- Email: [email protected]
battery pumping system in agricultural experiment station
Available online at www.sciencedirect.com
ScienceDirect Solar Energy 112 (2015) 319–338 www.elsevier.com/locate/solener
Power management of a photovoltaic/battery pumping system in agricultural experiment station Abla Khiareddine a,⇑, Chokri Ben Salah b,1, Mohamed Faouzi Mimouni a,2 a
Research Unit on Study of Industrial Systems and Renewable Energy (ESIER), Department of Electrical Engineering, National School of Engineers of Monastir, Avenue Ibn EL JAZZAR, 5019 Monastir, Tunisia b Control and Energy Management Lab. (CEMLab), Department of Electrical Engineering, National School of Engineers of Sfax, BP. W, 3038 Sfax, Tunisia Received 31 July 2014; received in revised form 14 November 2014; accepted 19 November 2014
Communicated by: Associate Editor Mukund Patel
Abstract This paper focuses on dynamic modeling, simulation, control and energy management in an agricultural experiment station located at Sahline–Tunisia consisting of a 1.5 kW photovoltaic panel (PV) and a 25 A h Lead Acid battery storage supplying an induction motor coupled to a centrifugal pump as mechanical load. The cost-optimally design and the new suitable power management approach are the two main objectives. An iterative optimization approach namely, the Deficiency of Power Supply Probability (DPSP), the Relative Excess Power Generated (REPG), the Energy Cost (EC) as well as the Total Net Present Cost (TNPC) have been developed in order to find the optimal configuration of PV/battery. To reach the second object, three new supervisory controllers are designed, a neurofuzzy controller, a fuzzy controller and an algorithm controller. In order to show the effectiveness of the first one, a comparison of the three controllers is proposed. The principal objectives of the three supervisory controllers are: (i) the design of an adequate tracking system maximum power point (MPPT) to extract the maximum power which is given by the theory of conservation of energy, (ii) the insurance of the control speed needed for the vectorial control of the induction motor, (iii) the regulation of the water in the tank which is taken as a second storage system and finally (iv) the insurance of the correct operation for all the conversion string in order to optimize the quantity of water pumped. Ó 2014 Elsevier Ltd. All rights reserved.
Keywords: Water level regulation; Supervisory controller; Neuro-fuzzy; Optimization
1. Introduction Energy plays the most vital role in the economic growth, progress, and development, as well as poverty eradication and security of any nation. Energy is an important factor ⇑ Corresponding author. Tel.: +216 95644567.
E-mail addresses: [email protected] (A. Khiareddine), [email protected] (C. Ben Salah), [email protected] (M.F. Mimouni). 1 Tel.: +216 98676138. 2 Tel.: +216 52852959. http://dx.doi.org/10.1016/j.solener.2014.11.020 0038-092X/Ó 2014 Elsevier Ltd. All rights reserved.
in all the sectors of any country’s economy. In the recent years, the diminishing supplies of fossil fuels and their impact on the environment have encouraged a growth in sustainable energies such as wind and solar power; such a significant growth was seen by the photovoltaic industry. Solar power generation is experiencing a remarkable growth in terms of installed power and energy generation in many countries. Autonomous photovoltaic panels are intermittent sustainable energy sources which require energy storage to balance generation and demand, as photovoltaic generation depends on time and weather.
320
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
Traditionally batteries are the most common storage technology for photovoltaic systems (Gibson and Kelly, 2010; Kaldellis et al., 2012; Steffen and Weber, 2013). Standalone photovoltaic systems are often used in remote areas, away from the national grid for water irrigation system (Bouzidi, 2011; Correˆa et al., 2012; Glasnovic and Margeta, 2007; Hamidat et al., 2003; Hegazy and Abou Hashema, 2013; Meah et al., 2008). Many works dealt with the choice of the drive system to interact with the PV source; PV pumping systems based on DC machines, AC machines (Correˆa et al., 2012; Meah et al., 2008), the type of pumps to use and the ways to
control and optimize the whole system. In fact, several control schemes are modeled and described as scalar control, vector control of the induction machine (Khiareddine et al., 2013; Mimouni et al., 2004) and DTC control (Rekioua and Matagne, 2012). Various algorithms have been developed over the years (Ben Salah and Ouali, 2011; Mazouz and Midoun, 2011; Shaiek et al., 2013) in order to determine the maximum power point such as Perturb and Observe Technique, Incremental Conductance Technique, Hill Climbing Control, Curve Fitting Method, Sliding Mode Controller, Fuzzy Logic Technique, Artificial Neural Networks,
Fig. 1. Schematic diagram of photovoltaic pumping system with battery storage.
Fig. 2. Pumping system: (a) the motopump, (b) general view of the tank and (c) greenhouse.
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
Neuro-Fuzzy Method, Genetic Algorithms. . .; a comparison of these different Algorithms has been studied (Rekioua and Matagne, 2012). These control techniques act on the duty cycle to place the generator in its optimal value regardless of changes in meteorological conditions. A combination of Lead Acid batteries, ultra capacitors, wind turbine and photovoltaic panel in a Hybrid Energy Storage System is examined (Brian and Byron, 2012; Diaf et al., 2007; Sarrias et al., 2012). Indeed, using a storage system as Lead Acid batteries, ultracapacitors increase the power density of the overall system. The addition of storage system to the solar and wind energy resources necessitates a power management strategy to provide good organization and distribution of power between the different energy sources as well as to protect the different elements of the conversion string from over/ under design of the system, on the other hand (Ben Salah and Ouali, 2010; Ben Salah et al., 2008). Various optimization techniques of hybrid PV/wind/battery/ultracapacitors systems sizing have been reported in the literature (Akbar and Alireza, 2014; Belmili et al., 2014; Bond et al., 2012; Glavin and Hurley, 2012; Koutroulis
321
et al., 2006). Using the autonomy level of the system, a set of different configurations which meet the load is obtained. The authors developed general methodology by considering design factors such as autonomy for sizing and optimization (Hamidat and Benyoucef, 2009; Hengsritawat et al., 2012; Kaldellis et al., 2010; Odeh et al., 2006; Papaefthymiou and Papathanassiou, 2014; Roy et al., 2010). Indeed, several approaches are developed to achieve the optimal configurations of hybrid systems PV/wind/battery/ ultracapacitors/diesel in terms of technical analysis. Namely, we find the technical approach called loss of power supply probability (LPSP) applied by (Diaf et al., 2007), the iterative optimization technique following the Deficiency of Power Supply Probability (DPSP), the Relative Excess Power Generated (REPG) developed by (Kaabeche et al., 2011), the iterative optimization approach following the Total Energy Deficit (TED) used by (Kaabeche and Ibtiouen, 2014) and the artificial intelligence by (Akbar and Alireza, 2014). For economic analysis, several economic criteria are exploited, such as the Total Net Present Cost (TNPC), the Total Annualized Cost (TAC), Break-Even Distance
8000
Daily load consumption (Wh/d)
7500
Load consumption
7000 6500 6000 5500 5000 4500 4000 3500 3000 January
February
March
April
May
Juin
July
August September October November December
Months
Fig. 3. Variation of the seasonal electricity consumption.
8000 Solar radiation
Daily solar radiation (Wh/m3/d)
7000 6000 5000 4000 3000 2000 1000 0
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Months
Fig. 4. Evolution of the solar illumination for 12 months of the year.
322
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
Fig. 5. Flow chart of the optimal sizing model.
Fig. 6. Relative Excess Power Generated REPG for different battery capacity and for 3 days of autonomy.
Analysis (BEDA), Energy Cost (EC) (Kaabeche et al., 2011; Kaabeche and Ibtiouen, 2014) and the levelized cost of energy (LCE) developed by Diaf et al. (2007). This paper deals with the modeling and control of a hybrid system integrating a photovoltaic panel and
batteries as an energy storage system. In the first part, a description of the whole system and its sizing systems is based on the Deficiency of Power Supply Probability (DPSP), the Relative Excess Power Generated (REPG) for power reliability, the Total Net Present Cost (TNPC)
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
323
Fig. 7. Deficiency of Power Supply Probability DPSP for different battery capacity and for 3 days of autonomy.
Deficiency of Power Supply Probability (%)
5 4.5 Battery capacity
4 20 Ah 25 Ah 50 Ah
3.5 3 2.5 2 1.5 1 0.5 0 1200
1250
1300
1350
1400
1450
1500
1550
1600
1650
PV power (W)
Fig. 8. Deficiency of Power Supply Probability DPSP for different battery capacity and for 1 day of autonomy.
Total Net Present Cost ($)
x 10
4
Optimal configurations
10 8 DPSP = 0%
6 4 2 X: 50 Y: 1500 Z: 1.995e+004
0 1700 1600
250 200
1500
150
1400
PV power (W)
100 1300
50
Battery capacity (Ah)
Fig. 9. System configurations and Total Net Present Cost for the studied system.
and Energy Cost (EC) for system costs. In the second part, the modeling of the three strings is presented. In the third part, specific focus is to be taken on the new supervisory controller in order to coordinate the operation of energy sources fulfilling several functions: firstly
to regulate the water level in the tank, taking into consideration the solar illumination, the battery’s state of charge (SOC) and the water level in the tank. Secondly to determine the reference speed needed for the vector control of induction motor and finally to ensure a good power
324
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
Optimal configurations
Energy Cost ($/kWh)
3 2.5 DPSP = 0%
2 1.5 1 0.5 1550 1500
250 200
1450
150
1400
PV power (W)
100 1350
50
Battery capacity (Ah)
Fig. 10. System configurations and Energy Cost for the studied system.
Table 1 PV modules specifications. Type
EG (eV)
A
Pmax
IMPP (A)
VMPP (V)
ISC (A)
VOC (V)
CT
NOCT
NS
Unit price (US$/W)
Lifetime (year)
Maintenance cost in the first year (%)
SM50-H
1.12
1.5
50Wp
3.15
15.9
3.35
19.8
1.2
45 ± 2 °C
33
3.6
25
1% of price
Table 2 Battery specifications. Nominal capacity (A h) 50
Voltage (V) 240 (12 * 20)
DOD (%)
gB (%)
Unit price (US$/W)
Lifetime (year)
Maintenance cost in the first year (%)
60
85
0.213
5
3% of price
Table 3 DC/AC inverter specification. Power rating (W)
Efficiency (%)
Unit price (US$/W)
Lifetime (year)
Maintenance cost in the first year (%)
1500
90
0.711
10
0
Table 4 Induction motor parameters (1.5 KW). N (tr/min)
p
Rs (X)
Rr (X)
Ls (H)
Lr (H)
M (H)
J (Kg m2)
f (Nm/rds1)
1435
2
5.72
4.2
0.462
0.452
0.44
0.0049
1.5.10–4
management. Photovoltaic pumping in a closed loop will allow us to win a lot of energy which can be used for the battery. Lastly, the simulation results and conclusions are presented. 2. System configuration and unit-sizing An overall power management strategy, provided by the opening and closing of three switches C1, C2, C3, has been designed for various purposes: Use all the energy supplied by the photovoltaic panel. Ensure proper operation for the entire hybrid system. Regulate the water level in the tank.
