Professional Documents
Culture Documents
Abstract
In this paper a novel ramp-rate control strategy based on exponential smoothing (ES) method is proposed. Different
from conventional ES method, the proposed method varies the smoothing parameter according to PV ramp-rate.
The proposed smoothing parameter is used to determine the switching status of the battery energy storage (BES). It is
also used in determining the amount of power to be injected or absorbed by the BES in order to smooth the PV output
power. The proposed ramp-rate control strategy is compared with moving average (MA) and conventional exponential
smoothing (CES) methods. It has been found that MA and CES method exhibit memory effect that caused BES to
operate all the time. The proposed ramp-rate control strategy overcomes this limitation by operating the BES only
during significant fluctuation. As a result, the size of the BES capacity can be reduced and increasing its life span.
Highlights.docx [Highlights]
To view all the submission files, including those not included in the PDF, click on the manuscript title on your EVISE
Homepage, then click 'Download zip file'.
1 Ramp-rate control approach based on dynamic smoothing parameter to mitigate solar
2 PV output fluctuations
5 University of Malaya, 50603 Kuala Lumpur, Malaysia. Tel.: +60 3 79675238; fax: +60 3
8 Malaysia
10 Kingdom
11
12 Abstract
13 In this paper a novel ramp-rate control strategy based on exponential smoothing (ES) method
14 is proposed. Different from conventional ES method, the proposed method varies the
16 used to determine the switching status of the battery energy storage (BES). It is also used in
17 determining the amount of power to be injected or absorbed by the BES in order to smooth
18 the PV output power. The proposed ramp-rate control strategy is compared with moving
19 average (MA) and conventional exponential smoothing (CES) methods. It has been found
20 that MA and CES method exhibit memory effect that caused BES to operate all the time. The
21 proposed ramp-rate control strategy overcomes this limitation by operating the BES only
22 during significant fluctuation. As a result, the size of the BES capacity can be reduced and
24 Keywords
27 Output power generated from solar photovoltaic (PV) is variable in nature, due to frequent
28 change in solar radiation level caused by cloud passing. Mitigating solar PV fluctuation is a
29 challenge since solar PV penetration with high ramp-rate introduces significant voltage
30 fluctuation in weak radial distribution network [1]. These negative effects have prompted
31 utilities like Puerto Rico Electric Power Authority (PREPA) to impose ramp rate limit to be
32 10% of rated capacity per minute for both wind and PV generation [2]. There are many ways
33 suggested in the literature such as use of dump load, operate PV below its maximum power
34 point (MPP) and use of storage technology to counter PV output power fluctuation.
35 Use of (i) battery technology, (ii) dump load and (iii) PV generator curtailment to smooth the
36 output power from solar PV plant is presented in [3]. The authors have also examined the
37 economic aspect of using these methods to smooth out PV output power. On analysing 10
38 min radiation data, it has been found that combining battery technology and generator
39 curtailment is the most economical solution. Smoothing PV output power using MPPT
40 control is presented in [4]. MPPT controls the PV output power ramp to 1% of PV capacity
41 per minute when solar radiation increases. This method is not effective when the solar
42 radiation decreases rapidly. Use of energy storage technologies such as battery energy storage
43 (BES), electric double layer capacitor (EDLC), superconducting magnetic energy storage
44 (SMES) and fuel cell have been proposed to smooth out short term solar PV output power
46 In [5], PV output power fluctuation is smoothed out effectively by dispatching EDLC based
47 on the reference value generated by moving average (MA) method. In [6] a modified euler
48 type moving average method is proposed to mitigate PV power fluctuation using EDLC. A
49 PV output power smoothing algorithm based on average irradiation level is presented in [9].
50 A smoothing algorithm using fuel cell was designed to smooth the PV output power in [10].
51 A fuzzy based wavelet transform smoothing filter is used in [11] to smooth PV and wind
53 wherein the BES is used to smooth the PV output power. Application of low pass filter to
54 smooth the PV output power fluctuations using BES can be found in [13]. In [14] a power
55 management module is developed where power fluctuation from PV is taken care by battery
56 source and the power supply to the load is provided by diesel generator and battery. Here
57 moving average smoothing method is used to smooth the fluctuation caused by 5-mins solar
58 radiation. In [15] conventional exponential smoothing (CES) method is used to suppress solar
59 PV fluctuation using proton exchange membrane (PEM) fuel cell and electrolyzer. In [16] the
60 ramp rate from solar PV is reduced by applying a high pass filter where battery and electric
61 vehicle charging facility were used to mitigate short term PV variability. In [17] BES is used
62 to smooth the output power fluctuation from hybrid system consisting of solar PV and wind
63 turbine. The smoothing functions dynamic filtering controller and dynamic rate limiter
64 proposed in [17] is based on power fluctuation rate from hybrid system. Application of
65 moving average filter and low pass filter to smooth out PV output power fluctuation is
66 presented in [18]. Use of battery energy and natural gas engine generator to smooth the PV
67 output power fluctuation is demonstrated in [19]. The smoothed output is produced using MA
68 method. Simulation results conclude that using gas engine generator along side of battery to
69 smooth PV output power reduces burden of the battery and increases its lifetime. Natural gas
70 engine generator is not fast enough to mitigate higher ramp-rates. Furthermore, if battery is
71 not available, the gas engine generator would not control the ramp-rates. The minimum
73 used to follow the smoothed reference waveform produced by moving average method to
74 smooth the PV output from 1.2MW PV plant [21]. It was recorded that the PV output power
75 varies 63% of rated capacity per minute. The PV penetration into the distribution system is
76 limited to 50% of PV capacity to counter any negative impact of high ramp-rate in
77 distribution system. On analysing the voltage and frequency parameters it was found that the
78 variation in PV does not affect the grid performance. Two ramp-rate control strategies, one
79 strategy based on PV inverter control and other strategy based on battery state of charge
80 (SOC) control and actual power from PV plant is presented in [22]. The advantage of using
81 MA method is that it requires smallest capacity of energy storage. On the other hand it will
83 From the discussion on previous works, it was found that MA and CES methods are
85 These methods give extra importance to the past history data than the present value of the
86 fluctuating value. This problem is referred as memory effect as described in [24]. The
87 detailed explanation of memory effect in MA and CES methods is described in Section II.
