Comparative impacts of wind and photovoltaic generation on energy storage for small islanded electricity systems

Renewable Energy 80 (2015) 793e805

Contents lists available at ScienceDirect

Renewable Energy journal homepage: www.elsevier.com/locate/renene

Comparative impacts of wind and photovoltaic generation on energy storage for small islanded electricity systems I.G. Mason* Department of Civil and Natural Resources Engineering, University of Canterbury, Christchurch, New Zealand

a r t i c l e i n f o

a b s t r a c t

Article history: Received 27 November 2014 Accepted 19 February 2015 Available online

This paper addresses the annual energy storage requirements of small islanded electricity systems with wind and photovoltaic (PV) generation, using hourly demand and resource data for a range of locations in New Zealand. Normalised storage capacities with respect to annual demand for six locations with winterpeaking demand profiles were lower for wind generation than for PV generation, with an average PV:wind storage ratio of 1.768:1. For two summer-peaking demand profiles, normalised storage capacities were lower for PV generation, with storage ratios of 0.613:1 and 0.455:1. When the sensitivity of storage was modelled for winter-peaking demand profiles, average storage ratios were reduced. Hybrid wind/PV systems had lower storage capacity requirements than for wind generation alone for two locations. Peak power for storage charging was generally greater with PV generation than with wind generation, and peak charging power increased for the hybrid systems. The results are compared with those for country-scale electricity systems, and measures for minimising storage capacity are discussed. It is proposed that modelling of storage capacity requirements should be included in the design process at the earliest possible stage, and that new policy settings may be required to facilitate a transition to energy storage in fully renewable electricity systems. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Storage Electricity Renewable Wind PV

1. Introduction Energy storage is a key element in balancing supply and demand within islanded electricity systems, especially when variable resources such as wind and solar are used for generation. Time frames for balancing may range from minutes to seasons. Where maximum long-term utilisation of generated energy, or minimal curtailment, is an objective, then seasonal storage is of particular interest. If supply and demand are highly correlated, seasonal storage requirements will be minimised. This may occur, for example, at a location employing solar photovoltaic (PV) generation, and where annual electricity demand peaks during the summertime, due to the widespread use of air conditioning. Conversely, where supply and demand are weakly correlated, required storage capacity will be greatest. Previous researchers have addressed the increased requirement for energy storage as the penetration of variable renewable * Department of Civil and Natural Resources Engineering, University of Canterbury, Private Bag 4800, Christchurch 8140, New Zealand. Tel.: þ64 3 366 7001; fax: þ64 3 364 2758. E-mail address: [email protected]. http://dx.doi.org/10.1016/j.renene.2015.02.040 0960-1481/© 2015 Elsevier Ltd. All rights reserved.

generation increases [1e5], storage design, including the development of optimisation techniques [6e11], and the financial value of energy storage systems [9,12e15]. In addition, life cycle energy and greenhouse gas (GHG) emissions associated with pumped hydro energy storage (PHES), compressed air energy storage (CAES) and battery energy storage have been reported [16]. Reported round trip efficiencies for technologies suitable, or potentially suitable, for seasonal storage have ranged from 70 to 81% for PHES [4,16] and 50e66% for CAES ( [17e19], to 36% for electrolysis and fuel cell systems [4]. The correlation of wind and solar generation with demand for the whole of Europe between 2000 and 2007 has been investigated, at monthly and hourly resolution, by Heide et al. [4]. For a 100% renewable electricity system, seasonal storage for a wind-only case was shown to be less than for a solar-only case. However, storage was minimised with a generation mix producing 55% of the energy from wind and 45% of the energy from solar. In a related panEuropean electricity system study Rasmussen et al. [20] explored combinations of high-efficiency short-term (6 h) storage and low efficiency, hydrogen-based, seasonal storage, at hourly resolution over an 8-year period. It was found that incorporation of the shortterm storage would reduce seasonal storage significantly in these

794

I.G. Mason / Renewable Energy 80 (2015) 793e805

16.8%. In a modelling study of a 100% renewable electricity system, comprising wind and PV generation, for the contiguous states of the USA divided into 10 regions, and using 32 years of data at hourly resolution, Becker et al. [6] found energy storage to be less for solaronly than for wind-only scenarios, in contrast to the results of Converse. Assuming 100% round trip efficiencies, normalised storage values, as a fraction of annual demand, for the whole country were approximately 0.21 for PV only, 0.30 for wind only, and 0.18 for an optimal generation mix of 35% wind and 65% PV. On a regional basis, normalised storage values ranged from 0.13 to 0.52. Whilst the literature has indicated the relative magnitudes of storage requirements where wind and PV generation are implemented on a country, or continental, scale, little is known about these relationships for more localised islanded systems. In addition, where generation and usage are geographically proximate, effects arising from widespread dispersion of generation and aggregation of demand will be absent, and optimal wind/PV generation mixes may, or may not, exist. Thus it is pertinent to investigate storage capacity requirements at these smaller scales. The specific objectives of this research were to: a) quantify the impact of wind and PV generation on annual energy storage capacities for islanded electricity systems; b) to investigate the sensitivity of energy storage capacity to variations in demand and resource patterns; and c) to determine whether hybrid wind/PV generation mixes require lower storage capacities than those for wind or PV generation individually. 2. Methods 2.1. Background

Fig. 1. Study locations.

circumstances. The energy outputs required to supply 100% of total demand to the entire USA electricity system, at monthly resolution, were determined by Converse [5], by scaling up reported wind and solar outputs between 2000 and 2007, from approximately 0.86% and 0.015% of annual demand respectively. Resulting annual storage values, expressed as a percentage of annual demand, and assuming 100% round trip efficiencies, ranged from 4.9% to 13.5% for wind, and from 21.3% to 31.7% for solar. For a 50:50 mix, the range was 7.5%e