Ensure a new maximum power point tracking (MPPT) by acting on the reference speed required for the vectorial control of the induction motor. Fig. 1 shows the system configuration for the proposed hybrid pumping system. In the hybrid system, the renewable PV power is taken as primary source, while the battery is used as a backup and storage system. The priority is given to the PV, not only to exploit the entire renewable energy but also to increase the life cycle of the battery. Indeed, the battery is charged in the following situations: The tank is full and the power panel is available.
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
When there is deficit in power generation, the battery will produce energy to the centrifugal pump. An agricultural experiment station located at Sahline region of Tunisia chose to automate the existing pumping system, as shown in Fig. 2, by using a renewable energy and especially the solar one. A photovoltaic pumping system will be used together with a storage component as shown in Fig. 1. This pumping system possesses a 1.5 kW motor and a 3 m tank in order to irrigate many crops cultivated under
325
greenhouses and in open fields such as tomato, pepper, potato, eggplant and melon (Fig. 2). To make the power available during night and cloudy days the battery banks are essential in solar system. Lead Acid batteries are used because of their high performance and long life. Before starting the calculations, we need to confirm the following data: The electricity usage per day Ej (W h). Number of day of autonomy JAUT.
Fig. 11. PV-Pump stand-alone system.
Fig. 12. Synoptic of the vectorial command of an asynchronous motor.
326
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
I
Depth of discharge limits DODMAX always 60%. The energy efficiency of the battery gBAT, (gBAT = 85%, (Rekioua and Matagne, 2012).
C¼
Ej J AUT V BAT DODMAX gBAT
Re
Rs
Vo
Battery of 240 V (VBAT = 12 20) having DODMAX of 60% is considered here. This leads to the calculation of the capacity of the battery C (A h) using the equation below:
Ib
Is
Rt
Csurface
Vcs Cbulk
Vcb
Fig. 15. Battery model diagram.
ð1Þ
During the summer period and depending on the nature of the cultivated crops, the electricity consumed by the agricultural experiment station is twice more than that used for the rest of the year, (Fig. 3). The agricultural experiment station needs a maximum of 5 h/day for filling the tank. That is why the priority is given to solar panels and the battery is used in order to store energy during high electricity production periods and then deliver it to the user at nights or at low sunlight periods. The number of day of autonomy in our case is between one and three days. The variation of solar radiation during the 12 months of the year in the Tunisian Sahel is shown in Fig. 4. To find an optimal combination of PV-Battery to satisfy the load demand, assessment may be performed based on reliability and economy of power supply. 2.1. Reliability criteria In this study, the technical sizing model is developed using the concept of DPSP and REPG to evaluate the reliability of PV-Battery systems.
The flow chart of the optimal sizing model is shown in Fig. 5, (Kaabeche et al., 2011; Diaf et al., 2007). This program has as inputs for each hour Dt: the PV power DPPV, the load consumption DEL, the maximum and minimum energy storage capacity of the battery respectively EBatmax, EBatmin and the maximum number of autonomy day NSDmax. where EBatmax ¼ C B V BAT
ð2Þ
EBatmin ¼ C B V BAT ð1 DODmax Þ
ð3Þ
Every hour, the storage capacity is subject to the following constraints: EBatmin 6 EBat ðtÞ 6 EBatmax
ð4Þ
In this program, the aim is to determine both the yearly DPSP and the yearly REPG, knowing the power of the PV and the load demand throughout the year (8760 h). The power of the panel is fixed since the beginning. The charge and discharge of the battery are described by the following equations. Battery charge:
EL ðtÞ EBat ðtÞ ¼ EBat ðt 1Þ ð1 rÞ þ EPV ðtÞ gB : ginv Battery discharge: EBat ðtÞ ¼ EBat ðt 1Þ ð1 rÞ
EL ðtÞ EPV ðtÞ gB : ginv
ð5Þ
ð6Þ
where r is the hourly self-discharge rate, ginv and gB present respectively the inverter efficiency and the battery bank efficiency. Fig. 13. PV-Battery stand-alone system.
The Excess Power Generated (EPG) is given by:
Fig. 14. PV-Battery stand-alone system and the buck converter.
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
EPGðtÞ ¼ EPV ðtÞ
EL ðtÞ þ ðEBatmax EBat ðt 1ÞÞ=gB ginv
ð7Þ
The Relative Excess Power Generated (REPG), defined as the ratio of power excess to the sum of load demand, is calculated by the following equation: , T T X X REPG ¼ EPGðtÞ EL ðtÞ ð8Þ t¼1
t¼1
The deficiency power supply (DPS) at hour t, is given by: DPSðtÞ ¼ EL ðtÞ ½EPV ðtÞ þ EBat ðt 1Þ EBatmin ginv
ð9Þ
And the Deficiency of Power Supply Probability (DPSP) can be expressed as: , T T X X DPSP ¼ DPSðtÞ EL ðtÞ ð10Þ t¼1
t¼1
327
The Relative Excess Power Generated REPG and the Deficiency of Power Supply Probability DPSP for different battery capacity and different power panel with NSdmax = 3 days are given respectively in Figs. 6 and 7. We can notice that the DPSP has a hyperbolic shape and it is clear that the DPSP increase with the decreases of the REPG. To obtain a DPSP of 0%, corresponding to the total autonomy of PV/Battery system, two solutions are possible, either to take a battery capacity equal to or higher than 50 A h, with a panel of 1500 W or more, or take a lower PV 1500 W with a strictly upper to 50 A h capacity. We can see that there is a compromise between increasing the PV power and increasing the battery capacity to achieve the reliability requested. The Deficiency of Power Supply Probability DPSP for different battery capacity and different power panel with NSdmax = 1 days is given by Fig. 8.
Fig. 16. Battery–Pump stand-alone system.
Fig. 17. Boost converter and DC bus voltage regulation.
328
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
As shown in Fig. 8, for 1 day of autonomy and with a capacity of less than 25 A h, we cannot reach the load autonomy (DPSP of 0%) whatever the PV power is. Although from 25 A h and more, it can be achieved. To choose the right pair, the economic analysis should be made to obtain the low cost with the total autonomy of the system. The economic criteria will be detailed in the next section.
where Crec is the recurring cost and Cnon-rec is the nonrecurring cost. d (8%) and e (4%) are respectively the interest and escalation rates. n is the system life period in years (25 years). P is the number of years between two successive payments for non-recurring costs. dadj, which is the adjusted interest rate, is given as follows:
2.2. Economic criteria
The second and the most excellent economic profitability indicators of system cost analysis is the Energy Cost (EC). EC can be defined as:
The economical approach based on the Total Net Present Cost (TNPC) and Energy Cost (EC), developed by Kaabeche and Ibtiouen (2014), is considered here for system configurations. The Total Net Present Cost (TNPC), which is a key financial indicator for economic viability appreciation of an investment project, can be expressed as follows: TNPCð$Þ ¼ IC þ PWCnon-rec þ PWCrec
ð11Þ
where IC is the initial cost of the system components, PWCnon-rec and PWCrec are factors for the conversion of the non-recurring and recurring costs to their present worth. These factors are given by Eqs. (12) and (13). h ih in 1þe 1þe 1 1þd 1þd h i PWCrec ¼ C rec ð12Þ 1þe 1 1þd h i h in 1þe 1þe 1 1þd adj 1þd adj h i ð13Þ PWCnon-rec ¼ C non-rec 1þe 1 1þd adj
d adj ¼
ð1 þ dÞP ð1 þ dÞ
P 1
1
ð14Þ
TNPC CRF ECð$=KW hÞ ¼ P8760 t¼1 E GenðtÞ
ð15Þ
where CRF is the capital recovery factor which is given by: CRFðd; nÞ ¼
dð1 þ dÞ
n
ð1 þ dÞn 1
ð16Þ
Figs. 9 and 10 give the TNPC and the EC respectively, with different battery capacity and different PV power for the optimal configurations. A careful observation into Figs. 9 and 10 show that the optimal configuration, which insures the system reliability requirement with the lowest values of TNPC and EC for three days of autonomy, is the battery capacity of 50 A h and 1500 W PV power. This couple has a REPG of 45.2%, a TNPC of 19,946 (US$) and EC of 0.61 (US$/ kW h). We can find another couple that provides 0% but the cost will be more significant.
Fig. 18. Supervisor controller.
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
For one day of autonomy, the optimal configuration is the battery capacity of 25 A h and 1500 W PV power. Tables 1–4 depict the technical and economical parameters of the system components. Once the optimal configuration is chosen, namely battery of 50 A h with a PV of 1500 W for three days of autonomy and a battery of 25 A h with a PV of 1500 W for one day of autonomy. We will search the most appropriate strategy to extract the maximum power from the PV and to ensure the smooth operation of the total system. The energy management strategy is described below using as simulation parameters a PV of 1500 W with a battery of 25 A h for one day of autonomy. 3. Models of system components Different energy sources are connected to the dc bus through appropriate power electronic interfacing circuits. The three strings will be detailed in the following parts.
329
where Kch is the proportionality constant of the pump, PPVMPPT the maximum power supplied by the PVP, gMOT is the induction motor efficiency and gINV is inverter efficiency which varies from 90% to 97%; we consider here an inverter efficiency of 90% (Ernst et al., 2011). To find the reference speed that will allow us first of all to get the maximum power and secondly to regulate the water level, we used the neuro fuzzy controller which as input. 3.2. String II «PV-BATTERY» To improve the performance of a photovoltaic system, a power converter was inserted between the PV generator and the battery (Fig. 13). The DC–DC converter “buck”, controlled by a tracking algorithm of maximum power point tracker MPPT, provides the proper voltage to the load.
3.1. String I «PV-MOTOPUMP» In the first part of this paper, we presented a description of the structure which consists of a photovoltaic generator coupled through an inverter to an induction motor (Fig. 11). Indeed, the PV is directly connected to a threephase inverter which is controlled by the SVPWM. Given the performance of the vector control, this last has been developed to control the induction motor. The synoptic of the vectorial command of an asynchronous motor is given in Fig. 12. In the majority of research works, MPPT is provided by the variation of the duty cycle a in order to control DC/DC converter. In this work, the MPPT is given by the theory of conservation of energy, in other words we must determine the reference speed Xref needed for the vectorial control of the induction motor which corresponds to the maximum reached by the solar panel power. During the programming of the supervisor with either fuzzy logic or neuro-fuzzy, we need a relationship between the maximum power supplied by the PV and the reference speed Xref. Mechanical power of the induction motor PM is given as: P M ¼ X Cr
ð17Þ
And the resistant torque of the load Cr is expressed by the Eq. (18): C r ¼ K ch :X2
ð18Þ 3
P M ¼ K ch X
ð19Þ
P M ¼ gINV gMOT P PVMPPT
ð20Þ
Then the speed is given by Eq. (21): rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3 gINV gMOT P PVMPPT X¼ K ch
ð21Þ
Fig. 19. Membership functions of input variables of the fuzzy logic.