88 The memory effect causes the BES to operate even though there are no significant PV ramp-
89 ups/downs. As the results of memory effect when MA or CES methods are applied, the BES
90 is forced to operate all the time decreasing its life span. In addition, the memory effect also
91 causes over smoothing of ramps. The term over smooth is referred as smoothing below the
92 desirable level of ramp rate. As a result of over smoothing, the BES charges/discharges more
93 energy resulting in reduction of ramp-rate below the desirable level which will eventually
95 Based on the above disadvantage, this paper aims to design a ramp-rate control strategy
96 which limits the PV ramp-rate within desirable level and also eliminate the memory effect.
97 The original contributions of this study are: (i) the root cause of memory effect in MA and
98 CES method is analysed based on its formulation. A detailed study on weights associated to
99 the PV data points for MA and CES methods are conducted, (ii) a sensitivity analysis on MA
100 and CES methods are carried out to analyse the variation of memory effect, ramp-rate control
101 ability and size of battery energy storage required, (iii) a ramp-rate control strategy is
102 proposed which restricts the PV ramp-rate within desirable level thus eliminating the memory
103 effect, (iv) smoothing parameter which is used in proposed ramp-rate strategy has been
104 formulated as a function of ramp-rate. A switching function is also introduced where the BES
105 is switched ON/OFF appropriately only to control PV output ramp ups/downs and (v) the
106 efficiency of the proposed ramp-rate control strategy is compared with MA method and CES
107 method for ramp-up/down events. From the comparison performed it has been found that the
108 proposed ramp-rate control approach is effective in minimising the ramp-rate to desirable
109 level for ramp-rate violation events. In addition, the proposed approach will not allow the
110 BES to operate for ramp-rates which are already within the limit. As a result, BES does not
111 operate all the time thereby increasing its life span. Moreover, larger BES capacity is not
114 In this section MA and CES methods with their limitations are presented in detail. MA and
115 CES methods are applied to smooth PV output power shown in Fig.1. Output power from
116 solar PV represents time series data, where the data points consist of successive measurement
117 of PV power over a time interval (see Fig.1). Average hourly solar radiation for a day in May
118 for a typical location in Malaysia is extracted using HOMER software. Twenty four hours
119 radiation data have been interpolated to obtain 1440 values by adding normal distribution
120 random noise with 0 mean and standard deviation as 1 [14]. The solar radiation is then
0.2
Power in MW
0.15
0.1
0.05
0
0 4 8 12 16 20 24
122 Time in hours
123 Fig.1 Solar power output power obtained from 1-min radiation data
124
125 2.1 n-Period Moving Average (MA) method
126 MA is essentially centered on finding out average for the set of input and the assumption is
127 attributed that it represents a constant or level model. Moving averages is made by assuming
128 that the latest n period is more relevant and others are left out. The principle of operation of
129 MA for n relevant data points for the input having k data points is given as,
130 F
i(k ) i(k 1) ... i(k n) (1)
n
131 where n is number of relevant data points used in MA smoothing, i and F are input (PPV) and
132 smoothed output respectively. For this work 41 data points is used as the moving average
133 window therefore the value of n=41. Elaborating (1) for 41-point MA is given as,
135 As mentioned above for a 41-period moving average, 41 PPV data points which are most
136 relevant for producing smoothed output is used and other data points are left out. A 41-period
137 moving average can be thought of giving weights 1 for the relevant data points considered
138 for producing smoothed output and 0 weights for the points left out.
139 MA is able to capture the principle that the most recent data points are important and older
140 points do not contribute for the smoothing. From the above explanation we would like to
141 highlight two main issues that are associated with MA which causes memory effect,
143 (ii) The weights associated with relevant data points are equal which is not logical or
144 acceptable.
145 The limitation of (MA) method is demonstrated using Fig.2. For illustration purpose the 41-
146 point MA smoothed waveform during the time period 13.5 hours and 18.5 hours is presented
147 in Fig.2 and it is obtained by smoothing the actual PV output power shown in Fig.1. The
148 actual solar PV power obtained from 1-min radiation data is compared with PV power
149 smoothed by MA method. Noticing portions a,b,c,d,e and f in Fig.2, there is no significant
150 fluctuation in PV for which the battery supplied or absorbed significant amount of power
0.3
Power smoothed by 41 point MA
0.25 Actual PV power
0.2 c
Power in MW
b
0.15 a d
0.1 e
0.05
f
0
14 15 16 17 18
152 Time in hours
153 Fig.2. Actual PV power and smoothed power using 41 point MA waveform
155 CES method actually considers all the points in the system and gives progressively increasing
156 weights to more recent data. The fundamental equation for CES is given as,
159 output power from solar PV and is the smoothing constant which should be selected
161
PPV (i ) PPV (i ) (1 ) PPV (i 1) (1 ) PPV (i 2) (4)
165 where n is the number of PV data points in the time window. In (5), PPV (i (n 1)) denotes
166 the forecast for the point n-1 for which the value is zero. Therefore (5) can be written as,
169 From (6) it is clear that the value of , (i-), (i- )n is treated as the weights associated
170 with the corresponding PV data points and is progressively decreasing to older data points.