The research was conducted using New Zealand electricity demand and energy resources data. New Zealand is a temperateclimate country in which overall electricity demand peaks during the winter months of June, July and August [21]. Solar insolation is naturally lowest during this period. However, monthly average wind speeds have historically peaked during the spring and summer months of October, November, December and January [22]. The ratio of maximum to minimum monthly values was approximately 1.3:1 for wind speed and 4.1:1 for solar radiation. October through January were confirmed as peak months for wind generation in a subsequent study using a 19-year synthetic wind data set to model electricity output from established, and potential, wind farm sites [23]. 2.2. Locations, study period and data sources Six residential locations were selected in order to cover a range of latitudes and climatic conditions, from the sub-tropical north, through the more temperate central regions to the cooler southern

Table 1 Selected annual average climate characteristics for the period 2000e2011. Location

Town

Climate station

Latitudea ( S)

Longitudea ( E)

Temperature ( C)

Global radiation (kWh/m2.y)

Wind speedb (m/s)

A

Whangarei

B C D

Hicks Bay Hawera Blenheim

E

Akaroa

F G

Invercargill “Farm Dairy”

Whangarei Aero Kaitaia Aeroc Hicks Bay Hawera Blenheim Aero Cape Campbellc Akaroa Le Bons Bayc Invercargill Aero Tiwai Point

35.769 35.067 37.564 39.612 41.523 41.731 43.809 43.946 46.417 46.587

174.364 173.287 178.314 174.292 173.865 174.274 172.966 173.119 168.330 168.376

16.0 e 15.0 12.8 12.7 e 12.8 e 10.0 10.7

1441 e 1516 1465 1521 e 1400 e 1217 1230

e 4.5 6.7 5.2 e 7.6 e 6.2 4.7 5.7

a b c

Climate station location. At 10 m. Closest useable wind resource.

I.G. Mason / Renewable Energy 80 (2015) 793e805

795

Table 2 Feeder and demand information. Location

Town

Annual load (MWh)

ICPsa (no.)

Domestic loads (% by no.)

Comments

A B C D E F G

Whangarei Hicks Bay Hawera Blenheim Akaroa Invercargill “Farm Dairy”

9610 1767 11,180 10,428 8464 7093 125

1256 302 1217 1320 (1194) 747 e

95 67 e 78 e 100 0

Other loads: pumps, communications, shops Other loads: 32% (97) < 30 kVA; 1% (3) 100 kVA Data for days 1e90 replaced by 2010 measurementsb Other loads: supermarkets; small businesses ICPs for 2011 unknown; bracketed figure is for 2014 Includes several small shops e

a b

Installation Control Points. Demand from days 1e90 in 2011 was higher than usual due to the feeder being used to back up another feeder.

part of the country (Fig. 1). In order to investigate cases where electricity usage peaks during the summer months, the demand profile for residential location F was inverted, by transposing measurements in time by 6 months, and a farm dairy sited near location G (Fig. 1) included. A one-year study period from 1 January 2011 to 31 December 2011 was chosen for the residential sites. The period for the farm dairy was constrained by the available demand data to 1 August 2006 to 31 July, 2007. Global solar radiation, 10-m wind speed and temperature data, all measured at the climate stations listed in Table 1 were sourced from a national database [24]. A summary of long-term climate characteristics is given in Table 1. All data inputs for modelling were hourly averages, or hourly totals, and were reported at New Zealand

Standard Time (NZST). Any gaps found in the climate station data sets were filled, either by interpolation, or by inserting data from adjacent, or similar, time periods. Electricity demand data, recorded at half-hourly resolution for specified 11 kV feeders with primarily residential loads, was obtained from local electricity distribution (lines) companies. This data was reported as either current (A), or as power (kW). Data recording times were adjusted from New Zealand Daylight Time to NZST where necessary, and gaps filled as described above. Annual loads, connection (ICP) numbers and other location-specific characteristics are given in Table 2. Farm dairy demand data, recorded as kWh per half-hour, was sourced from a previous research study [25]. For modelling purposes, all demand data were converted to hourly average demand (kWh).

Fig. 2. Daily demand: a) Locations A, D, E; b) Locations B, C, F.

Fig. 3. Daily demand: a) Location F-inverted; b) Location G (farm dairy).

Fig. 4. Weekly demand patterns for location A. a) Summer (20e26 February, 2011); b) Winter (14e20 August, 2011).

2.3. Modelling approach Solar PV electricity production was modelled using the HDRK anisotropic sky model [26] applied to 190 W(p) polycrystalline panels (Komaes KM190, Komaes, Japan). Performance was corrected for temperature, and then a de-rating factor of 0.77 [27] applied to account for losses due to DC-AC conversion, wiring and

Fig. 5. Weekly demand pattern for the farm dairy (Location G; 20e26 February, 2007).

Fig. 6. Daily Wind Run patterns. a) Location D; b) Location A.

I.G. Mason / Renewable Energy 80 (2015) 793e805

797

storage. Storage charging and discharging efficiencies of 0.9 each way were employed, which would simulate a high-efficiency PHES system [4]. Storage levels were normalised by expressing them as a fraction of annual demand and the required storage capacity computed from the difference between the maximum and minimum hourly levels. For clarity of presentation, initial storage capacities were adjusted until the minimum level differences closed to within 0.001% of zero, so that the peak level equalled the storage capacity. To explore the sensitivity of storage capacity to changes in demand and energy resource patterns, simulations were conducted in which: a) the energy resources were held constant and used with the location-specific demand profiles; and b) the demand profile was held constant and used with location-specific energy resources. It should be noted that the term ‘storage capacity’ refers to working storage capacity, and that the installed storage capacity will typically be greater, by an amount dependent on the technology employed. The impact of hybrid wind/PV systems on storage capacity was investigated by setting PV generation to generate 10% increments of annual demand and then adjusting wind generation in order to meet the remaining demand, and to return the final storage level to the initial storage level. Model outputs were: a) wind and solar PV energy generation (kWh); b) surpluses and deficits (kWh); and c) normalised storage levels. 3. Results 3.1. Demand and resource characteristics

Fig. 7. Daily Global Radiation patterns: a) Location B; b) Location F.

