Paper Study The Operation Of Wind/photovoltaic

Operation and Control Strategy of PV/WTG/EU Hybrid Electric Power System Using Neural Networks

Faculty of Engineering, Elminia University, Elminia, Egypt

Object of this paper This paper introduces an application of an artificial neural network on the operation control of the PV/WTG/EU to improve system efficiency and reliability.

This paper focus on a hybrid system consists of PV/WTG interconnected with utility grid taking into account the variation of solar radiation, Wind speed and load demand during the day. 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 load demand, wind speed, insolation and temperature at El'Zafranna site, Egypt as a case study.

2- System Model 2-1 Modeling of PV/WTG 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). A typical modeling of PV/WTG HEPS, in a gridconnected situation, is shown in the following Figure .

ad ia tio n

S te p -u p T ra n sfo rm e r

R

.App. And Res

Wind Speed

17

D C /D C

D C /A C

F ilte r

S te p -u p T ra n sfo rm e r

I .G .

G . B .

In p u t

D C /A C

A C /D C

S 4

S 5

F ilte r

O u tp u t

B us bar

N N fo r P V /W T G /E U

~ E U

S te p -d o w n T ra n sfo rm e r

S 3 S 1 S 2 B us bar S te p -d o w n T ra n sfo rm e r

L o ad

Fig. 1 Layout of PV/WTG interconnected with EU and control strategy

The /630 power generated by PV system 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

The following operating strategy have employed as follows:

Mode

S1

/730

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

1

2

3

4

OFF

ON

OFF

OFF

OFF

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

The ANN will send an ON-trip signal to switch S4 only if the following condition is realized:

430 < V

dcpv

<

550

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 dcw

Else, the switch state is OFF.

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 daily load curves for January, April, July and October [6]. Fig. 2. The daily load curves for January, April, July and October [6].

Application and Results

Fig.1

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.

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

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.

This Figure shows the DC voltages from WTG /1230

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

This Figure shows the DC voltage from PV system.

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

This Figure shows the evaluation of the 5+10+5 ANN errors.

This Figure sows the optimal Operation of the PV/Wind HEPS interconnected to EU to feed the load demand during December

17 From this 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.

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

Fig. 8 Relation between outputs and target for five months

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.

15 5

Fig. 9 Outputs of Neural Network for month of December

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.

Conclusions This paper presents one possible application of intelligent system. 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: 1. 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.

TheANN 5+10+5 is theaccuracy suitable neural network for 3.2. hasstrategy aANN very high optimal 4.The Using this minimizes theand lostachieve time ofthe switching optimal operation and for control of PV/WTG HEPS at in hour operation PV/WTG HEPS of as the shown ON by andhour switching OFF. Then, the reliability whole El'Zafarana site. Figures andbe 9.improved. system 8will

Thanks for your listening