171 If the value of is greater than 0.5 or closer to 1 the contribution of the present value or the
172 most recent value is weighed more and contribution of older terms is less for smoothing. If
173 the value of is smaller than 0.5 or closer to 0, present value or the most recent terms will
174 be weighed less and the older terms start contributing to the smoothing. Therefore, smoothing
175 constant is the main parameter which helps to smooth PV fluctuations. Choosing the
176 between 0 and 0.5 will over smooth PV output power and if the value is more than 0.5 or
178 There is no absolute reason presented in the literatures on choice of smoothing parameter
179 pertaining to this PV ramp-rate control problem. The major drawback of using CES method
180 is, the arbitrary choice of smoothing constant associated with the data points. For
181 comparison with the proposed control strategy the value of is chosen as 0.2 which is
182 predominantly found in the literature. For illustration purpose the smoothed waveform
183 produced by CES method during the time period 13.5h and 18.5h is presented in Fig.3. The
184 portions encircled in dotted line shows that there is no significant fluctuation in PV output
185 power, for which the battery will supply/absorb significant power due to memory effect. In
186 other words the weights , (i-), (i- )n distributed among the data points are not
187 logically related to PV ramp-rate, therefore, creating memory effect. For this reason the
188 BES is allowed to operate even though there is not significant fluctuation.
0.25
Actual PV
0.2 Smoothed PV
Power in MW
0.15
0.1
0.05
0
14 15 16 17 18
189 Time in hours
190 Fig.3. Actual PV output and PV power smoothed using CES method
191 The difference between CES and MA method is, CES method considers all the PV data
192 points in the system and gives progressively increasing weights to more recent data. On the
193 other hand MA method consider only relevant PV data points and give equal weights to it and
194 older data points are left out. Due to this memory effect, the CES and MA method will
195 invariably smooth the PV output power even though there is no significant fluctuation.
197 In this section, sensitivity analysis on MA and CES methods are carried out by using different
198 window size and smoothing constant values respectively for the data shown in Fig. 1. From
199 this analysis, the needs of new smoothing method can be justified.
200
201 3.1 Analysis on MA method
203 This analysis is conducted to show the variation of memory effect for different MA window
204 sizes. In this analysis 5-point, 11-point, 31-point, and 51-points MA window are applied. The
205 PV output power smoothed by using MA method for these window sizes is plotted in Fig.4.
206 In this figure, only time periods between 16.5 hours to 19 hours is shown due to the presence
207 of both significant and non-significant fluctuations in actual PV power. The corresponding
208 battery power utilization for MA with different windows is plotted in Fig.5.
209 From Fig.4, it is noticed that the actual PV output power can be smoothed using any window
210 size. However, the smoothing effect increases as the MA window size increases. The
211 difference between the smoothed waveform and actual PV power is the power required by the
212 battery source to supply or absorb in order to limit the fluctuations (plotted in Fig.5).When
213 the 5-point MA method is applied, the energy supplied by the BES to smooth the fluctuations
214 is 5.4kWh for the said time period. On the other hand, the energy supplied by BES using 51-
216 Thus, it can be concluded that memory effect varies from smaller to higher level as the value
217 of MA window increases. In addition, the variation of memory effect also has an influence in
219
0.18
Actual PV power
0.16 smooth 5point MA
smooth 11point MA
0.14 smooth 31point MA
smooth 51point MA
0.12
Power in MW
0.1
0.08
0.06
0.04
0.02
0
16.5 17 17.5 18 18.5 19
Time in hours
220
221 Fig.4. Smoothed PV output power produced using MA method for different windows
0.08
discharge/supply Pbat 5pt MA
Pbat 11pt MA
0.06 Pbat 31pt MA
Pbat 51pt MA
0.04
Power in MW
0.02
-0.02
-0.04
charging/absorb
-0.06
16.5 17 17.5 18 18.5 19
Time in hours
222
223 Fig.5. Battery utilization for smoothing PV output power using MA method for different
224 windows
226 In this analysis the ability of MA method with different window sizes to control the ramp-rate
227 to desirable limit is analyzed. The ramp-rate limit is set to 10% of PV rated capacity, which is
228 20kW/minute.