The expected peak in electricity usage during the winter months was demonstrated in most of the annual residential demand profiles (Fig. 2a, b). For clarity of presentation these show daily demand. Whilst Location B appears to have a relatively flat profile, this is primarily a scaling effect, and a slight increase in demand did occur during the winter period. A brief summer peak in late December and early January may be attributed to the holiday period at this time. The demand pattern for Location F demonstrated the greatest ‘swing’ between summer and winter, and was thus chosen as the basis for the inverted profile (Fig. 3a). The farm dairy demand profile (Fig. 3b) showed a slight downward trend over the season, followed by extremely low demand during the winter off-season. This pattern reflects the timing of the milking season in New Zealand, which typically begins in early August and runs through to the end of May. Weekly patterns showed typical morning and evening peaks, and late night troughs, during most of the year. Examples, plotted at half-hourly resolution, are shown for Location A (Fig. 4a, b). For two

other factors. Tilt angle was adjusted in order to maximise annual energy output. Wind speeds were scaled to a hub height of 30 m using the power law method with a roughness factor of 0.2 [28]. Wind electricity production was determined by modelling the power curve of a 500 kW turbine with a hub height of 30 m (Windflow Technologies, Christchurch, New Zealand), and calculating the hourly output. To simplify the analysis, array losses from multiple turbines were ignored. The required installed capacity of both PV and wind generation was determined by matching annual output to annual demand [29,30], followed by adjustments in installed capacity to: a) meet storage charging and discharging losses; and b) return the final storage level to the initial level. Fractions of the chosen turbine and PV panel were allowed for modelling purposes and the final storage level adjusted to with 0.001% of the initial level. Surpluses and deficits were computed at hourly resolution. All hourly surpluses were stored, and all hourly deficits were met from

Table 3 Annual average energy production values. Location

Average wind speed (m/s)a

Wind power density (W/m2)a,b

Wind CFc & energy yield (MWh/y)d

Global radiation (W/m2)

Mean ratio of diffuse to global radiation (Id/I)e

PV CFc & energy yield (kWh/y)f

A B C D E F G

5.7 8.3 6.7 9.1 8.0 5.9 7.7

219 638 354 911 823 375 904

0.13 0.31 0.21 0.42 0.31 0.18 0.29

171 174 170 171 163 136 133

0.72 0.69 0.71 0.67 0.71 0.78 0.79

0.115 0.123 0.117 0.122 0.117 0.097 0.096

a b c d e f

(554) (1379) (901) (1833) (1365) (793) (1271)

Calculated at 30 m, using power law roughness exponent alpha ¼ 0.2. Calculated using measured hourly wind speeds and air density ¼ 1.2 g/l. Capacity Factor. Per turbine. Modelled Id. Per panel.

(192) (205) (194) (203) (194) (162) (160)

798

I.G. Mason / Renewable Energy 80 (2015) 793e805

Table 4 Generation requirements and excess energy factors (local demand and local resources). Location

PV tilt

PV panelsa

Wind turbinesa

PV excess energy factor

Wind excess energy factor

A B C D E F G

30 32 27 33 33 30 31

56,969 9765 65,177 57,436 49,782 50,469 878

19.82 1.42 13.95 6.27 6.90 10.27 0.12

1.22 1.13 1.13 1.12 1.14 1.15 1.12

1.23 1.11 1.12 1.10 1.11 1.15 1.13

a

Rounded.

of the other locations, separate morning and evening peaks disappeared or were flattened during the summer months, resulting in a broad daytime peak only. The daily farm dairy demand pattern also showed morning and evening peaks, reflecting the New Zealand practice of twice-daily milking, typically between 0600 and 0900 h, and 1500e1800 h respectively (Fig. 5). Demand was very low in the periods between milking, and at night-time, with refrigeration and water heating being the major loads.

Seasonal trends in the wind resource were slight, and whilst large hourly and daily fluctuations were observed, these short-term patterns were relatively evenly distributed over the year. Examples are shown in Fig. 6 for the best residential resource (location D), and the poorest residential resource (location A), using daily 10-m wind run data. Solar resource patterns demonstrated the anticipated seasonal trends, with embedded shorter-term fluctuations due to varying daylight hours and weather fluctuations. Patterns for the best resource (location B), and the poorest resource (location F) are illustrated in Fig. 7, using daily global radiation data for clarity of presentation. Annual wind speed and power density values at 30 m indicated wind resources ranging from marginal (location A) to superb (locations D and G) (Table 3). When evaluated according to measured capacity factor and wind turbine energy yield, however, location D was found to be the best site, with a capacity factor of 0.42. Annual average global radiation data showed little variation between locations A and E, but there was a substantial reduction for the southern-most locations F and G (Table 3). The latter sites also had the highest annual average proportion of diffuse radiation. When

Fig. 8. Normalised storage level patterns for residential sites: a) Location A; b) Location B; c) Location C; d) Location D; e) Location E; f) Location F.

I.G. Mason / Renewable Energy 80 (2015) 793e805

evaluated using the capacity factor and annual PV energy output, the top locations were B and D, with locations A, C and E close behind.