330
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
Fig. 20. Membership functions of output variables of the fuzzy logic.
Fig. 21. Principle of neuro fuzzy system.
The duty cycle a is obtained crossing the supervisory controller using as inputs the solar irradiation G and the temperature T. The behavior of the buck converter shown in Fig. 14 can be captured by the following equations: Transistor is on
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
331
Fig. 22. Neuronal structure of the proposed model in Matlab.
dV PV ¼ iPV iBATT dt diBATT L ¼ V PV V BAT dt
ð22Þ
C Training data : FIS output 200 150
Transistor is off
100 50
0
500
1000
1500
2000
2500
3000
ð24Þ ð25Þ
Over one switching period Ts, the equations can be combined and represented as follows:
Tesng data : FIS output 200 150
dV PV ¼ ðiPV iBATT Þ a þ iPV ð1 aÞ dt diBATT L ¼ ðV PV V BAT Þ a þ ðV BAT Þ ð1 aÞ dt
C
100 50 0 -50
dV PV ¼ iPV dt diBATT L ¼ V BAT dt
C
0 -50
ð23Þ
ð26Þ ð27Þ
The finally equations of the buck converter are: 0
500
1000
1500
2000
2500
Checking data : FIS output 1.6 1.55 1.5 1.45
50
1.4
0
1.35
0
500
1000
1500
2000
Fig. 23. Learning process of the reference speed.
2500
0
5
10
15
20
25
Fig. 24. Training error of the reference speed.
30
332
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
dV PV ¼ iPV iBATT a dt diBATT L ¼ V PV a V BAT dt
Transistor T0 is on
ð28Þ
C
ð29Þ
A battery dynamic model is necessary for the prediction of the SOC. The considered model (Fig. 15) is made of a volume capacity, Cbulk, to characterize the battery capacity (Ben Salah and Ouali, 2010). ( V 0 ¼ IRt þ I b Re þ V cb ð30Þ V 0 ¼ IRt þ I s Rs þ V cs The complete state model is described by the following formula: 2 3 2 32 3 0 Cbulk ðR1 e þRs Þ Cbulk ðR1 e þRs Þ V cb V_ cb 6 7 6 76 7 6_ 7 6 76 6 V cs 7 ¼ 6 Csurface 1ðRe þRs Þ Csurface 1ðRe þRs Þ 0 74 V cs 7 5 4 5 4 5 _V o Vo Að3; 1Þ 0 Að3; 3Þ 2 3 Rs
diBATT ¼ V BAT dt dV dc C ¼ iload dt
L
ð35Þ ð36Þ
Transistor T0 is off diBATT ¼ V BAT V dc dt dV dc C ¼ iBATT þ iload dt
L
ð37Þ ð38Þ
The finally equations of the buck converter are: over one switching period Ts, the equations can be combined and represented as follows: diBATT ¼ V BAT þ ða 1Þ V dc dt dV dc C ¼ iload þ ð1 aÞ iBATT dt
L
ð39Þ ð40Þ
C bulk ðRe þRs Þ
6 7 6 7 þ 6 C RðRe þR Þ 7 4 surface e s 5
ð31Þ
For managing the maximum captured in PV and battery power, a supervisory controller is applied (Fig. 18). After adjusting the solar panel to produce the maximum energy (either by setting the reference speed of the induction motor if the panel is connected to the motor or by fixing aopt if the PV is connected to the battery) the supervisory
A where Að3; 1Þ ¼
Að3; 2Þ ¼
C bulk ðRe þ Rs Þ
2
Re
þ
C surface ðRe þ Rs Þ
R2s C bulk Re ðRe þ Rs Þ Re
C surface ðRe þ Rs Þ þ
Að3; 3Þ ¼
Rs
2
2
þ
4. Overall power management strategy
2
Rs 2
C surface ðRe þ Rs Þ
ð32Þ
Rs R t C bulk Re ðRe þ Rs Þ2
Rt Rs Re þ C surface ðRe þ Rs Þ C surface ðRe þ Rs Þ2
Re 1 C bulk Re ðRe þ Rs Þ C surface ðRe þ Rs Þ
ð33Þ ð34Þ
3.3. String III «BATTERY-MOTOPUMP» For the second part of the string, we have used a battery coupled through an adapter (DC/DC converter) to an online inverter to an asynchronous motor (Fig. 16). In order to increase the battery voltage (240 V) at 550 V and to ensure, a regulation of the DC bus voltage, as shown in Fig. 17, we inserted a boost converter. The DC–DC converter “boost”, is controlled by two regulation loops in order to provide the proper voltage to the load. The first loop is used for the regulation of voltage using a PI controller whereas the second is employed for the regulation of battery current using a hysteresis controller (Fig. 17). The behavior of the boost converter can be captured by the following equations:
Fig. 25. Learning process of the duty cycle a.
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
333
Fig. 26. Flow chart of the algorithm controller.
32
1000
30
900 28
Temperature (°C)
Solar illumination (KW/m2)
800 700 600 500 400 300
26
24
22
200 20
100 0 4
6
8
10
12
14
16
t (h) Fig. 27. Solar illumination variation.
18
20
18
4
6
8
10
12
14
t(h) Fig. 28. Temperature variation.
16
18
20
334
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338 10000 Real Energy Neuro-Fuzzy Fuzzy Algorithm
9000
Energy (Wh/day)
8000 7000 6000 5000 4000 3000 2000 1000 0
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Dec
Nov
Dec
Months Fig. 29. Daily maximum energy captured from the PV.
After data acquisition, Water level (H), solar illumination (G), temperature (T) and SOC, six cases are possible:
Table 5 Pump flow with different controllers (m3/h).
Neuro-fuzzy Fuzzy Algorithm
January
February
March
April
May
June
12.42 11.54 10.75
12.95 11.91 11.48
13.7 12.85 12.37
13.61 12.50 12.02
14.03 13.05 12.2
14.13 13.23 12.54
controller determines the state of switches C1, C2 and C3 according to the state of the inputs. The supervisory controller has as inputs:
Water level in the tank. Battery’s state of charge SOC. Solar illumination G. Temperature T. And as outputs:
Reference speed Xref. C1, C2 and C3 switches. The duty cycle a.
If H < Href (Href = 3 m). i. G – 0, the reference speed Xref is given from supervisory controller then C1 = 1, C2 = 1 and C3 = 0: the induction motor is supplied by the PV. ii. If G = 0; and the SOC > SOCmin, it must discharge the battery using the boost converter to power the motor, C1 = 0, C2 = 1 and C3 = 1. iii. If G = 0; and SOC = SOCmin, C1 = 0, C2 = 0 and C3 = 0. H = Href. iv. If G – 0 and SOC < SOCmax, The battery is in charge mode through a buck converter. C1 = 1, C2 = 0 and C3 = 1. v. If there is not any sunlight or SOC, the reference speed Xref = 0. C1 = 0, C2 = 0 and C3 = 0. vi. If G – 0 and SOC = SOCmax, Xref = 0. C1 = 0, C2 = 0 and C3 = 0. If the battery is charged and G – 0, the priority is given to PVP.
Reference Speed Real Speed
150
Speed (rad/s)
100 50 0 -50 -100 7:00
8:00
9:00
10:00
11:00
12:00
13:00
14:00
t(h) Fig. 30. Speed variation with neuro-fuzzy controller.
15:00
16:00
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
335
Flow (m3/h)
15 10 5 0 -5
7:00
8:00
9:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
t(h) Fig. 31. Flow variation with neuro-fuzzy controller.
3
Water level (m)
2.5 Real water level Reference water level
2
1.5
1
0.5
7:00
8:00
9:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
t(h) Fig. 32. Water level variation in the tank with neuro-fuzzy controller.
Three supervisors have been developed, the fuzzy controller, the nero-fuzzy controller and the algorithm controller. 4.1. Fuzzy controller In our application, the fuzzy inference is carried out by using Mam-dani’s method, and the defuzzification uses the centroid method to compute the output of the fuzzy logic. Fuzzy partitions are determined from the calculation of the triangular membership function. Membership functions of input variables of the fuzzy logic are given by Fig. 19 and the membership functions of output variables of the fuzzy logic are shown in Fig. 20. To determine aopt which corresponds to the optimal operation, the fuzzy logic controller uses only as input the temperature T and the sola irradiation G. However, to establish the other outputs, the fuzzy logic controller uses the water level in the tank, the SOC and the solar irradiation. For the fuzzy controller, we gave the error e between the reference level Href and the water level in the tank H.
4.2. Neuro-fuzzy controller The neuro-fuzzy systems can combine the advantages of two complementary techniques. Fuzzy systems provide a good representation of knowledge. The integration of neural networks in these fuzzy systems improves their performance through learning ability of neural networks (Fig. 21). Conversely, the injection of fuzzy rules in neural networks, often criticized for their lack of transparency, clarifies the meaning of the parameters of the network and facilitates their initialization, which represents a considerable gain computing time for their identification. The neuro-fuzzy controller developed in this section has four inputs and five outputs as shown in Fig. 14. Like the fuzzy controller, the two input variables T and G generate the control action aopt which will be applied to the buck converter. The SOC, water level in the tank and G are used for the determination of Xref, C1, C2 and C3. This controller allows automatic generation of fuzzy rules based on Sugeno inference model. The equivalent neuronal structure proposed in matlab of the reference speed is shown in Fig. 22.
336
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
A
a
B
D
C
E
F
1
C1
0.8 0.6 0.4 0.2 0
0
8:00
10:00
12:00
14:00
16:00
18:00
t(h)
b 1
C2
0.8 0.6 0.4 0.2 0
0
8:00
10:00
12:00
14:00
16:00
18:00
14:00
16:00
18:00
t(h)
c 1
C3
0.8 0.6 0.4 0.2 0
0
8:00
10:00
12:00
t(h)
d 0.83 0.82
SOC
0.81 0.8 0.79 0.78 0.77 0.76 0.75 7:00
8:00
9:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
t(h)
Fig. 33. (a)–(c); C1–C3 Switchers variation respectively and (d) SOC variation with neuro-fuzzy controller.
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
337
Learning process and training error of the reference speed are given respectively by Figs. 23 and 24. We also gave the learning process of the duty cycle a as shown in Fig. 25.
We can see clearly that the performance of the fuzzy and algorithm controller is less significant than the neuro-fuzzy controller. For that, we gave some simulation results using it.
4.3. Algorithm controller
Simulation results with neuro-fuzzy controller
In order to determine the relation between solar irradiation G and the approximated reference speed Xref, the curve fitting technique is used (Mansouri et al., 2008). The approximated reference speed is given by (41).