229 From the analysis it has been found that MA method with different window sizes have the
230 ability to control the PV ramp-rates within the limit. During the entire PV operation it has
231 been noticed that the extent of control of ramp-rate decreases as the MA window size
232 increases. For example when 5 point MA is applied the ramp-rate is controlled within
233 16kW/minute. On the other hand when 51-point MA is used, the ramp-rate is controlled
235 The ramp-rate control ability of MA with different window values is plotted in Fig.6 for time
40
Actual ramp-rate Actual ramp-rate
30 already within limit ramp-rate control 5ptMA
ramp-rate control 11pt MA
+ ramp limit ramp-rate control 31 pt MA
20
Ramp rate in (kW/minute)
ramp-rate control 51 pt MA
10
-10
-50
16.5 17 17.5 18 18.5 19
Time in hours
237
238 Fig.6. Ramp-rate control ability for MA method with different window values
239 There are instances where the actual PV ramp-rate is already within the limit (encircled in
240 Fig.6) for which the MA methods allow the battery storage to operate. The BES operation
241 during these instances is evident in Fig.5. This unnecessary operation will contribute to
243 Over smoothing of ramp-rate is also evident when the MA method is applied. For example, in
245 ramp-rate can be noticed. The ramp-rate violation should be controlled to the desirable level
246 (that is 20kW/minute). When 5 point, 11 point, 31 point and 51 point MA methods are
247 applied, the ramp-rate at that particular instance are over smoothed to 9kW/minute, -
248 2kW/minute, -0.75kW/minute, and -0.8kW/minute respectively, which are unnecessary. Here
249 the ramp-rate is controlled below the desirable level for which excessive battery power is
253 In this section an analysis on variation of memory effect for CES method with different
254 smoothing constant () is analyzed. The applied smoothing constants are 0.1, 0.5, and 0.8.
255 The results of PV output power smoothed by using CES method for these smoothing constant
256 values is plotted in Fig.7 for time periods of 16.5 hours to 19 hours. The corresponding
0.18
Actual PV power
0.16 CES smooth alpha=0.1
CES smooth alpha=0.5
0.14
CES smooth alpha=0.8
0.12
Power in MW
0.1
0.08
0.06
0.04
0.02
0
16.5 17 17.5 18 18.5 19
Time in hours
258
259 Fig.7. Smoothed PV output power produced using CES method for different
0.05
discharge/supply Pbat CES alpha 0.1
0.04 Pbat CES alpha 0.5
Pbat CES alpha 0.8
0.03
0.02
Power in MW
0.01
-0.01
-0.02
-0.03
charge/absorb
-0.04
16.5 17 17.5 18 18.5 19
Time in hours
260
261 Fig.8. Battery utilization for smoothing PV output power using CES method for different
262 From Fig.7, it can be noticed that the actual PV output power is smoothed for any value of
263 . However, the smoothing effect increases as the value of decreases from 1. The
264 difference between the smoothed waveform and actual PV power is the power required by the
265 battery source to supply or absorb in order to limit the fluctuations (plotted in Fig.8). When
266 the CES method with =0.8 is applied, the energy supplied by the BES to smooth the
267 fluctuations is 0.426kWh for the said time period. On the other hand when CES method with
269 Thus, it is evident that the memory effect varies from smaller to higher level as the values of
270 progressively decreases from 1. Moreover, the variation of memory effect also has an
273 In this analysis the ability of CES method with different smoothing constants to control the
274 ramp-rate to desirable limit is analyzed. During the entire time of PV operation it was noticed
275 that the extent of control of ramp-rate increases as the value of smoothing constant
276 decreases when CES method is applied. For example when CES method with =0.8 is
277 applied the ramp-rate is controlled within 40kW/minute. Similarly when CES method with
280 plotted in Fig.9 for time period 16.5 hours to 19 hours. It was found that CES method is
281 unable to limit the ramp-rate to the desirable limit for many instances when the smoothing
282 constant values are 0.5 and 0.8. For example, this inability is evident (encircled in Fig.9)
283 when =0.5 and 0.8. There are instances where the actual PV ramp-rate is already within
284 the limit (encircled in Fig.9), but CES method allows the BES to operate.
285 Over smoothing of ramp-rate is also evident when applying CES method. For example, in
286 portion b (encircled in Fig.9), the actual ramp-rate violation of -41.34kW/minute can be
287 noticed. The CES method with =0.1 caused over smoothed of ramp rate to -3.4kW/minute.
288 Here the ramp-rate is controlled below the desirable level for which excessive battery power
289 is utilized therefore resulting in large BES size. On the other hand, CES methods with =0.5
290 and 0.8 is unable to control the ramp-rate within the limit. For CES method with =0.5 the
291 ramp-rate is controlled till -22.7 kW/minute and for CES method with =0.8 case the ramp-
40
Actual PV ramp-rate
Actual ramp-rate
30 Ramp-rate CES alpha 0.1
already within limit
Ramp-rate CES alpha 0.5
+ramp limit Ramp-rate CES alpha 0.8
20
Ramp-rate in kW per minute
10
-10
-ramp limit
-20
294 Fig.9. Ramp-rate control ability for CES method with different smoothing constant values
295
296
297 4. Proposed Ramp-Rate Control Strategy
298 MA and CES method allows excessive use of BES which will reduce its operating life.
299 Moreover, this excessive usage will result in larger BES capacity when MA or CES methods
300 are applied. The limitations of MA and CES methods are addressed by the proposed ramp-
302 The proposed ramp-rate control strategy uses exponential smoothing method with
303 modifications. Therefore the fundamental equation of exponential smoothing method with the
306 where PPV (i ) is the smoothed PV output for the instant i, PPV (i 1) is the smoothed PV
307 output for the instant i-1 and PPV(i) is the actual PV output power for the instant i.
308 Unlike the smoothing parameter in (3), the proposed smoothing parameter in (7)
309 change for every time step depending on the ramp-rate violations. Therefore, the ramp-rate
310 has to be checked for any violation for every time step. Based on the level of violation the
311 smoothing parameter is calculated. The ramp rate of solar PV output power at instant i is
313
dPPV P (i) PPV (i 1)
(i ) PV (8)
dt t (i ) t (i 1)
314 where PPV(i) and PPV(i-1) are actual powers at ith and (i-1)th instant.