799

Table 5 Normalised storage capacities. Location Data inputsa,b

3.2. Generation requirements The required wind turbine and PV panel numbers are shown in Table 4. The amount of additional generation needed to provide for storage losses and final storage level adjustment is represented by the excess energy factor which is given by total energy generated divided by demand. Typically 10%e15% additional energy was required. Optimal tilt angles for the PV arrays generally ranged from 30 to 33 .

3.3. Storage requirements The annual storage level patterns for the residential locations with local demand and local resources showed a common pattern for PV generation, characterised by a peak toward the end of summer (around Julian day (J) ¼ 100) followed by a steady drawdown through until early spring (J ¼ 200e300), after which storage levels began to rise again (Fig. 8). This reflected the overlying trend of the solar resource, with minor variations due to location-specific resource and demand patterns. For wind generation, a similar, although less pronounced and less regular, pattern in storage levels was observed for locations A, D, E and F (Fig. 8a, d, e, f). Otherwise storage for wind generation showed either a winter peak (Fig. 8b) or a summer peak (Fig. 8c). For the inverted residential demand scenario, the storage level for wind peaked very late, and that for PV very early, in the year (Fig. 9a). In the case of the farm dairy, high storage levels for PV generation occurred in August (J ¼ 150) and again in February (J ¼ 50), whilst for wind generation the peak storage level was reached at the end of the milking season (J ¼ 200) (Fig. 9b).

Fig. 9. Normalised storage level pattern: a) Location F-inverted demand; b) Location G - farm dairy.

A B C D E F Average

Local Local Local Local Local Local Local

Demand Demand Demand Demand Demand Demand Demand

and and and and and and and

Local Local Local Local Local Local Local

A B C D E F Average

Local Local Local Local Local Local Local

Demand Demand Demand Demand Demand Demand Demand

and and and and and and and

Reference Reference Reference Reference Reference Reference Reference

A B C D E F Average

Local Local Local Local Local Local Local

Resources Resources Resources Resources Resources Resources Resources

and and and and and and and

Resources Resources Resources Resources Resources Resources Resources Resources Resources Resources Resources Resources Resources Resources

Reference Reference Reference Reference Reference Reference Reference

Demand Demand Demand Demand Demand Demand Demand

Wind PV

Storage ratio (PV:Wind)

0.138 0.075 0.128 0.122 0.062 0.178 0.114

0.167 0.112 0.234 0.167 0.190 0.345 0.192

1.211:1 1.500:1 1.822:1 1.368:1 3.073:1 1.934:1 1.768:1

0.122 0.080 0.131 0.122 0.099 0.196 0.116

0.169 0.133 0.180 0.167 0.158 0.249 0.167

1.383:1 1.674:1 1.371:1 1.368:1 1.588:1 1.275:1 1.492:1

0.184 0.167 0.170 0.196 0.157 0.178 0.175

0.247 0.223 0.304 0.249 0.282 0.345 0.275

1.339:1 1.338:1 1.794:1 1.275:1 1.799:1 1.934:1 1.580:1

F-invert Local Demandc and Local Resources 0.201 0.124 0.613:1 F-invert Local Demandc and Reference Resources 0.104 0.100 0.955:1 Gd Local Demand and Local Resources 0.176 0.080 0.455:1 Note: aReference Resources from Location D; bReference Demand from Location F; c Inverted Demand; dFarm Dairy.

Fig. 10. Storage versus PV penetration in wind/PV hybrid systems: a) Locations A, B; b) Locations CeF. Key: A-F ¼ Locations A-F.

800

I.G. Mason / Renewable Energy 80 (2015) 793e805

Fig. 11. Cumulative normalised charge and discharge patterns for locations with winter-peaking demand. Key: a-f ¼ wind for Locations A-F; a'-f' ¼ PV for Locations A-F.

I.G. Mason / Renewable Energy 80 (2015) 793e805

Normalised storage capacities for the residential locations, ranged from 0.062 to 0.178 for wind generation, and from 0.112 to 0.345 for PV generation, when using local demand and local energy resources (Table 5). PV generation required a greater storage capacity than wind generation in all cases, with the increases ranging from 21.1% (location D) to 205.9% (location E). The average storage ratio was 1.768:1. For the two scenarios where demand peaked in summer (Location F-invert and Location G), the normalised storage capacities were 0.201 and 0.176 for wind generation, 0.124 and 0.080 for PV generation (Table 5). In both cases, PV generation required substantially less storage capacity than wind generation, with decreases of 48.7% and 54.6% respectively. For the sensitivity analyses involving local energy resources with standard demand, the demand data from Location F were used as the reference. This site showed a distinct seasonal pattern of electricity usage, peaking during the winter months. For wind the normalised storage requirement ranged from 0.157 to 0.196, and for PV from 0.223 to 0.345 (Table 5). The storage ratio ranged from 1.275:1 to 1.934:1, a reduction in the range for local demand and local resources, with a mean value of 1.580:1. In comparison, for the sensitivity analyses involving local demand with standard energy resources, the normalised storage requirement for wind ranged from 0.080 to 0.196, whilst that for PV ranged from 0.133 to 0.249 (Table 5). Here the energy resources from Location D were used as the reference, since this site had the best wind resource, plus a very high quality solar resource. The storage ratio range was 1.275:1 to 1.674:1, again a contraction from the range for local demand and local resources. A lower mean storage ratio of 1.492:1 was found.

3.4. Optimal generation mixes An optimal mix of wind and solar generation in terms of storage capacity was found for two locations (Fig. 10a), but not for the remainder (Fig. 10b). For locations A and B, minimum storage values occurred at 60% and 40% PV penetration respectively. Normalised storage capacity decreased to 0.118 for Location A and to 0.055 for Location B.

