The Performance Of Photovoltaic/wind Turbine Generators/electric Utility

OPERATION AND CONTROL STRATEGY OF PV/WTG/EU HYBRID ELECTRIC POWER SYSTEM USING NEURAL NETWORKS Prof. Dr. H. H. El-Tamaly Adel A. El-baset Mohammed [email protected] [email protected] Faculty of Engineering, Elminia University, Elminia, Egypt, 61111

Abstract The performance of Photovoltaic/Wind Turbine Generators/Electric Utility (PV/WTG/EU) Hybrid Electric Power System (HEPS) can be improved through an application of advanced control method. This paper introduces an application of an artificial neural network on the operation and control strategy of the PV/WTG/EU HEPS to improve system efficiency and reliability. The neural network is utilized to identify the optimal operating voltage of PV/WTG/EU HEPS. The On-line computer programmed based on neural network generates the control signals. These signals will be generated according to insolation, temperature, wind speed, load demand and EU source. The generated power from PV/WTG HEPS has been calculated by a computer program under known insolation, temperature, wind speed and load demand. The computer program, which proposed here and applied to carry out these calculations, is based on the minimization of the energy purchase from EU. Different feed forward neural network architectures are trained and tested with data containing a variety of operation patterns. A simulation is carried out over one year using the hourly data of the insolation, temperature, wind speed, and load demand at El'Zafarana site, Egypt as a case study. The results show that the selected neural network architecture (5+10+5) gives reasonably accurate operation of PV/WTG/EU HEPS. Using this strategy maximizes the output power from PV/WTG then the efficiency of the whole HEPS will be improved. Also using this strategy minimizes the lost time of switching ON and switching OFF as well as the period between the switching ON and switching OFF. Then, the reliability of the whole HEPS will be improved. Keywords Hybrid Photovoltaic, Wind, Electric utility, Neural Network 1- Introduction As energy demands around the world increase, the need for a renewable energy source that will not harm the environment has been increased. Some projections indicate that the global energy demand will almost triple by 2050. So, oil can only supply the world for up to 150 years. Using renewable energy is one way to meet the future need. So, we can say that the renewable energy is the fuel of the future [1]. K. Mitchell, J. Rizk and M. Nagrial (2002) [2] discussed the potential system benefits of simple predictive control routines, using seasonal averaged load wind and solar data, in both stand-

alone and grid connected modes. O. Omari1, et. al. (2003) [3] the DC-coupled PV/WTG HEPS and its relation to the new criterion discussed, control and management strategies that applied to a simulation model of an example of this type presented. H. El-Tamaly and F. M. El-Kady (2000) [4]: In this Reference a design for PV system and WTG to be interfaced with EU investigated. Optimum design, cost and reliability for different penetration ratio estimated. E. Koutroulis, et. al. (2001) [5]: In this Reference a hybrid renewable energy system described which consists of twelve PV panels and a WTG and can supply continuous electric power of 1.5 kW. An energy management system have developed for this purpose in order to maximize the electric power produced using a MPPT method and consists of Buck-type DC/DC converters controlled by a microcontroller. This paper presents an application of an artificial neural network, ANN on the operation and control strategy of the PV/WTG interconnected with EU to improve system efficiency and reliability. Different FFNN architectures have been trained and tested with data containing a variety of operation patterns. 2- System Model 2-1 Modeling of PV/WTG The mathematical equations describing the modeling of photovoltaic array and wind turbine generator is given in [6], [7], [8], [9]. A typical modeling of PV/WTG HEPS, in a grid-connected situation, is shown in Fig 1. The design of PV/WTG HEPS interconnected to EU depends on dividing the load into two parts between photovoltaic (PV) and wind turbine generator (WTG). The power generated by PV solar cells array and WTG at any time, (t) can be expressed as the following:-

P (t) = Ppv (t) + PWTG (t) gtotal PL (t)

=P (t) ± P (t) gtotal g

(1) (2)

For the energy balance, the following conditions must be satisfied: 8760 8760 ∑ Pgtotal (t) − ∑ PL (t) ≅ 0 t =1 t =1 8760 ∑ P (t) ≅ 0 t =1 t 8760 ∑ P (t) ≅ 0 t =1 f

Where; Pgtotal(t) Ppv(t) PWTG(t) Pg(t) Pt(t) Pf(t) PL(t)

(3) (4) (5)