315 Likewise the difference between actual PV power at instant i(PPV(i)) and smoothed PV
316 power at instant i-1( PPV (i 1) ) also denotes the ramp-rate and is expressed as,
317 dPPV
P (i ) PPV (i 1)
(i ) PV
(9)
dt t (i ) t (i 1)
318 For this work 10% of solar PVs rated capacity per minute is taken as ramp-rate limit. The
321 is basically used to limit the PV ramp-rate. The f(RR) is illustrated as follows,
323 The proposed ramping function f(RR) is added with smoothed power at previous instant i-1
324 ( PPV (i 1) ) to obtain the smoothed power at instant i PPV (i ) and is given in (11).
326 By equating the (7) and (11) the proposed smoothing parameter () is given as,
327 f RR (12)
PPV (i ) PPV (i 1)
328 The calculated smoothing parameter () is substituted in (7) to find out smoothed output PPV
329 for every time instant. A discrete switching function S(i), based on the smoothing parameter
0, for 1
331 S (i ) (13)
1, for 1
332 By introducing the switching function S(i) in (7), the (7) is updated as follows,
333
PPV (i ) PPV (i ) (1 ) PPV (i 1) S (i ) (14)
334 Equation (14) is the final control equation from which the smoothed PV output power ( PPV (i )
336 The difference between the smoothed power ( PPV ) and the actual PV output power (PPV)
337 gives the target power required by the battery (PSET) to absorb or inject in order to control the
340 If PSET is positive the battery will discharge, and if negative the battery will operate in
341 charging mode. A simple illustration for generation of target power (PSET) for the battery is
PPV(i)
400V AC bus
Solar
PV
PPV(i)
PSET(i) PBES(i)
BES
-+
PPV (i )
Proposed strategy
Calculate eq.(8) and eq.(9)
Check for violation by eq.(10)
Calculate using eq.(12)
Decide battery switching eq.(13)
Calculate smoothed output P PV ( i ) using eq.(14)
343
344 Fig.10. Calculating PSET for battery
345 The illustration of the proposed ramp-rate control strategy is provided in Fig.11. It should be
346 noted that the time interval between subsequent points in Fig.11 and Fig.12 is 1 minute.
347 During the time period t1 to t4, the ramp-rates are calculated using (8) and (9). The calculated
348 ramp-rates are checked for any violation using (10) and the appropriate ramping function
349 f(RR) is calculated. Since there is no violation in ramp-rates the smoothing parameter
350 calculated using (12) is 1 (Fig.12) for which the battery is also switched OFF. Therefore the
351 values of smoothed PV power ( PPV ) and actual PV output (PPV) are same for the time period
352 t1 to t4. From t4 to t5 there is a significant fluctuation in PV output power (PPV). Therefore
353 the ramp-rates are calculated and checked for violation. Using the ramping function f(RR) the
354 appropriate value of smoothing parameter is calculated from which the battery switching
355 is also decided. Finally, the smoothed output PPV is calculated from (14) and since actual PV
356 power (PPV) is more than the smoothed power ( PPV )the battery has to be operated in
357 charging mode in order to limit the ramp-rate. There is no significant power fluctuation in
358 actual PV power from t5 to t6 but dPPV exceeds the ramp limit. Therefore the ramp-rate for
dt
359 the smoothed waveform is limited and the smoothed reference to time instant t6 is calculated
360 ( PPV (t 6) ). During this time period the battery is charged. Between t6 to t7 the ramp-rates are
361 within the limit therefore PPV (t 7) is equal to PPV(t7). Therefore the smoothing parameter
362 for the time period t7 reaches 1 for which the battery is switched OFF.
363
365 Similarly the ramp-down is controlled by discharging the battery between the period t7 to t10
366 where the variation in the smoothing parameter during this period is presented in Fig.12.
367
368 Fig.12. Variation in smoothing parameter for the proposed ramp-rate strategy
369 From t10 to t14 there is no significant fluctuation in PV output (PPV), therefore the smoothed
370 output ( PPV ) and actual PV output (PPV) remains the same. During this time period the value
371 of smoothing parameter is also 1 and the battery is also switched OFF. It is clear from
372 Fig.11 that battery is used only to smooth out the ramp ups and downs. The proposed ramp-
373 rate control strategy is applied to a 200kW PV plant where the ramp-rate is controlled by a
376 The test system consists of grid connected PV and BES as shown in Fig.13. The test system
377 is modelled and the simulation is carried out using PSCAD software. The specifications of
378 the test system are given in Table 1. A single diode model of solar PV cell is implemented to
379 build the PV plant. The PV plant is operated in unity power factor. In this work single stage
380 controller is utilised to control the voltage source converter (VSC) of PV plant. In the single
381 stage controller mechanism, the power from the DC side is transferred to AC side by
382 regulating the voltage at DC link. The error between the square of the reference voltage set
383 by MPPT and square of the measured DC bus voltage is sent to PI controller. The output
384 from the PI controller is added to instantaneous power from PV array to obtain the reference
385 power. This reference power is then sent to inner power control loop.