801

4. Discussion 4.1. Charge and discharge patterns Correlations between the temporal patterns of cumulative input and output to storage, i.e. charge and discharge, are shown in Figs. 11 and 12. These profiles reflect the balance between supply and demand, and provide a visual explanation for the observed differences in storage profile shapes (Fig. 8aef) and in required storage capacity. Two elements are pertinent to the discussion: a) the overlying seasonal trends; and b) the short-term fluctuations over periods of hours and days. In the wind generation scenarios, the overlying charging trends for the year from almost linear (Fig. 11b, e) to what may be described as inverse sigmoidal (Fig. 11a, d, f). Over shorter time intervals, charging patterns were typically characterised by a series of relatively short ramping up periods followed by relatively short plateaux (Fig. 11aef; Fig. 12a, b). In contrast, charging patterns with PV generation can be observed to show an overlying, and more pronounced, inverse sigmoidal trend in all cases, and to build relatively smoothly over smaller time intervals (Fig. 11a'-f'; Fig. 12a', b'). These observations reflect the relatively even availability of wind throughout the year versus the seasonal trends in the solar resource, along with the hourly variations in both. Discharge patterns for wind generation were typically more linear, whilst those for PV generation reflected the solar trends. In relation to storage capacity, which can also be derived from these graphs by measuring the maximum difference between the two patterns, it can be seen that the charge and discharge profiles for winter-peaking demand patterns are consistently closer together in the wind scenarios than in the PV scenarios, and that the maximum differences are smaller. (It should be noted that when the charge and discharge profiles cross e.g. as in all the PV graphs, the storage capacity is given by the sum of the absolute maximum differences). Those locations with the lowest storage capacity requirement for wind generation (B, D) had the highest degree of correlation between charge and discharge (Fig. 11b, d), and thus supply and demand. Similarly location F, with the highest wind storage requirement, had

Fig. 12. Cumulative normalised charge and discharge patterns for locations with summer-peaking demand. Key: a-b ¼ wind for Locations F-invert & G; a'-b' ¼ PV for Locations F-invert & G.

802

I.G. Mason / Renewable Energy 80 (2015) 793e805

the poorest degree of correlation, with profiles further apart (Fig.11f). The same relationships may be observed for PV generation (Fig. 11f') and also for the summer-peaking demand patterns (Fig. 12). 4.2. Sensitivity When constant resources (Location D) were applied to local demand profiles, the changes in normalised storage capacity for the winter-peaking demand patterns were relatively small, varying from 0.034 to þ0.037 for wind generation (Fig. 13a) and from 0.096 to þ0.021 for PV generation (Fig. 13b). In contrast when a constant demand profile (Location F) was applied to local resources, an increase in storage capacity resulted in all cases except for location F with inverted demand. The ranges for wind and PV were þ0.046 to þ0.095, and þ0.080 to þ0.135, respectively, for locations A-F. In the case of location F with inverted demand, application of the reference resources resulted in a large decrease in storage capacity for both wind and PV generation, which may be attributed to the use of better energy resources compared to those actually available at Location F. The design implications here are that changes in demand profiles can potentially be used to substantially influence storage capacity. In this study a more pronounced winter-peaking profile resulted in notable storage capacity increases. Thus the investigation of any prospects for bringing demand into closer alignment with the resource patterns is suggested as a worthy topic for further investigation. 4.3. Charging and discharging power The magnitude and frequency of hourly power requirements provided important storage design information. In this regard,

Fig. 14. Surplus and deficit frequency distributions in 50 kW bins for Location D: a) wind; b) PV.

surplus and deficit frequency patterns differed for wind generation versus PV generation, in almost all cases. Typically, frequency profiles for PV generation were characterised by longer tails for surpluses, and higher peaks for deficits, as shown for Location D in Fig. 14. For winter-peaking demand, PV scenarios were also characterised, with one exception, by maximum hourly surpluses of considerably greater magnitude than those for wind generation (Table 6). This means that PV systems, in these scenarios, would typically require 2 to 3 times the average hourly power capacity for input to storage, e.g. a larger capacity pump in a PHES system. In contrast, the maximum deficit hourly power ratios for wind and PV generation were quite similar, varying by 0% to approximately 15%. This indicated a minor influence of generation type on hourly average discharge power. Where optimal generation mixes existed (Locations A and B), the maximum surplus power ratio with respect to wind increased in the case of Location A, and decreased in the case of Location B. This gave surplus maxima ratios (hybrid:wind) of 0.720:1 and 1.237:1 respectively. Maximum deficit values showed minor changes only.

Table 6 Ratio of maximum surplus and deficit power (local demand and local resources).

Fig. 13. Sensitivity of storage capacity to changes in demand and resource patterns. a) Wind generation; b) PV generation. Note: Reference resources: Location D; Reference demand: location F.

