Unlocking the Potential of Renewable Energy with Storage

Abstract

The role of storage in managing the variability in wind, solar and wave energy generation is well understood. Shifting energy from periods of high generation to low generation is seen as an ideal role for storage. However, the rapid fluctuations in wind and solar generation with periods less than 1 h can lead to very significant problems on the grid, reducing carrying capacity of lines and increasing the amount of spinning reserve and regulation services required to unachievable levels. A number of electrical storage technologies which are aimed at removing these rapid fluctuations are now being successfully demonstrated at the MW-scale. The various storage technologies are outlined and an example of wind smoothing system is examined in detail.

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

The issue of how to cope with the inherent variability in renewable energy generation from weather-driven sources such as wind, solar and wave is well known. Shifting energy from windy or sunny days to those with less wind or sun is seen as an ideal role for storage utilising technologies such as pumped hydro and compressed air. The current power generation, transmission and market systems are, in fact able to cope with these longer term trends over several hours or days while variable renewables remain at modest penetration levels. The more pressing issue is fluctuations in wind speeds with periods of 1 h or less which occur during highly convective or stormy weather conditions. Similarly, intermittent cloud can produce very sharp changes in PV solar power generation. These conditions can lead to very significant problems on the grid, reducing carrying capacity of power lines and increasing the amount of spinning reserve and regulation services required to unachievable levels. The only alternative is to curtail the renewable generation which is already being done in several markets.

A number of electrical storage technologies are being developed to both remove these rapid fluctuations and provide support to grid systems with large amounts of solar and wind power. Most of these are now being demonstrated at the MW-scale.

Whilst the variability of renewable energy sources such as wind and solar is well known, what is less understood is that the variability occurs over a wide range of timescales, including quite short timescales. The longer timescale variability (longer than 1 h) is associated with the passage of weather systems and daily cycles and the shorter timescale variability (shorter than 1 h) is associated with the turbulence or gustiness of the wind. The longer timescale variability in wind power can be mitigated by moving energy from periods of high generation to times of lower generation. The shorter timescale variability in wind power can cause instabilities in the power system.

Figure 1 shows the consequences of longer timescale variability in an examination of generation sources in the German grid from a week in Summer 2012 (Burger 2012). Clearly, the contribution of wind and solar power to satisfying the load is continuously varying. Longer-term storage, such as pumped hydro can play a significant role at these timescales.

Fig. 1

PV and Wind contribution to power generation in Germany, June 25th to July 1st 2012 (Yellow—solar, green—wind, grey—conventional). Burger (2012)

Figure 2 shows short-term variability in wind power and the predicted consequences. It shows some modelling results, calculated in 2004 by NEMMCO, predicting flows in the main interconnector between the states of South Australia and Victoria, based on simulations of wind power production from wind speed records. With 400 MW of wind power installed in South Australia (the approximate installed capacity at the time) there is little difference from the no-wind power case, but with a projected 1,000 MW installed the wind power causes short-term violation of the connector flow limits. Fast storage systems could smooth this excess variability. The alternative solutions are to curtail the wind farm output or reduce the interconnector flow. As of October 2011, South Australia has 1,200 MW of installed wind power.

Fig. 2

Effect of increasing wind power in South Australia on Heywood interconnector flow—model results (12 h sample—blue trace is background flow, pink trace is with indicated installed wind power)

Rapid fluctuations in generation can cause an increased demand in regulation power, which is used to compensate between predicted demand and actual demand. This is shown in Fig. 3, which shows the difference between the predicted demand (the smooth blue curve) and the actual demand. Normally, the generation is scheduled against the predicted demand and the difference (the red curve) is supplied by separate regulation services. Wind power is generally treated as negative demand and most systems with significant contributions from wind power have wind forecasting systems, which calculate the likely contribution from wind. However, these forecasting systems do not predict the rapid fluctuation component which remains as an error, which needs to be corrected through regulation services.

Fig. 3

Definition of regulation services (Kirby 2004), blue trace is predicted demand, green trace is actual demand and the red trace is the difference of the two—the regulation power required

Studies by the New York Independent System Operator (ISO), predict a significant rise in the regulation power requirement with increasing wind power on their system (Fig. 4). The Californian ISO also projected an increase in regulation power requirement with increasing intermittent renewable energy on their system. They predicted a rise from a frequency regulation requirement of 1 % of peak load dispatch (approx. 350 MW) to 2 % regulation as renewables rose to 20 % contribution in 2010 and to 4 % regulation required (1,400 MW) as the contribution rises to 33 % by 2020. Regulation services are mainly supplied by peak generating plants with high emission levels or by modulating the output of base load generating plant. This is an ideal application for fast storage systems and we will see later that several technologies are already being trialled in this role.