386 The BES is modeled by incorporating the effect of internal resistance and Peukert equation in
387 order to obtain battery voltage and SOC [14]. The charge level in the battery is determined
388 from the battery voltage level. The charge controllers were used to protect the battery from
389 deep discharging and over charging. Therefore the voltage level is continuously monitored
390 using charge controllers to prevent the battery from deep discharging. Bulk or current
391 charging and constant voltage charging schemes were used to protect the battery from over
392 charging. Current control mode is incorporated to control the VSC of BES plant. The inputs
393 to the inner current control loop are the dq reference current vectors which are produced from
394 output power control loop. The maximum charging and discharging limit for the BES is set at
BKR Grid
0.400/11kV
1 MVA
0.4kV bus
BKR load BKR PV BKR BES
L
o Photovoltaic BES
a array 200kWh
d 200kW
396
401 In this section the overall results on application of the proposed ramp-rate control strategy to
402 control the ramp-rate from solar PV within the desirable level is presented. The proposed
403 ramp-rate strategy is compared with 41-point MA and CES method. The aim is to limit the
404 ramp-rate of PV output power by 10% of PV rated capacity which is 20kW/minute. In other
405 words, any ramps from the PV output power within this limit is allowable and above the limit
407 A 200kWh lead-acid BES is used in the simulation. The BES is permitted to operate between
408 30% to 100% state of charge (SOC). At initial condition the SOC of the battery is set to 70%
409 thereby allowing it to charge/discharge. The overall results for mitigation of PV output power
410 fluctuation using MA, CES with =0.2 and proposed ramp-rate control strategy are shown in
413 PV output power. However the proposed ramp-rate control strategy has the ability to control
414 the ramp-rates to the desirable level which is highlighted from the results presented in next
415 section.
PPV
0.25 PV smoothed using MA
PBES
0.2
Power in MW
0.15
0.1
0.05
-0.05
0 4 8 12 16 20 24
416 Time in hours
0.3
PPV
0.25 PV smoothed using CES
PBES
0.2
Power in MW
0.15
0.1
0.05
-0.05
0 4 8 12 16 20 24
418 Time in hours
0.3
PPV
0.25 PV smoothed using proposed strategy
PBES
0.2
Powerin MW
0.15
0.1
0.05
0
0 4 8 12 16 20 24
420 Time in hours
423 In this section the results on ability of the proposed method to control the ramp-rate to
424 desirable level are discussed. Fig.17 presents actual PV power with smoothed power using
425 MA, CES and the proposed method for the time period 17 hours to 19 hours. The ramp-rate
426 for this time period for the three methods is presented in Fig.18. The figure shows the ability
427 of the methods to control the ramp rate to the desirable level, which is 20kW/minute.
428 Corresponding BES utilization for the three methods are shown in Fig.19. In order to
429 highlight the effectiveness of the proposed method, three portions are selected to present
0.3
Actual PPV
PV smoothed using MA
0.25
PV smoothed using CES
PV smoothing using proposed approach
0.2 Ramp-rate violation event
Power in MW
0.15
e
0.1
c
0.05
d
433 Fig.17. Smoothing PV output power for specific hours using MA, CES, and proposed ramp-
-20
-ramp limit
-30
e
-40
Ramp-rate violation event
-50
17 17.2 17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19
Time in hours
435
0.08
Battery usage for ramp-
rate within the limit for PBES-MA
MA and CES methods. PBES-CES
0.06
PBES=0, for proposed PBES-proposed approach
approach
e
0.04 d
Power in MW
c
0.02
-0.02
Battery utilization during
ramp-rate violation events
-0.04
17 17.2 17.4 17.6 17.8 18 18.2 18.4 18.6 18.8 19
Time in hours
437
439 In Fig.18, during the time periods c and d the actual ramp-rate is found to be within the
440 limit. However, when MA and CES methods are applied the ramp-rate is further reduced. As
441 a result in Fig. 19, it can be seen that the BES is operated during the time duration of c and
442 d when both methods are applied. These non-significant fluctuating events supposedly
443 shouldnt be considered for battery operation since the ramp-rate is within the limit. The
444 continuous BES operation will affect the BESs life span in long run. Deep discharging and
445 frequent charging decreases battery life [25]. On the other hand, when the proposed method is
446 applied, the output power from the BES is maintained zero or switched OFF for ramp-rates
447 which are within the limit. This is evident from Fig.19 for the c and d.
448 An incident of ramp-rate violation in actual PV power is shown in portion e of Fig.17 and
449 Fig.18 at 18.01 hours and is recorded as -41.34kW/minute. The BES has to discharge power
450 in order to control this negative ramp-rate violation. When MA and CES methods are applied
452 a result the BES has to discharge about 34.3kW and 33.04kW when MA and CES methods
453 are applied respectively. On the other hand when the proposed method is applied the BES
454 discharges only 21.34kW in order to limit the ramp-rate exactly to the desirable level (-
455 20kW/minute). This can be seen in portion e in Fig.19. The BES will supply or absorb
456 excessive power when MA or CES method is applied. On the other hand, the proposed
457 method causes the BES to charge or discharge exactly the required power to ensure the ramp-
459 The ramp-rate profile of the smoothed PV output for the proposed method for entire duration
460 of PV operation is presented in Fig.20. From the figure it is clear that the proposed method
0.05
Ramp-rate of actual PV
0.04
Ramp-rate of proposed strategy
0.03
0.02
+ve Ramp limit
Power in MW
0.01
0
-0.01
-0.02
-ve Ramp limit
-0.03
-0.04
-0.05
0 4 8 12 16 20 24
462 Time in hours
465 Variation of BES SOC level for all the three methods are shown in Fig.21. It can be noticed
466 that the variation in SOC is large when MA and CES methods are applied. This variation in
467 SOC level is due to, (i) unnecessary operation of BES even though the ramp-rates are within
468 the limits and (ii) excessive charge or discharge of BES power to over smooth the ramp-rates.