Location

Ratio of surplus maxima (PV:Wind)

Ratio of deficit maxima (PV:Wind)

A B C D E F F-invert G

0.643:1 3.210:1 1.592:1 2.498:1 2.064:1 1.901:1 1.995:1 1.818:1

1.000:1 1.065:1 0.971:1 0.846:1 1.052:1 0.989:1 1.000:1 0.958:1

I.G. Mason / Renewable Energy 80 (2015) 793e805

4.4. Comparison with country-scale requirements In their modelling of electricity system storage requirements for Europe, Heide et, al. [4] found storage capacities for 100% wind, and 100% solar, generation of approximately 3.1 and 2.8 times the average monthly demand, respectively. These ratios translate to 0.258 times the annual demand for wind, and 0.233 times the annual demand for solar, with a PV:wind storage ratio of 0.903:1. For their optimum generation mix (50e60% wind and 40e50% PV energy), the storage capacity was about 1.5 times the monthly demand, or 0.125 times the annual demand. Compared to the findings of the present study, the value for wind is considerably higher, but the ratios for solar and for the optimum generation mix are within the range found here for localised systems. The surplus generation factor of Heide et al. is similar to the excess energy factor in the present study, with the range of approximately 1.12e1.16 for PHES close to the range found here of 1.10e1.23 (Table 4). In one study involving the continental USA [5] reported data showed considerably greater storage capacities for solar than for wind (Fig. 15), which, along with the storage ratio of 3.235:1, is similar to findings obtained for Location E in the present study. However the data of Becker et al. [6] revealed a storage ratio of 0.686:1 for the contiguous states of the USA, and a range of 0.366:1 to 0.700:1 for the 10 regions. These ratios differ markedly from those for the 6 residential locations in the present study, and, as the authors indicate, likely reflect summer-peaking demand profiles, along with differing solar resource patterns. The presence of an optimal generation mix in relation to storage capacity for two locations in the present study, and the absence of an optimal mix for the remainder, requires further investigation. Geographical dispersion of resources, and geographical aggregation of loads, may be hypothesised to account for an optimal mix at country scale, however the fact that this also occurred at smallscale, where such effects were absent or minor, indicates that other factors need to be considered. 4.5. Policy implications A primary suggestion for consideration by policy makers, planners and designers, is to ensure that modelling of storage requirements for proposed islanded electricity systems occurs at the earliest possible stage in the planning and design process. The findings of the present study indicate that normalised working storage capacities may differ by a factor of as much as 5, from as low as 0.062 to as high as 0.345, representing substantial differences in required infrastructure and investment. The study has also shown that, from a storage capacity point of view, wind generation will be

Fig. 15. Normalised storage levels for the USA electricity system (data source: [5]).

803

preferred in some cases, PV generation will be preferred in others, and that an optimal generation mix may exist, but not in all situations. Thus an early indication of the likely status and magnitude these issues will be valuable. Evaluation will be facilitated where historic demand data is readily available, along with wind and solar resource data obtained on-site or from climate stations. If data is limited, synthetic demand profiles and climate data from satellite observations and meteorological modelling may be employed. Such analyses would also be valuable where grid-tied systems are contemplated, as imported and exported energy, and power, profiles will be quantified. One limitation of the present study was the use of 1-year data sets, and whilst this enabled relative differences to be established, evaluation over longer time frames is recommended for design purposes. An evaluation of sub-hourly issues, including power quality, will also be needed at the detailed design level. Further studies, covering a range of demand profiles and energy resource patterns in other countries, will be valuable in building a more comprehensive data set of normalised storage capacity values for consideration by researchers, policy makers, planners and designers. Different, typically lower, storage efficiencies than employed in the present study will have a marked impact on required working capacity, and thus specification of the most efficient practical option is important. PHES can have a very respectable round-trip efficiency of approximately 80%, although it may be as low as 70%, and less in some circumstances. In practice however, not every location will have suitable topography for PHES or social acceptance for its construction. Increased interest in using coastal sites, with the ocean as the lower reservoir e.g. Hearps et al. [31] is a promising option which may assist in enabling the implementation of PHES, as may the use of underground reservoirs e.g. Steffen [32]. Battery systems, which can suit relatively small storage capacity requirements, have previously reported efficiencies ranging from 95 to 98% for Li-ion batteries to 40e50% for zinc-air batteries [33]. Hydrogen and CAES are also options, although present efficiencies are comparatively low. The installed storage capacity, as opposed to the working storage capacity values reported here, is also important. For example, PHES reservoirs will be subject to minimum water levels and will require an allowance for freeboard, making the physical structure somewhat larger than required for working capacity alone. Similarly, battery installed capacity will be a function of recharge and discharge level constraints. Given the substantial storage capacity requirements modelled, an alternative design approach, involving specification of significant over-capacity in generation, which results in reduced storage capacity, at the expense of increased energy spillage e.g. Budischak et al. [34] deserves evaluation for islanded electricity systems. In their study of storage requirements for off-grid residential PV/battery systems at five locations in the USA, Bronski et al. [35] reported data from which the storage requirements and the extent of spillage for individual residences when using this approach may be quantified. In this case, using 2014 results, normalised storage ratios ranged from 0.007 to 0.020, which is considerably lower than the range found in the present study. Part of this effect may be due to the 85.5% round trip efficiency calculated for this system (DC-battery-DC-AC), versus 81% used in the present study (AC-storage-AC), although it is likely to be relatively small as indicated by a decrease in normalised storage for Location D from 0.167 to 0.159 when a round trip efficiency of 85.5% was employed. However, the excess energy ratios were much higher ranging from 2.03 to 2.21, along with lower capacity factors of 0.068e0.090 based on demand. Whether such energy spillage is justified in terms of an overall systems analysis incorporating thermodynamic considerations and life cycle analyses provides an interesting topic of debate and investigation for policy makers,