: The total generated power from PV/WTG, kW. : The output power from photovoltaic array, kW. : The output power from wind turbine generated power, kW : The electrical utility power, kW. : The surplus power, kW : The deficit power ,kW : The load demand, kW

The following operating strategy have employed by using ANN as follows: Mode 1: When there is a radiation but there is no wind speed, i.e. there is no power from WTGs then the load demand will be supplied from PV system. The surplus energy supplied to EU and deficit energy taken from EU. (S1=ON, S2=OFF, S3 =ON, S4 =ON, S5 =OFF) in the case of surplus energy or (S1 =ON, S2 =ON, S3 =OFF, S4 =ON, S5 =OFF) in the case of deficit energy. Mode 2: When there is no radiation, i.e. there is no PV power and there is a wind speed then the load demand will be supplied from WTG. The surplus energy supplied to EU and deficit energy taken from EU. (S1 =ON, S2 =OFF, S3 =ON, S4 =OFF, S5 =ON) in the case of surplus energy or (S1 =ON, S2 =ON, S3=OFF, S4=OFF, S5=ON) in the case of deficit energy. Mode 3: When there is a radiation and there is a wind speed then the load demand will be supplied from PV/WTG and surplus/deficit energy supplied/taken from EU. (S1 =ON, S2 =OFF, S3 =ON, S4 =ON, S5 =ON) in the case of surplus energy or (S1 =ON, S2 =ON, S3 =OFF, S4=ON, S5 =ON) in the case of deficit energy.

Mode 4: When there is no radiation and there is no wind speed then the load demand will be supplied from EU. (S1 =OFF, S2=ON, S3=OFF, S4=OFF, S5=OFF).

Wind Speed

Ra dia tio n

The Artificial Neural Network, (ANN) not only connects PV system or WTG with EU when the DC input voltages lay in the allowed range of the power-conditioning unit (PCU) but also disconnects from EU when the DC input voltages are out of range of the power-conditioning unit. The ANN detects the value of the DC input voltage from PV system, (Vdcpv) or WTG, (Vdcw) and then it sends an ON, 1 or OFF, 0 trip signal to the switches S1, S2, S3, S4 and S5. Step-up Transformer DC/DC

DC/AC

Filter

Step-up Transformer

I.G.

G. B.

Input

DC/AC

AC/DC

S4

S5

Filter

Output

NN for PV/WTG/EU

Bus bar

Step-down Transformer

S3 S1

~ EU

S2 Bus bar

Step-down Transformer

Load

Fig. 1 Layout of PV/WTG interconnected with EU and control strategy The ANN will send an ON-trip signal to switch S4 only if the following condition is realized: 430 < V < 550 dcpv

(6)

Else, the switch state is OFF. On the other hand, the ANN will send an ON-trip signal to switch S5 only if the following condition is realized: 850 < V < 1370 (7) dcw

Else, the switch state is OFF. The operation of the five switches shown in Fig. 1 can be summarized as shown in Table (1). 2-2 Load Characteristics It is assumed here that the load demand varies monthly. This means that each month has daily load curve different from other months. Therefore, there are twelve daily load curves through the year. Fig. 2 shows the load demand for January, April, July and October [6].

Table (1) Operational modes of PV/WTG HEPS Mode

S1

S2

S3

S4

S5

ON

OFF

ON

ON

OFF

ON

ON

OFF

ON

OFF

ON

OFF

ON

OFF

ON

ON

ON

OFF

OFF

ON

ON

OFF

ON

ON

ON

ON

ON

OFF

ON

ON

OFF

ON

OFF

OFF

OFF

1

2

3

4

Generated power vs. Load demand Pgtotal > PL, Ppv(t)>0, PWTG(t)=0 PV DC voltage within limits WTG DC voltage out of limits Pgtotal < PL, Ppv(t)>0, PWTG(t)=0 PV DC voltage within limits WTG DC voltage out of limits Pgtotal > PL, Ppv(t)=0, PWTG(t)>0 PV DC voltage out of limits WTG DC voltage within limits Pgtotal < PL, Ppv(t)=0, PWTG(t)>0 PV DC voltage out of limits WTG DC voltage within limits Pgtotal > PL, Ppv(t)>0, PWTG(t)>0 PV DC voltage within limits WTG DC voltage within limits Pgtotal < PL , Ppv(t)>0, PWTG(t)>0 PV DC voltage within limits WTG DC voltage within limits Pgtotal =0, Ppv(t)=0, PWTG(t)=0 WTG DC voltage out of limits PV DC voltage out of limits