469 As a result large capacity of BES is required when MA and CES methods are applied, which
470 will eventually increase its capital cost. On the other hand, when the proposed method is
471 applied the variation of BESs SOC level is less. This is because the proposed method will
472 allow the BES to operate only during ramp-rate violation events and for remaining hours the
473 power from the BES is zero or switched OFF. From the analysis of results shown in Fig.21,
474 it was found the total amount of energy needed to control the ramp-rate using the proposed
475 method is around 8.13kWh. This value is much lower than the size of the battery in the
476 system, which is 200kWh. Based on this finding, the proposed method can also used to find
477 suitable battery size for the system in order to avoid battery oversize.
1
SOC CES
SOC MA
0.95 SOC proposed
0.9
Battery SOC in %
0.85
0.8
0.75
0.7
0.65
0 2 4 6 8 10 12 14 16 18 20 22 24
Time in hours
478
479 Fig.21. Variation of BES SOC level for the three methods
480 Fig.22 gives clear picture of variation of smoothing parameter for entire time of operation.
481 It can be observed that the value of smoothing parameter is 1 when the ramp-rate is within
482 the limits. The value of smoothing parameter will be less than 1 for ramp-rate violations.
Variation in smoothing parameter (sigma)
1
0.8
0.6
0.4
0.2
0 4 8 12 16 20 24
483 Time in hours
485 6. Discussion
486 Table 2 shows a comparison between the three methods for the number of instances BES is
487 switched ON. It was noticed that there are 85 instances of ramp-rate violation events. Using
488 the proposed ramp-rate control strategy, the BES is switched ON only during those instances.
489 The smoothing parameter calculated using (12) will be less than 1 for those instances.
490 During remaining instances, the BES is switched OFF since the PV ramp-rate is within the
491 limit. Here the value of smoothing parameter is 1. On the other hand, when MA or CES
492 method is used to smooth PV output power fluctuation the BES is switched ON all the time.
493 Table 2. Comparison on number of instances BES is switched OB for three methods
495 From Table 2, it is evident that when the proposed ramp-rate control strategy is used the
496 number of charging and discharging instances of BES is very less because the battery source
498 The proposed ramp-rate control method can be applied for both planning and real-time
499 application. For planning, it can be used to design the microgrid system, mainly on the
500 estimating appropriate battery size. For real time application, real time data (irradiance) or
501 short time forecasting of the PV output (based on predicted irradiance) can be used by the
502 proposed method to determine the smoothing of PV output. In general, the level of accuracy
503 of the forecasting will influence the proposed method outcomes for planning and real time
504 applications.
505 The following are the advantages of the proposed ramp-rate control strategy over MA and
507 (i) The proposed ramp-rate control strategy considers all data points present in the system.
508 (ii) The weights associated with the data points are not equal.
509 (iii) The choice of smoothing parameter associated with the data points for the proposed
511 (iv) Since the weights distributed in PV data points are based on PV ramp-rate, the
513 (v) The proposed ramp-rate control strategy does not require BES to operate all the time
514 therefore reducing the size of BES capacity considerably and also contribute to increase in
516 7. Conclusion
517 A new method to control the PV output power ramp-rate to desirable level using BES is
518 proposed in this paper. The proposed method has been compared with MA and CES methods.
519 From the analysis it was found that the distribution of weights associated with PV data points
520 in MA and CES methods is not based on PV ramp-rates, thus creating memory effect. This
521 causes switching of BES all the time to smooth the PV output power. On the other hand, the
522 proposed strategy does not exhibit memory effect as the weights associated with the PV data
523 points are based on PV ramp-rates. The proposed control strategy allows the BES to operate
524 only during ramp violation events for which the BES optimally charges/discharges to control
525 the ramp-rate to the desirable level. The test results show that the proposed method caused
526 the BES to switch ON for 85 minutes, which constitutes to only 10.9% of BES operation as
527 compared to MA or CES methods. With the less switching of BES using the proposed
528 method, the life span of BES can be increased. Moreover the proposed approach does not
529 allow the BES to over smooth the ramp-rate. Therefore, when the proposed method is
530 applied, the size of BES capacity required for PV ramp-rate control application can be
531 reduced.
532 References
533 [1] Sugihara, H., Yokoyama, K., Saeki, O., Tsuji, K., Funaki, T. Economic and efficient
534 voltage management using customer-owned energy storage systems in a distribution
535 network with high penetration of photovoltaic systems, IEEE Transactions on Power
536 System, 28 (2013) 102-111.
537 [2] Gevorgian, V., Booth, S., Review of Prepa Technical Requirements for Interconnecting
538 Wind and Solar Generation, National Renewable Energy Laboratory: Golden, CO, U.S.
539 Department of Energy, USA, 2013.
540 [3] Omran, W. A., Kazerani, M., Salama, M. M. A. Investigation of methods for reduction of
541 power fluctuations generated from large grid-connected photovoltaic systems, IEEE
542 Transactions on Energy Conversion, 26 (2011) 318-327.
543 [4] Ina, N., Yanagawa, S., Kato, T., Suzuoki, Y. Smoothing PV System's Output by Tuning
544 MPPT Control, IEEJ Transaction on Power and Energy, 124 (2004) 455-461.
545 [5] Kakimoto, N., Satoh, H., Takayama, S., Nakamura, K. Ramp-rate control of photovoltaic
546 generator with electric double-layer capacitor, IEEE Transactions on Energy Conversion,
547 24 (2009) 465-473.