804

I.G. Mason / Renewable Energy 80 (2015) 793e805

designers and researchers. If suitable secondary loads can be identified, then this approach may have additional merit, and high associated capacity factors of 0.148e0.183, based on energy production, were indicated by the data of Bronski et al. [35]. As indicated above, demand-side management (DSM) may have potential to minimise storage capacity and this should also be included at the planning stage. This approach was signalled by Denholm and Hand [36] in discussing load shifting, and Bronski et al. [35] found very large reductions in storage capacity when DSM was implemented. In their study the latter reported data showing 44%e67% reductions in storage capacity resulting from DSM, but an increased amount of spillage as indicated by the extended upper values for the excess energy ratio, which ranged from 2.06 to 2.85. In terms of further research, an interesting challenge for DSM researchers will be to investigate the prospects of more closely aligning long-term (e.g. yearly) demand profiles with long-term variable generation patterns, in addition to the more common application of DSM to ‘peak-shaving’ at hourly and daily time scales. Finally, it is suggested that new policy settings may be needed in order to promote energy storage in fully renewable electricity systems. In this context it is informative to reflect upon the energy storage viewpoints of two New Zealand government-owned generators, the New Zealand grid operator, the New Zealand Electricity Authority, two NGOs and eight individual experts, as gleaned from questionnaires and semi-structured interviews [37]. Pumped hydro and battery storage were seen by most respondents as prohibitively, or extremely, expensive at utility scale. Thermal ‘peakers’ fuelled by fossil gas or fossil diesel were viewed as the most likely solution to increased penetration of variable renewables, and batteries only likely to be employed for very high value applications such as providing 6-s instantaneous reserve capacity. However “a price of carbon of around NZ$100/tCO2-e was seen as making these technologies much more competitive, and climate change mitigation was seen as a strong driver of these storage options.” In these responses there is an implicit recognition of the energy stored, by nature, within fossil fuels, the bulk storage of which requires only the provision of simple containment structures, such as tanks for oil and gas, or fences for coal. Thus the deficits modelled here could, in a partially renewable system, be met using a wind/diesel hybrid generation mix, for example, and the surpluses either spilled, or dumped into secondary loads. At the present time however, when the call from the scientific community is for reductions in global GHG emissions of 40e70% by 2050 in order to limit warming to  2 C, with action sooner rather than later [38], or as stated by the World Bank ‘The time to act is now’ [39], the continued widespread use of fossil fuels to balance supply and demand, in lieu of renewable energy storage, needs to be progressively phased out. A transitional policy setting where the energy consumption of fossilfuelled ‘peakers’ is reduced over time may be appropriate in some situations, with the objective of minimising, if not eliminating, GHG emissions. To facilitate a full transition, however, it is suggested that complimentary measures, such as the equivalent of a feed-in-tariff for low-GHG-emitting energy storage technologies, and capacity payments, may need to be implemented and retained until carbon pricing, and/or other policy initiatives, provide adequate support. 5. Conclusions For six small islanded electricity systems with winter-peaking demand profiles, wind generation was shown to be preferable to PV generation in relation to seasonal storage requirements and in the case of summer-peaking demand profiles, the reverse was demonstrated. The ratio of storage required for PV generation to that needed for wind generation averaged 1.768:1, with a range of

1.211:1 to 3.073:1. Storage capacities, when normalised as a fraction of annual demand, averaged 0.114 for wind generation, with a range of 0.062e0.178, and for PV generation averaged 0.192, with a range of 0.112e0.345. In the case of two simulated systems with summerpeaking demand profiles, the ratios of storage capacity when comparing PV to wind generation were 0.613:1 for a residential location, and to 0.455:1 for a farm dairy. When demand was held constant for the systems with winterpeaking demand profiles, the mean storage ratio decreased to 1.580:1, and when the energy resources were held constant a ratio of 1.492:1 was obtained. In these situations the range of storage capacities for both wind and PV generation contracted slightly. Optimal generation mixes with respect to storage capacity, at PV penetration values of 60% and 40%, were found for two locations. Under these circumstances normalised storage capacities were reduced from 0.138 to 0.118, and from 0.075 to 0.055, respectively. Storage capacity was demonstrated to be related to the degree of correlation between cumulative charging and discharging patterns. Charging profiles for wind generation and PV generation reflected differing daily and seasonal variations in the respective resources. Peak power requirements for storage charging were found to be greater for PV generation than for wind generation, in all but one case, with the ratio of peak power for PV systems to peak power for wind systems ranging from 1.592:1 to 3.210:1. In contrast, the maximum deficit power showed only minor variation. It is suggested that consideration and modelling of storage requirements should occur at the earliest possible stage in a project, and that policy settings which facilitate a transition to low-GHGemitting energy storage options in renewable electricity systems be considered. Acknowledgements The author thanks those staff at six New Zealand lines companies who kindly supplied residential demand data and associated information, and Matthew Hughes for preparing Fig. 1. References [1] Dell RM, Rand DAJ. Energy storage e a key technology for global energy sustainability. J Power Sources 2001;100(1e2):2e17. [2] Anderson D, Leach M. Harvesting and redistributing renewable energy: on the role of gas and electricity grids to overcome intermittency through the generation and storage of hydrogen. Energy Policy 2004;32(14):1603e14. [3] Caralis G, Zervos A. Analysis of the combined use of wind and pumped storage systems in autonomous Greek islands. IET Renew Power Gener 2007;1(1): 49e60. [4] Heide D, von Bremen L, Greiner M, Hoffmann C, Speckmann M, Bofinger S. Seasonal optimal mix of wind and solar power in a future, highly renewable Europe. Renew Energy 2010;35(11):2483e9. [5] Converse AO. Seasonal energy storage in a renewable energy system. Proc Ieee 2012;100(2):401e9. [6] Becker S, Frew BA, Andresen GB, Zeyer T, Schramm S, Greiner M, et al. Features of a fully renewable US electricity system: optimized mixes of wind and solar PV and transmission grid extensions. Energy 2014;72:443e58. [7] Bueno C, Carta JA. Technical-economic analysis of wind-powered pumped hydrostorage systems. Part 1: model development. Sol Energy 2005;78(3): 382e95. [8] Bueno C, Carta JA. Technical-economic analysis of wind-powered pumped hydrostorage systems. Part II: model application to the island of El Hierro. Sol Energy 2005;78(3):396e405. [9] Brown PD, Lopes JAP, Matos MA. Optimization of pumped storage capacity in an isolated power system with large renewable penetration. Ieee Trans Power Syst 2008;23(2):523e31. [10] Katsaprakakis DA, Christakis DG, Zervos A, Papantonis D, Voutsinas S. Pumped storage systems introduction in isolated power production systems. Renew Energy 2008;33(3):467e90. [11] Bridier L, David M, Lauret P. Optimal design of a storage system coupled with intermittent renewables. Renew Energy 2014;67:2e9. [12] Bathurst GN, Strbac G. Value of combining energy storage and wind in short-term energy and balancing markets. Electr Power Syst Res 2003;67(1):1e8.