Fig. 2 The hourly load demand for January, April, July and October [6]. 3- Application and Results Figure 1 shows an overview of the power circuit and control circuit of the proposed PV/WTG HEPS interconnected with EU. The power circuit has been controlled by using the proposed three layers neural network as shown in Fig. 3. X1, X2, X3, X4 and t are the Five-input training matrix which represent DC output voltage from PV system, DC output voltage from WTG system, AC voltage of electric utility power, load demand, and time respectively. W(1) and W(2) represents the weight matrices. The network consists of five input layers, ten nodes in hidden layers and five nodes in output layer which sigmoid transfer function. The network has been found after a series of tests and modifications. Fig. 4 shows the DC voltages from WTG. On the other hand, Fig. 5 shows the DC voltage from PV system. The DC voltages are the input to ANN. Note that the DC voltages lay in the allowed range of PCU as shown in the figures 4 and 5. Figure 6 shows the evaluation of the 5+10+5

ANN errors. Table (2) shows weights and biases for 5+10+5 ANN. Figure 7 displays the optimal operation of the PV/WTG HEPS interconnected to EU hour by hour through the day, which represents the months of December.

Fig. 3 Structure of the proposed three layers ANN used to interconnect PV/WTG HEPS.

Fig. 4 DC output voltage from WTG during March, June, September and December

Fig. 5 DC output voltage from PV array during March, June, September and December

Fig. 6 Relation between Error and Epoch 5+10+5 ANN Table (2a) Weights and Biases for 5+10+5 ANN for PV/Wind HEPS Interconnected with EU W1 Bias -0.01 -0.85 -1.61 -2.75 -2.97 -7.14 0.19 -2.09 2.44 -2.08 1.05 -4.07 -0.15 1.24 8.68 -1.72 2.32 5.43 -0.37 -0.70 -0.57 -1.72 -0.75 -1.78 0.25 -0.95 1.09 -1.25 -0.13 -1.05 0.30 -0.85 1.41 -0.20 0.18 2.14 0.13 0.22 -0.44 4.76 2.05 6.58 0.09 -1.60 1.74 -5.66 -3.19 -7.93 0.11 -0.16 -0.39 -1.70 0.08 -1.45 0.33 -5.47 5.45 11.26 12.85 5.82 Table (2b) Weights W2 and Biases for 5+10+5 ANN for PV/Wind Interconnected with EU Weights W2 bias -8.97 -0.47 8.74 -3.26 -3.72 5.19 6.92 -8.26 -2.35 0.02 5.01 0.63 -4.58 0.89 0.38 0.90 -3.35 -3.43 -4.29 1.45 -16.93 -0.48 -2.53 6.29 -0.90 -0.09 -0.87 2.79 1.04 1.68 -1.70 17.01 0.43 -2.65 -1.06 16.51 0.79 1.27 1.62 -3.61 -1.04 -1.58 0.02 -1.5 -0.84 -4.57 -1.78 -2.04 -2.73 4.00 11.69 -10.29 -5.24 2.43 -5.24

Fig. 7 Optimal Operation of the PV/Wind HEPS interconnected to EU to feed the load demand during December

From Fig. 7 it can be seen that the deficit energy has been taken from EU and surplus energy has been injected to EU through the day, which represents the month of December. Figure 8 shows the difference between output from ANN and the desired output for the test data of 120 examples (Five months). These differences are displayed for switches S1, S2, S3, S4 and S5. From this Figure, it can be seen that the ANN of 5+10+5 operates with a high accuracy. Figure 9 displays the output of the proposed ANN of 5+10+5 for month of December using test data. This output may be 1 or 0 for each switch. From Figures 7 and 9 (December) it can be noticed that the trip signal which produced from ANN sent to switch S1 at hours 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 19, 20, 21, 22 and 23. This means that the PV/WTG feed the load demand at these hours. On the other hand, switch S2 (for example) equal to 1 at hours 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 15, 16, 17, 18, 19, 22, 23 and 24 This means that the EU should supply the load demand at these hours. On the other hand, the power injected to EU through switch S3 at hours 1, 2, 3, 13, 20 and 21. From switch S1 and S2 it can be noticed that the hybrid PV/WTG with EU feed the load demand at hours 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 15, 16, 19, 22 and 23. The electric utility feed the load demand without PV/WTG HEPS at hour 24. From switch S4 it can be seen that the PV system feed the load demand at hours 8, 9, 10,11, 12, 13, 14, 15, and 16 which there is no radiation at hours 1, 2, 3, 4, 5, 6, 7, 17, 18, 19, 20, 21, 22, 23 and 24. On the other hand, the WTG feed the load demand at hours 1, 2, 4, 5, 6, 7, 9, 10, 13, 19, 20, 21, 22 and 23. Which there is no wind speed or the DC output voltages not lay within acceptable limits of PCU at hours 8, 11, 12, 14, 15, 16, 17, 18 and 24 as shown in switch S5.