The speed of the machine follows its set point without overshooting showing the effectiveness of the loop speed control which eventually allows achieving the desired flow (Figs. 30 and 31). Speed takes different values depending on the solar irradiation value and according to the state of the water level and battery state of charge. The value of the reference speed applied to the motor must be match to the maximum power supplied by the solar panel as illustrated in Fig. 30. The water level reaches its reference value (Href = 3 m) without error and without overshoot (Fig. 32). Even if a disturbance was introduced, the water level drops and then returns to its reference value. Part A: Before 7:00 morning, the tank is not full and the solar irradiation is not enough. So the battery will power the motor through the boost and the inverter. Consequently we will have: C1 = 0, C2 = 1, C3 = 1, SOC will decrease, as shown in Fig. 33. Part B: In this section, the solar irradiation is sufficient and the tank is not yet filled, so there will be a failover of C1 and C3 (C1 = 1, C2 = 1, C3 = 0). SOC remains constant since there is no charge or discharge of the battery. The motor is supplied by the panel via the inverter and the reference speed will be given by the neuro-fuzzy controller ensuring operation at maximum power. Part C: Once the tank is filled and the battery is not fully charged, so the panel will charge the battery through buck converter and the aopt will be given by the supervision controller. C1 = 1 C2 = 0; C3 = 1. SOC will increase. Part D: In this portion, we introduced a perturbation as shown in Fig. 32, hence the panel will power the motor again. C1 = 1; C2 = 1; C3 = 0. SOC = cst. Part E: The tank is filled again, so we will get the same result as in part C. i.e., C1 = 1 C2 = 0; C3 = 1 and SOC will increase. Part F: A new perturbation is introduced from 17:00 and considering that solar irradiation is not quite enough, then the battery will supply the pump again. As shown in Fig. 33, changing states of switches according to solar illumination G, SOC and the water level in the tank, allows us to ensure a very good power management to avoid energy loses. The surplus energy produced can be exploited in other applications that will be the aim of future works.
Xref ¼ 33:325 þ 0:3511 G 5:7951 104 G2 þ 5:611 107 G3 2:088 1010 G4
ð41Þ
This equation is used when the PV will supply the induction motor but when the battery will power this latter, the reference speed is equal to the nominal speed Xn. The charge and discharge of the battery are described by the following equations. Battery charge: IB ¼ þ
P PV : V BAT
ð42Þ
Battery discharge: IB ¼
Pm : V BAT
ð43Þ
where IB is the battery current, PPV is the PV power, Pm is the motor power and VBAT is the battery voltage. The flow chart of the algorithm controller is described by Fig. 26. 5. Simulations results Based on the above models and control methods, photovoltaic pumping system with storage element providing regulation of water level are modeled and simulated in a user-friendly MATLAB/Simulink environment. To test the behavior of the system subjected to climatic variations, we used an irradiance profile illustrated in Fig. 27 and temperature profile given in Fig. 28. To compare the performance of the three control strategies, we have calculated the daily maximum energy which can be captured from the PV with the three methods, regardless of the load consumption, by applying the value of the reference speed corresponding to the maximum output power (Fig. 29). It is clear from Fig. 29 that the neuro-fuzzy controller is the best one: while the energy gained with neuro-fuzzy controller is about 99.5% of the real energy, the energy gained with fuzzy controller is about 92.4% and the energy gained with algorithm controller is about 82%. A second valorization consists of the calculation of the mean daily pump flow in Table 5 for six months.
6. Conclusion In this work, a photovoltaic pumping system with a battery Lead Acid as storage element has been studied. The
338
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
priority is given to PV. After choosing the optimal configuration, a supervisor has been developed to provide several tasks: extracting maximum PV power acting on the reference speed, regulating the water level in the tank and providing a good energy management. This paper examines three types of supervisors, neuro-fuzzy controller, fuzzy controller and algorithm controller, to show the effectiveness of the first one to extract the maximum energy by optimizing the flow and the energy. Simulation studies were conducted to verify the performance of the system. The simulation results show the effectiveness of the proposed system, which means that if there is an excess power; it will be stored in the battery which increases the power density of the overall system. The neuro fuzzy controller deserves to be implemented in agricultural experimental station located at Sahline. References Akbar, M., Alireza, A., 2014. Comparative study of artificial intelligence techniques for sizing of a hydrogen-based stand-alone photovoltaic/ wind hybrid system. Int. J. Hydrogen Energy 39, 9973–9984. Belmili, H., Haddadi, M., Bacha, S., Almi, M.F., Bendib, B., 2014. Sizing stand-alone photovoltaic–wind hybrid system: techno-economic analysis and optimization. Renew. Sustain. Energy Rev. 30, 821–832. Ben Salah, C., Ouali, M., 2010. Energy management of a hybrid photovoltaic system. Int. J. Energy Res.. Ben Salah, C., Ouali, M., 2011. Comparison of fuzzy logic and neural network in maximum power point tracker for PV systems. Electr. Power Syst. Res. 81, 43–50. Ben Salah, C., Chaabene, M., Ben Ammar, M., 2008. Multi-criteria fuzzy algorithm for energy management of a domestic photovoltaic panel. Renewable Energy 33, 993–1001. Bond, M., Fulle, R.J., Aye, Lu., 2012. Sizing solar home systems for optimal development impact. Energy Policy 42, 699–709. Bouzidi, B., 2011. Viability of solar or wind for water pumping systems in the Algerian Sahara regions – case study Adrar. Renew. Sustain. Energy Rev. 15, 4436–4442. Brian, D.V., Byron, A.N., 2012. Analysis of off-grid hybrid wind turbine/ solar PV water pumping systems. Sol. Energy 86, 1197–1207. Correˆa, T.P., Seleme Jr, S.I., Silva, S.R., 2012. Efficiency optimization in stand-alone photovoltaic pumping system. Renewable Energy 41, 220– 226. Diaf, S., Diaf, D., Belhamel, M., Haddadi, M., Louche, A., 2007. A methodology for optimal sizing of autonomous hybrid PV/wind system. Energy Policy 35, 5708–5718. Ernst, C.S., Hackbarth, A., Madlener, R., Lunz, B., Sauer, D.U., Eckstein, L., 2011. Battery sizing for serial plug-in hybrid electric vehicles: a model-based economic analysis for Germany. Energy Policy 39, 5871–5882. Gibson, T.L., Kelly, N.A., 2010. Solar photovoltaic charging of lithiumion batteries. J. Power Sources 195, 3928–3932. Glasnovic, Z., Margeta, J., 2007. A model for optimal sizing of photovoltaic irrigation water pumping systems. Sol. Energy 81, 904– 916. Glavin, M.E., Hurley, W.G., 2012. Optimisation of a photovoltaic battery ultracapacitor hybrid energy storage system. Sol. Energy 86, 3009– 3020.
Hamidat, A., Benyoucef, B., 2009. Systematic procedures for sizing photovoltaic pumping system, using water tank storage. Energy Policy 37, 1489–1501. Hamidat, A., Benyoucef, B., Hartani, T., 2003. Small-scale irrigation with photovoltaic water pumping system in Sahara regions. Renewable Energy 28, 1081–1096. Hegazy, R., Abou Hashema, M.E., 2013. Sizing of a stand alone concentrated photovoltaic system in Egyptian site. Int. Electr. Power Energy Syst. 45, 325–330. Hengsritawat, V., Tayjasanant, T., Nimpitiwan, N., 2012. Optimal sizing of photovoltaic distributed generators in a distribution system with consideration of solar radiation and harmonic distortion. Int. Electr. Power Energy Syst. 39, 36–47. Kaabeche, A., Ibtiouen, R., 2014. Techno-economic optimization of hybrid photovoltaic/wind/diesel/battery generation in a stand-alone power system. Sol. Energy 103, 171–182. Kaabeche, A., Belhamel, M., Ibtiouen, R., 2011. Techno-economic valuation and optimization of integrated photovoltaic/wind energy conversion system. Sol. Energy 85, 2407–2420. Kaldellis, J.K., Zafirakis, D., Kondili, E., 2010. Optimum sizing of photovoltaic-energy storage systems for autonomous small islands. Int. Electr. Power Energy Syst. 32, 24–36. Kaldellis, J.K., Zafirakis, D., Stavropoulou, V., Kaldelli, El., 2012. Optimum wind- and photovoltaic-based stand-alone systems on the basis of life cycle energy analysis. Energy Policy 50, 345–357. Khiareddine, A., Ben Salah, C., Mimouni, M.F., 2013. New methodology of speed-control of photovoltaic pumping system. Renewable Sustainable Energy 5, 053109-1–053109-27. Koutroulis, E., Kolokotsa, D., Potirakis, A., Kalaitzakis, K., 2006. Methodology for optimal sizing of stand-alone photovoltaic/windgenerator systems using genetic algorithms. Sol. Energy 80, 1072–1088. Mansouri, M.N., Ghanmi, N., Mimouni, M.F., 2008. Commande et analyse des performances d’une station de pompage photovoltaı¨que fonctionnant en re´gime optimal. Revue des Energies Renouvelables, 1– 17. Mazouz, N., Midoun, A., 2011. Control of a DC/DC converter by fuzzy controller for a solar pumping system. Int. Electr. Power Energy Syst. 33, 1623–1630. Meah, K., Fletcher, S., Ula, S., 2008. Solar photovoltaic water pumping for remote locations. Renew. Sustain. Energy Rev. 12, 472–487. Mimouni, M.F., Mansouri, M.N., Benghanem, B., Annabi, M., 2004. Vectorial command of an asynchronous motor fed by a photovoltaic generator. Renewable Energy 29, 433–442. Odeh, I., Yohanis, Y.G., Norton, B., 2006. Economic viability of photovoltaic water pumping systems. Sol. Energy 80, 850–860. Papaefthymiou, S.V., Papathanassiou, S.A., 2014. Optimum sizing of wind-pumped-storage hybrid power stations in island systems. Renewable Energy 64, 187–196. Rekioua, D., Matagne, E., 2012. Optimization of Photovoltaic Power Systems. Springer Verlag, London. Roy, A., Kedare, S.B., Bandyopadhyay, S., 2010. Optimum sizing of wind-battery systems incorporating resource uncertainty. Appl. Energy 87, 2712–2727. Sarrias, R., Ferna´ndez, L.M., Garcı´a, C.A., Jurado, F., 2012. Coordinate operation of power sources in a doubly-fed induction generator wind turbine/battery hybrid power system. J. Power Sour. 205, 354–366. Shaiek, Y., Ben Smida, M., Sakly, A., Mimouni, M.F., 2013. Comparison between conventional methods and GA approach for maximum power point tracking of shaded solar PV generators. Sol. Energy 90, 107–122. Steffen, B., Weber, C., 2013. Efficient storage capacity in power systems with thermal and renewable generation. Energy Econom. 36, 556–567.
