satisfy greenhouses that do not consume electricity for heating, using semi-transparent photovoltaics on their roofing without affecting the crop and offering competitive electricity prices to farmers. For a diesel price of 1.5 €/l, the different hybrid system configurations resulted in a cost of energy from 0.188 to 0.254 €/kWh for the default renewable potential, and can be reduced to 0.17 €/kWh for favored solar potential, and up to 0.13 €/kWh for favored wind potential. When the greenhouse also uses electricity for heating, then the photovoltaic field must be strengthened with a power greater than that which can be installed on the roofing, utilizing other available surfaces of the unit (e.g., storage areas or product processing areas). Therefore, even in cases where a heat pump is used for heating and cooling, it is worth investigating the use of a hybrid system, the economic viability of which depends to a large extent on the price of grid electricity as well as on the prices of conventional fossil fuels. For the same price of diesel, the energy production cost for the default renewable potential varies from 0.322 to 0.37 €/kWh, but these values can be reduced to 0.31 €/kWh for favored solar potential, and to 0.25 €/kWh for favored wind potential.

In all the cases examined, the design was done in such a way as to ensure a predetermined level of energy demand satisfaction. However, in the optimization process, there was a large benefit (in terms of cost of energy produced) when this level was not 100%. The higher the allowable failure margin, the lower the cost of the energy produced. Thus, the possibility of determining this safe limit of failure (safe in the sense that it does not endanger the crop) emerges as an important factor both in the design and in the final energy cost. As a rule of thumb, it can be said that this margin corresponds to actions that can be shifted in time (e.g., the packaging).

Finally, with respect to overall energy performance, the two greenhouse technologies examined result in comparable energy costs when their electrical energy demand is served by an appropriate hybrid system. However, the greenhouse which uses a heat pump for heating results in a much lower environmental footprint.