Fig. 8 Relation between outputs and target for five months

Fig. 9 Outputs of Neural Network for month of December

4- Conclusions This paper presents one possible application of intelligent system. For the case here studied, ANN has been applied to the operation of electric PV/WTG HEPS, using the learning capacity of neural networks, applied to the parameters of hybrid systems. The ANN proposed shows the importance of establishing an optimized control, both in terms of the selection of the optimal strategy, and of the relationship between the power generated by the PV system, wind system, EU and load profile. From the results obtained above the following conclusions can be drawn from this paper: • A novel technique based on ANN is proposed to achieve the optimal operation control strategy of PV/WTG HEPS. This ANN operates the PV/WTG HEPS to feed the load demand. • The 5+10+5 ANN is the suitable neural network for optimal operation and control of PV/WTG HEPS at El'Zafarana site. • The ANN has a very high accuracy and achieve the optimal hour by hour operation for PV/WTG HEPS as shown in Figures 8 and 9. • Using this strategy minimizes the lost time of switching ON and switching OFF. Then, the reliability of the whole system will be improved. 5- REFERENCES [1] International Energy agency report " Key Issues in Developing Renewables", 1997. [2] K. Mitchell, J. Rizk and M. Nagrial, “A Simple Predictive Controller for Stand-alone and Gridconnected Renewable Energy Systems”, Australasian Universities Power Engineering Conference (AUPEC) AUPEC2002, Australia, 29th Sept. to 2nd Oct. 2002. [3] O. Omari1, et. al. , "A Simulation Model For Expandable Hybrid Power Systems", 2nd European PV-Hybrid and Mini-Grid Conference, Thursday, 25th Sept. and Friday, 26th Sept. 2003, Kassel, Germany [4] H. H. El-Tamaly and F. M. El-Kady, "Optimizing the Integration of Photovoltaic/Wind System with Electric Utility", Proceeding on Al-Azhar Engineering sixth International Conference, Sept. 1-4-2000, pp. 69-80. [5] E. Koutroulis, et. al. ,"A Hybrid PV-Wind Generator System Using a Maximum Power Point Tracking Technique", International Conference Renewable Energies for Islands Towards 100% RES Supply, Chania-Crete, Greece, 14-16 June 2001. [6] H. H. El-Tamaly and Adel A. El-Baset ,"Design and Control Strategy of Utility Interfaced PV/WTG Hybrid System", The Ninth International Middle East Power System Conference, MEPCON'2003, Minoufiya University, Faculty of Eng., Shebin El-Kom, Egypt, Vol. 2, Dec. 1618, 2003, pp. 699-674. [7] Chihchiang Hua, Jongrong Lin and Chihming Shen, " Implementation of a DSP-controlled photovoltaic system with peak power tracking", IEEE Trans. Industrial electronics, Vol. 45, No. 1, pp.99-107, Feb. 1998. [8] Gary L. Johnson, "Wind energy systems", Book, Prentice-Hall.,1985

Biographies H. H. El-Tamaly was born in Kafer El-sakh-Egypt, on June 15, 1952. He received the B.S. degree in 1975 from Mansoura university-Egypt, M. Sc. Degree in 1980 from Mansoura university-Egypt, Ph. D. Degree in 1984 form Cairo university-Egypt, Professor of electrical power engineering in 1994. Currently, he is working as professor and chairman of electrical engineering Dept.- Faculty of Engineering- Elminia university-Egypt. His major interests are renewable energy system, power systems, power electronics, neural network and Fuzzy system. Adel A. Elbaset Mohamed was born in Nag Hamadi- Egypt, on Oct. 24. 1971. He received the B.S. degree in 1995 from Elminia university-Egypt, M. Sc. Degree in 2000 from Elminia university- Egypt, where he is currently working toward the Ph.D. degree in the department of electrical Engineering, Elminia university. His major interests are renewable energy system, power systems, power electronics, neural network and Fuzzy system.