ScienceDirect Solar Energy 112 (2015) 319–338 www.elsevier.com/locate/solener
Power management of a photovoltaic/battery pumping system in agricultural experiment station Abla Khiareddine a,⇑, Chokri Ben Salah b,1, Mohamed Faouzi Mimouni a,2 a
Research Unit on Study of Industrial Systems and Renewable Energy (ESIER), Department of Electrical Engineering, National School of Engineers of Monastir, Avenue Ibn EL JAZZAR, 5019 Monastir, Tunisia b Control and Energy Management Lab. (CEMLab), Department of Electrical Engineering, National School of Engineers of Sfax, BP. W, 3038 Sfax, Tunisia Received 31 July 2014; received in revised form 14 November 2014; accepted 19 November 2014
Communicated by: Associate Editor Mukund Patel
Abstract This paper focuses on dynamic modeling, simulation, control and energy management in an agricultural experiment station located at Sahline–Tunisia consisting of a 1.5 kW photovoltaic panel (PV) and a 25 A h Lead Acid battery storage supplying an induction motor coupled to a centrifugal pump as mechanical load. The cost-optimally design and the new suitable power management approach are the two main objectives. An iterative optimization approach namely, the Deficiency of Power Supply Probability (DPSP), the Relative Excess Power Generated (REPG), the Energy Cost (EC) as well as the Total Net Present Cost (TNPC) have been developed in order to find the optimal configuration of PV/battery. To reach the second object, three new supervisory controllers are designed, a neurofuzzy controller, a fuzzy controller and an algorithm controller. In order to show the effectiveness of the first one, a comparison of the three controllers is proposed. The principal objectives of the three supervisory controllers are: (i) the design of an adequate tracking system maximum power point (MPPT) to extract the maximum power which is given by the theory of conservation of energy, (ii) the insurance of the control speed needed for the vectorial control of the induction motor, (iii) the regulation of the water in the tank which is taken as a second storage system and finally (iv) the insurance of the correct operation for all the conversion string in order to optimize the quantity of water pumped. Ó 2014 Elsevier Ltd. All rights reserved.
Keywords: Water level regulation; Supervisory controller; Neuro-fuzzy; Optimization
1. Introduction Energy plays the most vital role in the economic growth, progress, and development, as well as poverty eradication and security of any nation. Energy is an important factor ⇑ Corresponding author. Tel.: +216 95644567.
E-mail addresses: [email protected] (A. Khiareddine), [email protected] (C. Ben Salah), [email protected] (M.F. Mimouni). 1 Tel.: +216 98676138. 2 Tel.: +216 52852959. http://dx.doi.org/10.1016/j.solener.2014.11.020 0038-092X/Ó 2014 Elsevier Ltd. All rights reserved.
in all the sectors of any country’s economy. In the recent years, the diminishing supplies of fossil fuels and their impact on the environment have encouraged a growth in sustainable energies such as wind and solar power; such a significant growth was seen by the photovoltaic industry. Solar power generation is experiencing a remarkable growth in terms of installed power and energy generation in many countries. Autonomous photovoltaic panels are intermittent sustainable energy sources which require energy storage to balance generation and demand, as photovoltaic generation depends on time and weather.
320
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
Traditionally batteries are the most common storage technology for photovoltaic systems (Gibson and Kelly, 2010; Kaldellis et al., 2012; Steffen and Weber, 2013). Standalone photovoltaic systems are often used in remote areas, away from the national grid for water irrigation system (Bouzidi, 2011; Correˆa et al., 2012; Glasnovic and Margeta, 2007; Hamidat et al., 2003; Hegazy and Abou Hashema, 2013; Meah et al., 2008). Many works dealt with the choice of the drive system to interact with the PV source; PV pumping systems based on DC machines, AC machines (Correˆa et al., 2012; Meah et al., 2008), the type of pumps to use and the ways to
control and optimize the whole system. In fact, several control schemes are modeled and described as scalar control, vector control of the induction machine (Khiareddine et al., 2013; Mimouni et al., 2004) and DTC control (Rekioua and Matagne, 2012). Various algorithms have been developed over the years (Ben Salah and Ouali, 2011; Mazouz and Midoun, 2011; Shaiek et al., 2013) in order to determine the maximum power point such as Perturb and Observe Technique, Incremental Conductance Technique, Hill Climbing Control, Curve Fitting Method, Sliding Mode Controller, Fuzzy Logic Technique, Artificial Neural Networks,
Fig. 1. Schematic diagram of photovoltaic pumping system with battery storage.
Fig. 2. Pumping system: (a) the motopump, (b) general view of the tank and (c) greenhouse.
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
Neuro-Fuzzy Method, Genetic Algorithms. . .; a comparison of these different Algorithms has been studied (Rekioua and Matagne, 2012). These control techniques act on the duty cycle to place the generator in its optimal value regardless of changes in meteorological conditions. A combination of Lead Acid batteries, ultra capacitors, wind turbine and photovoltaic panel in a Hybrid Energy Storage System is examined (Brian and Byron, 2012; Diaf et al., 2007; Sarrias et al., 2012). Indeed, using a storage system as Lead Acid batteries, ultracapacitors increase the power density of the overall system. The addition of storage system to the solar and wind energy resources necessitates a power management strategy to provide good organization and distribution of power between the different energy sources as well as to protect the different elements of the conversion string from over/ under design of the system, on the other hand (Ben Salah and Ouali, 2010; Ben Salah et al., 2008). Various optimization techniques of hybrid PV/wind/battery/ultracapacitors systems sizing have been reported in the literature (Akbar and Alireza, 2014; Belmili et al., 2014; Bond et al., 2012; Glavin and Hurley, 2012; Koutroulis
321
et al., 2006). Using the autonomy level of the system, a set of different configurations which meet the load is obtained. The authors developed general methodology by considering design factors such as autonomy for sizing and optimization (Hamidat and Benyoucef, 2009; Hengsritawat et al., 2012; Kaldellis et al., 2010; Odeh et al., 2006; Papaefthymiou and Papathanassiou, 2014; Roy et al., 2010). Indeed, several approaches are developed to achieve the optimal configurations of hybrid systems PV/wind/battery/ ultracapacitors/diesel in terms of technical analysis. Namely, we find the technical approach called loss of power supply probability (LPSP) applied by (Diaf et al., 2007), the iterative optimization technique following the Deficiency of Power Supply Probability (DPSP), the Relative Excess Power Generated (REPG) developed by (Kaabeche et al., 2011), the iterative optimization approach following the Total Energy Deficit (TED) used by (Kaabeche and Ibtiouen, 2014) and the artificial intelligence by (Akbar and Alireza, 2014). For economic analysis, several economic criteria are exploited, such as the Total Net Present Cost (TNPC), the Total Annualized Cost (TAC), Break-Even Distance
8000
Daily load consumption (Wh/d)
7500
Load consumption
7000 6500 6000 5500 5000 4500 4000 3500 3000 January
February
March
April
May
Juin
July
August September October November December
Months
Fig. 3. Variation of the seasonal electricity consumption.
8000 Solar radiation
Daily solar radiation (Wh/m3/d)
7000 6000 5000 4000 3000 2000 1000 0
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Months
Fig. 4. Evolution of the solar illumination for 12 months of the year.
322
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
Fig. 5. Flow chart of the optimal sizing model.
Fig. 6. Relative Excess Power Generated REPG for different battery capacity and for 3 days of autonomy.
Analysis (BEDA), Energy Cost (EC) (Kaabeche et al., 2011; Kaabeche and Ibtiouen, 2014) and the levelized cost of energy (LCE) developed by Diaf et al. (2007). This paper deals with the modeling and control of a hybrid system integrating a photovoltaic panel and
batteries as an energy storage system. In the first part, a description of the whole system and its sizing systems is based on the Deficiency of Power Supply Probability (DPSP), the Relative Excess Power Generated (REPG) for power reliability, the Total Net Present Cost (TNPC)
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
323
Fig. 7. Deficiency of Power Supply Probability DPSP for different battery capacity and for 3 days of autonomy.
Deficiency of Power Supply Probability (%)
5 4.5 Battery capacity
4 20 Ah 25 Ah 50 Ah
3.5 3 2.5 2 1.5 1 0.5 0 1200
1250
1300
1350
1400
1450
1500
1550
1600
1650
PV power (W)
Fig. 8. Deficiency of Power Supply Probability DPSP for different battery capacity and for 1 day of autonomy.
Total Net Present Cost ($)
x 10
4
Optimal configurations
10 8 DPSP = 0%
6 4 2 X: 50 Y: 1500 Z: 1.995e+004
0 1700 1600
250 200
1500
150
1400
PV power (W)
100 1300
50
Battery capacity (Ah)
Fig. 9. System configurations and Total Net Present Cost for the studied system.
and Energy Cost (EC) for system costs. In the second part, the modeling of the three strings is presented. In the third part, specific focus is to be taken on the new supervisory controller in order to coordinate the operation of energy sources fulfilling several functions: firstly
to regulate the water level in the tank, taking into consideration the solar illumination, the battery’s state of charge (SOC) and the water level in the tank. Secondly to determine the reference speed needed for the vector control of induction motor and finally to ensure a good power
324
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
Optimal configurations
Energy Cost ($/kWh)
3 2.5 DPSP = 0%
2 1.5 1 0.5 1550 1500
250 200
1450
150
1400
PV power (W)
100 1350
50
Battery capacity (Ah)
Fig. 10. System configurations and Energy Cost for the studied system.
Table 1 PV modules specifications. Type
EG (eV)
A
Pmax
IMPP (A)
VMPP (V)
ISC (A)
VOC (V)
CT
NOCT
NS
Unit price (US$/W)
Lifetime (year)
Maintenance cost in the first year (%)
SM50-H
1.12
1.5
50Wp
3.15
15.9
3.35
19.8
1.2
45 ± 2 °C
33
3.6
25
1% of price
Table 2 Battery specifications. Nominal capacity (A h) 50
Voltage (V) 240 (12 * 20)
DOD (%)
gB (%)
Unit price (US$/W)
Lifetime (year)
Maintenance cost in the first year (%)
60
85
0.213
5
3% of price
Table 3 DC/AC inverter specification. Power rating (W)
Efficiency (%)
Unit price (US$/W)
Lifetime (year)
Maintenance cost in the first year (%)
1500
90
0.711
10
0
Table 4 Induction motor parameters (1.5 KW). N (tr/min)
p
Rs (X)
Rr (X)
Ls (H)
Lr (H)
M (H)
J (Kg m2)
f (Nm/rds1)
1435
2
5.72
4.2
0.462
0.452
0.44
0.0049
1.5.10–4
management. Photovoltaic pumping in a closed loop will allow us to win a lot of energy which can be used for the battery. Lastly, the simulation results and conclusions are presented. 2. System configuration and unit-sizing An overall power management strategy, provided by the opening and closing of three switches C1, C2, C3, has been designed for various purposes: Use all the energy supplied by the photovoltaic panel. Ensure proper operation for the entire hybrid system. Regulate the water level in the tank.