548 [6] Monai, T., Takano, I., Nishikawa, H., Sawada, Y. A collaborative operation method
549 between new energy-type dispersed power supply and EDLC, IEEE Transactions on
550 Energy Conversion, 19 (2004) 590-598.
551 [7] Tam, K. S., Kumar, P., Foreman, M. Enhancing the utilization of photovoltaic power
552 generation by superconductive magnetic energy storage, IEEE Transactions on Energy
553 Conversion, 4 (1989) 314-321.
554 [8] Rahman, S., Tam, K.S. A feasibility study of photovoltaic-fuel cell hybrid energy system,
555 IEEE Transactions on Energy Conversion, 3 (1988) 50-55.
556 [9] Hund, T.D., Gonzalez, S., Barrett, K. Grid-Tied PV system energy smoothing, in 35th
557 IEEE Photovoltaic Specialists Conference (PVSC), (2010) 2762-2766.
558 [10] Darras, C., Muselli, M., Poggi, P., Voyant, C., Hoguet, J. C., Montignac, F. PV output
559 power fluctuations smoothing: The MYRTE platform experience, International Journal
560 of Hydrogen Energy, 37 (2012) 14015-14025.
561 [11]Li, X., Li, Y., Han, X., Hui, D. Application of fuzzy wavelet transform to smooth
562 wind/PV hybrid power system output with battery energy storage system, Energy
563 Procedia, 12 (2011) 994-1001.
564 [12]Saez-de-Ibarra, A., Martinez-Laserna, E., Stroe, D. I., Swierczynski, M., Rodriguez, P.
565 Sizing Study of Second Life Li-ion Batteries for Enhancing Renewable Energy Grid
566 Integration, IEEE Transactions on Industrial Applications, 52 (2016) 4999-5008.
567 [13]Datta, M., Senjyu, T., Yona, A., Funabashi, T., Kim, C. H. Photovoltaic Output Power
568 Fluctuations Smoothing Methods for Single and Multiple PV Generators, Current
569 Applied Physics, 10 (2010) S265-S270.
570 [14] Koohi-Kamali, S., Rahim, N. A., Mokhlis, H. Smart power management algorithm in
571 microgrid consisting of photovoltaic, diesel, and battery storage plants considering
572 variations in sunlight, temperature, and load, Energy Conversion and Management 84
573 (2014) 562-582.
574 [15] Tesfahunegn, S. G., Ulleberg, ., Vie, P. J., Undeland, T. M. PV Fluctuation Balancing
575 Using Hydrogen Storagea Smoothing Method for Integration of PV Generation into the
576 Utility Grid, Energy Procedia, 12 (2011) 1015-1022.
577 [16]Traube, J. et al., Mitigation of solar irradiance intermittency in photovoltaic power
578 systems with integrated electric-vehicle charging functionality, IEEE Transactions on
579 Power Electronics, 28 (2013) 3058-3067.
580 [17]Li, X., Hui, D., Lai, X. Battery energy storage station (BESS)-based smoothing control of
581 photovoltaic (PV) and wind power generation fluctuations, IEEE Transactions on
582 Sustainable Energy, 4 (2013) 464-473.
583 [18] Ellis, A., Schoenwald, D., Hawkins, J., Willard, S., Arellano, B. PV output smoothing
584 with energy storage, in 38th IEEE Photovoltaic Specialists Conference (PVSC), (2012)
585 001523-001528.
586 [19] Johnson, J. et al., PV output smoothing using a battery and natural gas engine-generator,
587 in 39th IEEE Photovoltaic Specialists Conference (PVSC), (2013) 1811-1816.
588 [20] Marcos, J., Storkl, O., Marroyo, L., Garcia, M., Lorenzo, E. Storage requirements for
589 PV power ramp-rate control, Solar Energy, 99 (2014) 28-35.
590 [21] Johnson, J., Schenkman, B., Ellis, A., Quiroz, J., Lenox, C. Initial Operating Experience
591 of the La Ola 1.2-MW Photovoltaic System, Sandia Report, U.S. Department of Energy,
592 2011.
593 [22] De la Parra, I., Marcos, J., Garca, M., Marroyo, L. Control strategies to use the
594 minimum energy storage requirement for PV power ramp-rate control, Solar Energy, 111
595 (2015) 332-343.
596 [23] Marcos, J., de la Parra, I., Garca, M., Marroyo, L. Control Strategies to Smooth Short-
597 Term Power Fluctuations in Large Photovoltaic Plants Using Battery Storage Systems,
598 Energies, 7 (2014) 65936619.
599 [24] Alam, M. J. E., Muttaqi, K. M., Sutanto, D. A novel approach for ramp-rate control of
600 solar PV using energy storage to mitigate output fluctuations caused by cloud passing,
601 IEEE Transactions on Energy Conversion, 29 (2014) 507-518.
602 [25] Amiri, M., Esfahanian, M., Hairi-Yazdi, M. R., Esfahanian, V. Minimization of power
603 losses in hybrid electric vehicles in view of the prolonging of battery life, Journal of
604 Power Sources, 190 (2009) 372-379.
A novel ramp-rate control strategy based on CES method is proposed
The proposed method is compared with MA and CES methods
Proposed method does not show memory effect that caused BES to operate endlessly
As a result, the size of the BES capacity can be reduced