I.G. Mason / Renewable Energy 80 (2015) 793e805 [13] Black M, Strbac G. Value of storage in providing balancing services for electricity generation systems with high wind penetration. J Power Sources 2006;162(2):949e53. [14] Black M, Strbac G. Value of bulk energy storage for managing wind power fluctuations. Ieee Trans Energy Convers 2007;22(1):197e205. [15] Anagnostopoulos JS, Papantonis DE. Simulation and size optimization of a pumped-storage power plant for the recovery of wind-farms rejected energy. Renew Energy 2008;33(7):1685e94. [16] Denholm P, Kulcinski GL. Life cycle energy requirements and greenhouse gas emissions from large scale energy storage systems. Energy Convers Manag 2004;45(13e14):2153e72. [17] Pickard WF, Hansing NJ, Shen AQ. Can large-scale advanced-adiabatic compressed air energy storage be justified economically in an age of sustainable energy? J Renew Sustain Energy 2009;1(3):10. [18] Pickard WF, Shen AQ, Hansing NJ. Parking the power: strategies and physical limitations for bulk energy storage in supply-demand matching on a grid whose input power is provided by intermittent sources. Renew Sustain Energy Rev 2009;13(8):1934e45. [19] Cavallo A. In: Kaldellis JK, editor. Compressed air energy storage technologies for wind energy systems. Stand alone and hybrid wind energy systems. Cambridge, UK: Woodhead Publishing Ltd; 2010. [20] Rasmussen MG, Andresen GB, Greiner M. Storage and balancing synergies in a fully or highly renewable pan-European power system. Energy Policy 2012;51:642e51. [21] Transpower System Operator. Recent electricity demand. 2014. Available at: https://www.systemoperator.co.nz/security-supply/sos-weekly-reporting/ electricity-demand. last accessed 28 July, 2014, Transpower, Wellington, New Zealand. [22] Leathwick JR, Wilson G, Stephens RTT. Climate surfaces for New Zealand. Hamilton, New Zealand: Landcare Research; 2002. [23] Bull B. Correlation between wind generation output and hydro inflows. Wellington, New Zealand: Electricity Commision; 2010. [24] NIWA. National Institute for Water and Atmospheric Research. Hamilton: New Zealand; 2014. Cliflo. Available at, http://cliflo.niwa.co.nz/. [25] CAE. 2008. Unpublished data. Centre for Advanced Engineering, Christchurch, New Zealand.

805

[26] Duffie JA, Beckman WA. Solar Engineering of Thermal Processes. 3rd ed. New York, USA: Wiley; 2006. [27] NREL. How to change parameters in legacy calculators. Colorado, USA: National Renewable Energy Laboratory, Golden; 2013. Available at: http:// www.nrel.gov/rredc/pvwatts/changing_parameters.html. [28] Manwell JF, McGowan JG, Rogers AL. Wind energy explained; theory, design and application. New York, USA: Wiley; 2002. [29] Mason IG, Page SC, Williamson AG. A 100% renewable electricity generation system for New Zealand utilising hydro, wind, geothermal and biomass resources. Energy Policy 2010;38(8):3973e84. [30] Mason IG, Page SC, Williamson AG. Security of supply, energy spillage control and peaking options within a 100% renewable electricity generation system for New Zealand. Energy Policy 2013;60(1):324e33. [31] Hearps P, Dargaville R, McConnell D, Sandiford M, Forcey T, Seligman P. Pumped hydro energy storage: Arup-MEI research report, February 2014. Melbourne, Australia: Melbourne Energy Institute; 2014. [32] Steffen B. Prospects for pumped hydro storage in Germany. Energy Policy 2012;45(1):420e9. [33] Skyllas-Kazacos M. Electro-chemical energy storage technologies for wind energy systems. In: Kaldellis JK, editor. Stand-alone and hybrid wind energy systems. Oxford, UK: Woodhead Publishing Ltd; 2010. [34] Budischak C, Sewell D, Thompson H, Mach L, Veron DE, Kempton W. Costminimised combinations of wind power, solar power and electrochemical storage, powering the grid up to 99.9% of the time. J Power Sources 2013;225:60e74. [35] Bronski P, Creyts J, Guccione L, Madrazo M, Mandel J, Rader B, et al. The economics of grid defection. Boulder, USA: Rocky Mountain Institute; 2014. [36] Denholm P, Hand M. Grid flexibility and storage required to achieve very high penetration of variable renewable electricity. Energy Policy 2011;39: 1817e30. [37] Kear G, Chapman R. 'Reserving judgement': perceptions of pumped hydro and utility-scale batteries for electricity storage and reserve generation in New Zealand. Renew Energy 2013;57:249e61. [38] IPCC. Climate change 2014 synthesis report. Geneva, Switzerland: IPCC; 2014. [39] World Bank. Turn down the heat: confronting the new climate normal. Washington DC, USA: World Bank; 2014.