Ensure a new maximum power point tracking (MPPT) by acting on the reference speed required for the vectorial control of the induction motor. Fig. 1 shows the system configuration for the proposed hybrid pumping system. In the hybrid system, the renewable PV power is taken as primary source, while the battery is used as a backup and storage system. The priority is given to the PV, not only to exploit the entire renewable energy but also to increase the life cycle of the battery. Indeed, the battery is charged in the following situations: The tank is full and the power panel is available.
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
When there is deficit in power generation, the battery will produce energy to the centrifugal pump. An agricultural experiment station located at Sahline region of Tunisia chose to automate the existing pumping system, as shown in Fig. 2, by using a renewable energy and especially the solar one. A photovoltaic pumping system will be used together with a storage component as shown in Fig. 1. This pumping system possesses a 1.5 kW motor and a 3 m tank in order to irrigate many crops cultivated under
325
greenhouses and in open fields such as tomato, pepper, potato, eggplant and melon (Fig. 2). To make the power available during night and cloudy days the battery banks are essential in solar system. Lead Acid batteries are used because of their high performance and long life. Before starting the calculations, we need to confirm the following data: The electricity usage per day Ej (W h). Number of day of autonomy JAUT.
Fig. 11. PV-Pump stand-alone system.
Fig. 12. Synoptic of the vectorial command of an asynchronous motor.
326
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
I
Depth of discharge limits DODMAX always 60%. The energy efficiency of the battery gBAT, (gBAT = 85%, (Rekioua and Matagne, 2012).
C¼
Ej J AUT V BAT DODMAX gBAT
Re
Rs
Vo
Battery of 240 V (VBAT = 12 20) having DODMAX of 60% is considered here. This leads to the calculation of the capacity of the battery C (A h) using the equation below:
Ib
Is
Rt
Csurface
Vcs Cbulk
Vcb
Fig. 15. Battery model diagram.
ð1Þ
During the summer period and depending on the nature of the cultivated crops, the electricity consumed by the agricultural experiment station is twice more than that used for the rest of the year, (Fig. 3). The agricultural experiment station needs a maximum of 5 h/day for filling the tank. That is why the priority is given to solar panels and the battery is used in order to store energy during high electricity production periods and then deliver it to the user at nights or at low sunlight periods. The number of day of autonomy in our case is between one and three days. The variation of solar radiation during the 12 months of the year in the Tunisian Sahel is shown in Fig. 4. To find an optimal combination of PV-Battery to satisfy the load demand, assessment may be performed based on reliability and economy of power supply. 2.1. Reliability criteria In this study, the technical sizing model is developed using the concept of DPSP and REPG to evaluate the reliability of PV-Battery systems.
The flow chart of the optimal sizing model is shown in Fig. 5, (Kaabeche et al., 2011; Diaf et al., 2007). This program has as inputs for each hour Dt: the PV power DPPV, the load consumption DEL, the maximum and minimum energy storage capacity of the battery respectively EBatmax, EBatmin and the maximum number of autonomy day NSDmax. where EBatmax ¼ C B V BAT
ð2Þ
EBatmin ¼ C B V BAT ð1 DODmax Þ
ð3Þ
Every hour, the storage capacity is subject to the following constraints: EBatmin 6 EBat ðtÞ 6 EBatmax
ð4Þ
In this program, the aim is to determine both the yearly DPSP and the yearly REPG, knowing the power of the PV and the load demand throughout the year (8760 h). The power of the panel is fixed since the beginning. The charge and discharge of the battery are described by the following equations. Battery charge:
EL ðtÞ EBat ðtÞ ¼ EBat ðt 1Þ ð1 rÞ þ EPV ðtÞ gB : ginv Battery discharge: EBat ðtÞ ¼ EBat ðt 1Þ ð1 rÞ
EL ðtÞ EPV ðtÞ gB : ginv
ð5Þ
ð6Þ
where r is the hourly self-discharge rate, ginv and gB present respectively the inverter efficiency and the battery bank efficiency. Fig. 13. PV-Battery stand-alone system.
The Excess Power Generated (EPG) is given by:
Fig. 14. PV-Battery stand-alone system and the buck converter.
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
EPGðtÞ ¼ EPV ðtÞ
EL ðtÞ þ ðEBatmax EBat ðt 1ÞÞ=gB ginv
ð7Þ
The Relative Excess Power Generated (REPG), defined as the ratio of power excess to the sum of load demand, is calculated by the following equation: , T T X X REPG ¼ EPGðtÞ EL ðtÞ ð8Þ t¼1
t¼1
The deficiency power supply (DPS) at hour t, is given by: DPSðtÞ ¼ EL ðtÞ ½EPV ðtÞ þ EBat ðt 1Þ EBatmin ginv
ð9Þ
And the Deficiency of Power Supply Probability (DPSP) can be expressed as: , T T X X DPSP ¼ DPSðtÞ EL ðtÞ ð10Þ t¼1
t¼1
327
The Relative Excess Power Generated REPG and the Deficiency of Power Supply Probability DPSP for different battery capacity and different power panel with NSdmax = 3 days are given respectively in Figs. 6 and 7. We can notice that the DPSP has a hyperbolic shape and it is clear that the DPSP increase with the decreases of the REPG. To obtain a DPSP of 0%, corresponding to the total autonomy of PV/Battery system, two solutions are possible, either to take a battery capacity equal to or higher than 50 A h, with a panel of 1500 W or more, or take a lower PV 1500 W with a strictly upper to 50 A h capacity. We can see that there is a compromise between increasing the PV power and increasing the battery capacity to achieve the reliability requested. The Deficiency of Power Supply Probability DPSP for different battery capacity and different power panel with NSdmax = 1 days is given by Fig. 8.
Fig. 16. Battery–Pump stand-alone system.
Fig. 17. Boost converter and DC bus voltage regulation.
328
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
As shown in Fig. 8, for 1 day of autonomy and with a capacity of less than 25 A h, we cannot reach the load autonomy (DPSP of 0%) whatever the PV power is. Although from 25 A h and more, it can be achieved. To choose the right pair, the economic analysis should be made to obtain the low cost with the total autonomy of the system. The economic criteria will be detailed in the next section.
where Crec is the recurring cost and Cnon-rec is the nonrecurring cost. d (8%) and e (4%) are respectively the interest and escalation rates. n is the system life period in years (25 years). P is the number of years between two successive payments for non-recurring costs. dadj, which is the adjusted interest rate, is given as follows:
2.2. Economic criteria
The second and the most excellent economic profitability indicators of system cost analysis is the Energy Cost (EC). EC can be defined as:
The economical approach based on the Total Net Present Cost (TNPC) and Energy Cost (EC), developed by Kaabeche and Ibtiouen (2014), is considered here for system configurations. The Total Net Present Cost (TNPC), which is a key financial indicator for economic viability appreciation of an investment project, can be expressed as follows: TNPCð$Þ ¼ IC þ PWCnon-rec þ PWCrec
ð11Þ
where IC is the initial cost of the system components, PWCnon-rec and PWCrec are factors for the conversion of the non-recurring and recurring costs to their present worth. These factors are given by Eqs. (12) and (13). h ih in 1þe 1þe 1 1þd 1þd h i PWCrec ¼ C rec ð12Þ 1þe 1 1þd h i h in 1þe 1þe 1 1þd adj 1þd adj h i ð13Þ PWCnon-rec ¼ C non-rec 1þe 1 1þd adj
d adj ¼
ð1 þ dÞP ð1 þ dÞ
P 1
1
ð14Þ
TNPC CRF ECð$=KW hÞ ¼ P8760 t¼1 E GenðtÞ
ð15Þ
where CRF is the capital recovery factor which is given by: CRFðd; nÞ ¼
dð1 þ dÞ
n
ð1 þ dÞn 1
ð16Þ
Figs. 9 and 10 give the TNPC and the EC respectively, with different battery capacity and different PV power for the optimal configurations. A careful observation into Figs. 9 and 10 show that the optimal configuration, which insures the system reliability requirement with the lowest values of TNPC and EC for three days of autonomy, is the battery capacity of 50 A h and 1500 W PV power. This couple has a REPG of 45.2%, a TNPC of 19,946 (US$) and EC of 0.61 (US$/ kW h). We can find another couple that provides 0% but the cost will be more significant.
Fig. 18. Supervisor controller.
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
For one day of autonomy, the optimal configuration is the battery capacity of 25 A h and 1500 W PV power. Tables 1–4 depict the technical and economical parameters of the system components. Once the optimal configuration is chosen, namely battery of 50 A h with a PV of 1500 W for three days of autonomy and a battery of 25 A h with a PV of 1500 W for one day of autonomy. We will search the most appropriate strategy to extract the maximum power from the PV and to ensure the smooth operation of the total system. The energy management strategy is described below using as simulation parameters a PV of 1500 W with a battery of 25 A h for one day of autonomy. 3. Models of system components Different energy sources are connected to the dc bus through appropriate power electronic interfacing circuits. The three strings will be detailed in the following parts.
329
where Kch is the proportionality constant of the pump, PPVMPPT the maximum power supplied by the PVP, gMOT is the induction motor efficiency and gINV is inverter efficiency which varies from 90% to 97%; we consider here an inverter efficiency of 90% (Ernst et al., 2011). To find the reference speed that will allow us first of all to get the maximum power and secondly to regulate the water level, we used the neuro fuzzy controller which as input. 3.2. String II «PV-BATTERY» To improve the performance of a photovoltaic system, a power converter was inserted between the PV generator and the battery (Fig. 13). The DC–DC converter “buck”, controlled by a tracking algorithm of maximum power point tracker MPPT, provides the proper voltage to the load.
3.1. String I «PV-MOTOPUMP» In the first part of this paper, we presented a description of the structure which consists of a photovoltaic generator coupled through an inverter to an induction motor (Fig. 11). Indeed, the PV is directly connected to a threephase inverter which is controlled by the SVPWM. Given the performance of the vector control, this last has been developed to control the induction motor. The synoptic of the vectorial command of an asynchronous motor is given in Fig. 12. In the majority of research works, MPPT is provided by the variation of the duty cycle a in order to control DC/DC converter. In this work, the MPPT is given by the theory of conservation of energy, in other words we must determine the reference speed Xref needed for the vectorial control of the induction motor which corresponds to the maximum reached by the solar panel power. During the programming of the supervisor with either fuzzy logic or neuro-fuzzy, we need a relationship between the maximum power supplied by the PV and the reference speed Xref. Mechanical power of the induction motor PM is given as: P M ¼ X Cr
ð17Þ
And the resistant torque of the load Cr is expressed by the Eq. (18): C r ¼ K ch :X2
ð18Þ 3
P M ¼ K ch X
ð19Þ
P M ¼ gINV gMOT P PVMPPT
ð20Þ
Then the speed is given by Eq. (21): rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3 gINV gMOT P PVMPPT X¼ K ch
ð21Þ
Fig. 19. Membership functions of input variables of the fuzzy logic.
330
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
Fig. 20. Membership functions of output variables of the fuzzy logic.
Fig. 21. Principle of neuro fuzzy system.
The duty cycle a is obtained crossing the supervisory controller using as inputs the solar irradiation G and the temperature T. The behavior of the buck converter shown in Fig. 14 can be captured by the following equations: Transistor is on
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
331
Fig. 22. Neuronal structure of the proposed model in Matlab.
dV PV ¼ iPV iBATT dt diBATT L ¼ V PV V BAT dt
ð22Þ
C Training data : FIS output 200 150
Transistor is off
100 50
0
500
1000
1500
2000
2500
3000
ð24Þ ð25Þ
Over one switching period Ts, the equations can be combined and represented as follows:
Tesng data : FIS output 200 150
dV PV ¼ ðiPV iBATT Þ a þ iPV ð1 aÞ dt diBATT L ¼ ðV PV V BAT Þ a þ ðV BAT Þ ð1 aÞ dt
C
100 50 0 -50
dV PV ¼ iPV dt diBATT L ¼ V BAT dt
C
0 -50
ð23Þ
ð26Þ ð27Þ
The finally equations of the buck converter are: 0
500
1000
1500
2000
2500
Checking data : FIS output 1.6 1.55 1.5 1.45
50
1.4
0
1.35
0
500
1000
1500
2000
Fig. 23. Learning process of the reference speed.
2500
0
5
10
15
20
25
Fig. 24. Training error of the reference speed.
30
332
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
dV PV ¼ iPV iBATT a dt diBATT L ¼ V PV a V BAT dt
Transistor T0 is on
ð28Þ
C
ð29Þ
A battery dynamic model is necessary for the prediction of the SOC. The considered model (Fig. 15) is made of a volume capacity, Cbulk, to characterize the battery capacity (Ben Salah and Ouali, 2010). ( V 0 ¼ IRt þ I b Re þ V cb ð30Þ V 0 ¼ IRt þ I s Rs þ V cs The complete state model is described by the following formula: 2 3 2 32 3 0 Cbulk ðR1 e þRs Þ Cbulk ðR1 e þRs Þ V cb V_ cb 6 7 6 76 7 6_ 7 6 76 6 V cs 7 ¼ 6 Csurface 1ðRe þRs Þ Csurface 1ðRe þRs Þ 0 74 V cs 7 5 4 5 4 5 _V o Vo Að3; 1Þ 0 Að3; 3Þ 2 3 Rs
diBATT ¼ V BAT dt dV dc C ¼ iload dt
L
ð35Þ ð36Þ
Transistor T0 is off diBATT ¼ V BAT V dc dt dV dc C ¼ iBATT þ iload dt
L
ð37Þ ð38Þ
The finally equations of the buck converter are: over one switching period Ts, the equations can be combined and represented as follows: diBATT ¼ V BAT þ ða 1Þ V dc dt dV dc C ¼ iload þ ð1 aÞ iBATT dt
L
ð39Þ ð40Þ
C bulk ðRe þRs Þ
6 7 6 7 þ 6 C RðRe þR Þ 7 4 surface e s 5
ð31Þ
For managing the maximum captured in PV and battery power, a supervisory controller is applied (Fig. 18). After adjusting the solar panel to produce the maximum energy (either by setting the reference speed of the induction motor if the panel is connected to the motor or by fixing aopt if the PV is connected to the battery) the supervisory
A where Að3; 1Þ ¼
Að3; 2Þ ¼
C bulk ðRe þ Rs Þ
2
Re
þ
C surface ðRe þ Rs Þ
R2s C bulk Re ðRe þ Rs Þ Re
C surface ðRe þ Rs Þ þ
Að3; 3Þ ¼
Rs
2
2
þ
4. Overall power management strategy
2
Rs 2
C surface ðRe þ Rs Þ
ð32Þ
Rs R t C bulk Re ðRe þ Rs Þ2
Rt Rs Re þ C surface ðRe þ Rs Þ C surface ðRe þ Rs Þ2
Re 1 C bulk Re ðRe þ Rs Þ C surface ðRe þ Rs Þ
ð33Þ ð34Þ
3.3. String III «BATTERY-MOTOPUMP» For the second part of the string, we have used a battery coupled through an adapter (DC/DC converter) to an online inverter to an asynchronous motor (Fig. 16). In order to increase the battery voltage (240 V) at 550 V and to ensure, a regulation of the DC bus voltage, as shown in Fig. 17, we inserted a boost converter. The DC–DC converter “boost”, is controlled by two regulation loops in order to provide the proper voltage to the load. The first loop is used for the regulation of voltage using a PI controller whereas the second is employed for the regulation of battery current using a hysteresis controller (Fig. 17). The behavior of the boost converter can be captured by the following equations:
Fig. 25. Learning process of the duty cycle a.
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
333
Fig. 26. Flow chart of the algorithm controller.
32
1000
30
900 28
Temperature (°C)
Solar illumination (KW/m2)
800 700 600 500 400 300
26
24
22
200 20
100 0 4
6
8
10
12
14
16
t (h) Fig. 27. Solar illumination variation.
18
20
18
4
6
8
10
12
14
t(h) Fig. 28. Temperature variation.
16
18
20
334
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338 10000 Real Energy Neuro-Fuzzy Fuzzy Algorithm
9000
Energy (Wh/day)
8000 7000 6000 5000 4000 3000 2000 1000 0
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Dec
Nov
Dec
Months Fig. 29. Daily maximum energy captured from the PV.
After data acquisition, Water level (H), solar illumination (G), temperature (T) and SOC, six cases are possible:
Table 5 Pump flow with different controllers (m3/h).
Neuro-fuzzy Fuzzy Algorithm
January
February
March
April
May
June
12.42 11.54 10.75
12.95 11.91 11.48
13.7 12.85 12.37
13.61 12.50 12.02
14.03 13.05 12.2
14.13 13.23 12.54
controller determines the state of switches C1, C2 and C3 according to the state of the inputs. The supervisory controller has as inputs:
Water level in the tank. Battery’s state of charge SOC. Solar illumination G. Temperature T. And as outputs:
Reference speed Xref. C1, C2 and C3 switches. The duty cycle a.
If H < Href (Href = 3 m). i. G – 0, the reference speed Xref is given from supervisory controller then C1 = 1, C2 = 1 and C3 = 0: the induction motor is supplied by the PV. ii. If G = 0; and the SOC > SOCmin, it must discharge the battery using the boost converter to power the motor, C1 = 0, C2 = 1 and C3 = 1. iii. If G = 0; and SOC = SOCmin, C1 = 0, C2 = 0 and C3 = 0. H = Href. iv. If G – 0 and SOC < SOCmax, The battery is in charge mode through a buck converter. C1 = 1, C2 = 0 and C3 = 1. v. If there is not any sunlight or SOC, the reference speed Xref = 0. C1 = 0, C2 = 0 and C3 = 0. vi. If G – 0 and SOC = SOCmax, Xref = 0. C1 = 0, C2 = 0 and C3 = 0. If the battery is charged and G – 0, the priority is given to PVP.
Reference Speed Real Speed
150
Speed (rad/s)
100 50 0 -50 -100 7:00
8:00
9:00
10:00
11:00
12:00
13:00
14:00
t(h) Fig. 30. Speed variation with neuro-fuzzy controller.
15:00
16:00
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
335
Flow (m3/h)
15 10 5 0 -5
7:00
8:00
9:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
t(h) Fig. 31. Flow variation with neuro-fuzzy controller.
3
Water level (m)
2.5 Real water level Reference water level
2
1.5
1
0.5
7:00
8:00
9:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
t(h) Fig. 32. Water level variation in the tank with neuro-fuzzy controller.
Three supervisors have been developed, the fuzzy controller, the nero-fuzzy controller and the algorithm controller. 4.1. Fuzzy controller In our application, the fuzzy inference is carried out by using Mam-dani’s method, and the defuzzification uses the centroid method to compute the output of the fuzzy logic. Fuzzy partitions are determined from the calculation of the triangular membership function. Membership functions of input variables of the fuzzy logic are given by Fig. 19 and the membership functions of output variables of the fuzzy logic are shown in Fig. 20. To determine aopt which corresponds to the optimal operation, the fuzzy logic controller uses only as input the temperature T and the sola irradiation G. However, to establish the other outputs, the fuzzy logic controller uses the water level in the tank, the SOC and the solar irradiation. For the fuzzy controller, we gave the error e between the reference level Href and the water level in the tank H.
4.2. Neuro-fuzzy controller The neuro-fuzzy systems can combine the advantages of two complementary techniques. Fuzzy systems provide a good representation of knowledge. The integration of neural networks in these fuzzy systems improves their performance through learning ability of neural networks (Fig. 21). Conversely, the injection of fuzzy rules in neural networks, often criticized for their lack of transparency, clarifies the meaning of the parameters of the network and facilitates their initialization, which represents a considerable gain computing time for their identification. The neuro-fuzzy controller developed in this section has four inputs and five outputs as shown in Fig. 14. Like the fuzzy controller, the two input variables T and G generate the control action aopt which will be applied to the buck converter. The SOC, water level in the tank and G are used for the determination of Xref, C1, C2 and C3. This controller allows automatic generation of fuzzy rules based on Sugeno inference model. The equivalent neuronal structure proposed in matlab of the reference speed is shown in Fig. 22.
336
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
A
a
B
D
C
E
F
1
C1
0.8 0.6 0.4 0.2 0
0
8:00
10:00
12:00
14:00
16:00
18:00
t(h)
b 1
C2
0.8 0.6 0.4 0.2 0
0
8:00
10:00
12:00
14:00
16:00
18:00
14:00
16:00
18:00
t(h)
c 1
C3
0.8 0.6 0.4 0.2 0
0
8:00
10:00
12:00
t(h)
d 0.83 0.82
SOC
0.81 0.8 0.79 0.78 0.77 0.76 0.75 7:00
8:00
9:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
t(h)
Fig. 33. (a)–(c); C1–C3 Switchers variation respectively and (d) SOC variation with neuro-fuzzy controller.
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
337
Learning process and training error of the reference speed are given respectively by Figs. 23 and 24. We also gave the learning process of the duty cycle a as shown in Fig. 25.
We can see clearly that the performance of the fuzzy and algorithm controller is less significant than the neuro-fuzzy controller. For that, we gave some simulation results using it.
4.3. Algorithm controller
Simulation results with neuro-fuzzy controller
In order to determine the relation between solar irradiation G and the approximated reference speed Xref, the curve fitting technique is used (Mansouri et al., 2008). The approximated reference speed is given by (41).
The speed of the machine follows its set point without overshooting showing the effectiveness of the loop speed control which eventually allows achieving the desired flow (Figs. 30 and 31). Speed takes different values depending on the solar irradiation value and according to the state of the water level and battery state of charge. The value of the reference speed applied to the motor must be match to the maximum power supplied by the solar panel as illustrated in Fig. 30. The water level reaches its reference value (Href = 3 m) without error and without overshoot (Fig. 32). Even if a disturbance was introduced, the water level drops and then returns to its reference value. Part A: Before 7:00 morning, the tank is not full and the solar irradiation is not enough. So the battery will power the motor through the boost and the inverter. Consequently we will have: C1 = 0, C2 = 1, C3 = 1, SOC will decrease, as shown in Fig. 33. Part B: In this section, the solar irradiation is sufficient and the tank is not yet filled, so there will be a failover of C1 and C3 (C1 = 1, C2 = 1, C3 = 0). SOC remains constant since there is no charge or discharge of the battery. The motor is supplied by the panel via the inverter and the reference speed will be given by the neuro-fuzzy controller ensuring operation at maximum power. Part C: Once the tank is filled and the battery is not fully charged, so the panel will charge the battery through buck converter and the aopt will be given by the supervision controller. C1 = 1 C2 = 0; C3 = 1. SOC will increase. Part D: In this portion, we introduced a perturbation as shown in Fig. 32, hence the panel will power the motor again. C1 = 1; C2 = 1; C3 = 0. SOC = cst. Part E: The tank is filled again, so we will get the same result as in part C. i.e., C1 = 1 C2 = 0; C3 = 1 and SOC will increase. Part F: A new perturbation is introduced from 17:00 and considering that solar irradiation is not quite enough, then the battery will supply the pump again. As shown in Fig. 33, changing states of switches according to solar illumination G, SOC and the water level in the tank, allows us to ensure a very good power management to avoid energy loses. The surplus energy produced can be exploited in other applications that will be the aim of future works.
Xref ¼ 33:325 þ 0:3511 G 5:7951 104 G2 þ 5:611 107 G3 2:088 1010 G4
ð41Þ
This equation is used when the PV will supply the induction motor but when the battery will power this latter, the reference speed is equal to the nominal speed Xn. The charge and discharge of the battery are described by the following equations. Battery charge: IB ¼ þ
P PV : V BAT
ð42Þ
Battery discharge: IB ¼
Pm : V BAT
ð43Þ
where IB is the battery current, PPV is the PV power, Pm is the motor power and VBAT is the battery voltage. The flow chart of the algorithm controller is described by Fig. 26. 5. Simulations results Based on the above models and control methods, photovoltaic pumping system with storage element providing regulation of water level are modeled and simulated in a user-friendly MATLAB/Simulink environment. To test the behavior of the system subjected to climatic variations, we used an irradiance profile illustrated in Fig. 27 and temperature profile given in Fig. 28. To compare the performance of the three control strategies, we have calculated the daily maximum energy which can be captured from the PV with the three methods, regardless of the load consumption, by applying the value of the reference speed corresponding to the maximum output power (Fig. 29). It is clear from Fig. 29 that the neuro-fuzzy controller is the best one: while the energy gained with neuro-fuzzy controller is about 99.5% of the real energy, the energy gained with fuzzy controller is about 92.4% and the energy gained with algorithm controller is about 82%. A second valorization consists of the calculation of the mean daily pump flow in Table 5 for six months.
6. Conclusion In this work, a photovoltaic pumping system with a battery Lead Acid as storage element has been studied. The
338
A. Khiareddine et al. / Solar Energy 112 (2015) 319–338
priority is given to PV. After choosing the optimal configuration, a supervisor has been developed to provide several tasks: extracting maximum PV power acting on the reference speed, regulating the water level in the tank and providing a good energy management. This paper examines three types of supervisors, neuro-fuzzy controller, fuzzy controller and algorithm controller, to show the effectiveness of the first one to extract the maximum energy by optimizing the flow and the energy. Simulation studies were conducted to verify the performance of the system. The simulation results show the effectiveness of the proposed system, which means that if there is an excess power; it will be stored in the battery which increases the power density of the overall system. The neuro fuzzy controller deserves to be implemented in agricultural experimental station located at Sahline. References Akbar, M., Alireza, A., 2014. Comparative study of artificial intelligence techniques for sizing of a hydrogen-based stand-alone photovoltaic/ wind hybrid system. Int. J. Hydrogen Energy 39, 9973–9984. Belmili, H., Haddadi, M., Bacha, S., Almi, M.F., Bendib, B., 2014. Sizing stand-alone photovoltaic–wind hybrid system: techno-economic analysis and optimization. Renew. Sustain. Energy Rev. 30, 821–832. Ben Salah, C., Ouali, M., 2010. Energy management of a hybrid photovoltaic system. Int. J. Energy Res.. Ben Salah, C., Ouali, M., 2011. Comparison of fuzzy logic and neural network in maximum power point tracker for PV systems. Electr. Power Syst. Res. 81, 43–50. Ben Salah, C., Chaabene, M., Ben Ammar, M., 2008. Multi-criteria fuzzy algorithm for energy management of a domestic photovoltaic panel. Renewable Energy 33, 993–1001. Bond, M., Fulle, R.J., Aye, Lu., 2012. Sizing solar home systems for optimal development impact. Energy Policy 42, 699–709. Bouzidi, B., 2011. Viability of solar or wind for water pumping systems in the Algerian Sahara regions – case study Adrar. Renew. Sustain. Energy Rev. 15, 4436–4442. Brian, D.V., Byron, A.N., 2012. Analysis of off-grid hybrid wind turbine/ solar PV water pumping systems. Sol. Energy 86, 1197–1207. Correˆa, T.P., Seleme Jr, S.I., Silva, S.R., 2012. Efficiency optimization in stand-alone photovoltaic pumping system. Renewable Energy 41, 220– 226. Diaf, S., Diaf, D., Belhamel, M., Haddadi, M., Louche, A., 2007. A methodology for optimal sizing of autonomous hybrid PV/wind system. Energy Policy 35, 5708–5718. Ernst, C.S., Hackbarth, A., Madlener, R., Lunz, B., Sauer, D.U., Eckstein, L., 2011. Battery sizing for serial plug-in hybrid electric vehicles: a model-based economic analysis for Germany. Energy Policy 39, 5871–5882. Gibson, T.L., Kelly, N.A., 2010. Solar photovoltaic charging of lithiumion batteries. J. Power Sources 195, 3928–3932. Glasnovic, Z., Margeta, J., 2007. A model for optimal sizing of photovoltaic irrigation water pumping systems. Sol. Energy 81, 904– 916. Glavin, M.E., Hurley, W.G., 2012. Optimisation of a photovoltaic battery ultracapacitor hybrid energy storage system. Sol. Energy 86, 3009– 3020.
Hamidat, A., Benyoucef, B., 2009. Systematic procedures for sizing photovoltaic pumping system, using water tank storage. Energy Policy 37, 1489–1501. Hamidat, A., Benyoucef, B., Hartani, T., 2003. Small-scale irrigation with photovoltaic water pumping system in Sahara regions. Renewable Energy 28, 1081–1096. Hegazy, R., Abou Hashema, M.E., 2013. Sizing of a stand alone concentrated photovoltaic system in Egyptian site. Int. Electr. Power Energy Syst. 45, 325–330. Hengsritawat, V., Tayjasanant, T., Nimpitiwan, N., 2012. Optimal sizing of photovoltaic distributed generators in a distribution system with consideration of solar radiation and harmonic distortion. Int. Electr. Power Energy Syst. 39, 36–47. Kaabeche, A., Ibtiouen, R., 2014. Techno-economic optimization of hybrid photovoltaic/wind/diesel/battery generation in a stand-alone power system. Sol. Energy 103, 171–182. Kaabeche, A., Belhamel, M., Ibtiouen, R., 2011. Techno-economic valuation and optimization of integrated photovoltaic/wind energy conversion system. Sol. Energy 85, 2407–2420. Kaldellis, J.K., Zafirakis, D., Kondili, E., 2010. Optimum sizing of photovoltaic-energy storage systems for autonomous small islands. Int. Electr. Power Energy Syst. 32, 24–36. Kaldellis, J.K., Zafirakis, D., Stavropoulou, V., Kaldelli, El., 2012. Optimum wind- and photovoltaic-based stand-alone systems on the basis of life cycle energy analysis. Energy Policy 50, 345–357. Khiareddine, A., Ben Salah, C., Mimouni, M.F., 2013. New methodology of speed-control of photovoltaic pumping system. Renewable Sustainable Energy 5, 053109-1–053109-27. Koutroulis, E., Kolokotsa, D., Potirakis, A., Kalaitzakis, K., 2006. Methodology for optimal sizing of stand-alone photovoltaic/windgenerator systems using genetic algorithms. Sol. Energy 80, 1072–1088. Mansouri, M.N., Ghanmi, N., Mimouni, M.F., 2008. Commande et analyse des performances d’une station de pompage photovoltaı¨que fonctionnant en re´gime optimal. Revue des Energies Renouvelables, 1– 17. Mazouz, N., Midoun, A., 2011. Control of a DC/DC converter by fuzzy controller for a solar pumping system. Int. Electr. Power Energy Syst. 33, 1623–1630. Meah, K., Fletcher, S., Ula, S., 2008. Solar photovoltaic water pumping for remote locations. Renew. Sustain. Energy Rev. 12, 472–487. Mimouni, M.F., Mansouri, M.N., Benghanem, B., Annabi, M., 2004. Vectorial command of an asynchronous motor fed by a photovoltaic generator. Renewable Energy 29, 433–442. Odeh, I., Yohanis, Y.G., Norton, B., 2006. Economic viability of photovoltaic water pumping systems. Sol. Energy 80, 850–860. Papaefthymiou, S.V., Papathanassiou, S.A., 2014. Optimum sizing of wind-pumped-storage hybrid power stations in island systems. Renewable Energy 64, 187–196. Rekioua, D., Matagne, E., 2012. Optimization of Photovoltaic Power Systems. Springer Verlag, London. Roy, A., Kedare, S.B., Bandyopadhyay, S., 2010. Optimum sizing of wind-battery systems incorporating resource uncertainty. Appl. Energy 87, 2712–2727. Sarrias, R., Ferna´ndez, L.M., Garcı´a, C.A., Jurado, F., 2012. Coordinate operation of power sources in a doubly-fed induction generator wind turbine/battery hybrid power system. J. Power Sour. 205, 354–366. Shaiek, Y., Ben Smida, M., Sakly, A., Mimouni, M.F., 2013. Comparison between conventional methods and GA approach for maximum power point tracking of shaded solar PV generators. Sol. Energy 90, 107–122. Steffen, B., Weber, C., 2013. Efficient storage capacity in power systems with thermal and renewable generation. Energy Econom. 36, 556–567.