Fig. 4

Regulation requirement versus installed wind power, New York ISO

In a report on the role of energy storage with renewable electricity generation, NREL (Denholm et al. 2010) concluded that high penetration of variable generation increases the need for all flexibility options including storage. They also concluded that it creates market opportunities for these technologies; however, storage has been difficult to sell into the market because of the challenges it has in quantifying the value of its services. Indeed the role which storage systems can play in the future power grids is very diverse. A recent study by Sandia Labs (Eyer and Corey 2010) lists 17 applications in five major categories such as Electric Supply, Ancillary (regulation) Services, Grid System, End User/Utility Customer and Renewables Integration (Table ES 1 from Sandia report) shows the potential value of the application, the likely market in the USA in a 10-year period and the timescale of the application. There is a very wide range in each case.

Figure 5 shows a similar view from a report by EPRI (Rastler 2010) which orders the storage applications by value and indicates the size of the market. For some high value applications, Transmission and Distribution (T&D) system support and area frequency regulation (part of regulation services), some technologies are already economic (See Table ES-4 and Figure ES-14 in the EPRI report). One problem identified is that many technologies are still only at the large-scale demonstration phase and need more experience to earn the trust of Utilities and other customers.

Fig. 5

Estimated target market size and target value analysis (Rastler 2010)

The EPRI report also shows the range of storage applications in a diagrammatic form (Fig. 6). This representation is intended to show that there is often an overlap between applications in when size and timescale are considered. In fact, appropriately designed storage systems could provide multiple roles in a single system—achieving so-called “benefit stacking” and multiplying the possible monetary value which could be achieved. This may be essential to the initial viability of storage solutions.

Fig. 6

EPRI analysis of storage value (Rastler 2010)

However, it is acknowledged that currently storage value is difficult to extract when its operation may be simultaneously spread over a number of service areas. While it may be possible to calculate the benefits in areas where there are established markets or where there established markets—e.g. retail energy trading in time-shifted wind and solar energy (arbitrage) and regulation services, it is more difficult to access the value where there is strong regulation such as grid infrastructure services and upgrade deferral. Indeed it becomes even more complex when devices may fill both roles simultaneously.

Regulatory incentives are starting to appear. The US Federal Energy Regulatory Commission (FERC) issued a ruling on pay for performance in October 2011, which gives extra payment for faster ramping services, benefitting fast storage-based services. The California State Energy Storage Bill AB2514, March 2010 now requires electrical corporations and locally owned utilities to create energy storage systems in their distribution networks to either reduce emissions of greenhouse gases, reduce demand for peak electrical generation, or improve the reliable operation of the electrical transmission or distribution grid. This law mandates storage equal to 2.25 % of daytime peak power by 2014 and 5 % of daytime peak power by 2020.

2 The Technologies

Rastler (2010) (Table ES-4) lists a number of technology options and costs (in today’s terms) when applied to applications in bulk energy storage, fast frequency regulation, renewables integration and grid support. The list of technologies which are deemed to be commercial is quite short with many more in the demonstration phase of development. Some examples are listed below.

For bulk energy storage (several hours duration) pumped hydro, underground compressed air, sodium sulphur and advanced lead-acid is deemed to be commercial. More innovative pumped storage systems include custom built units such as at Taum Sauk in Missouri—a 450 MW system first installed in 1963 (Fig. 7) and the Okinawa Yanbaru Seawater Pumped Storage System. Constructed in 1999, this system utilises a cliff-top reservoir adjacent to the ocean to give a system capable of a throughput of 31 MW with approximately 400 MWhr of storage.

Fig. 7

Taum sauk 450 MW custom pumped hydro storage system in Missouri

Compressed air storage systems use reversible turbines to compress or expand air which is stored in depleted underground hydrocarbon reservoirs, which can provide many hours of storage. There have been two long established systems at Huntdorf, Germany (290 MW peak power, established 1978) and McIntosh, Alabama (110 MW peak power, established 1991). An example of commercial electrical storage is the Sodium-sulphur (NaS) system from NGK from Japan which is rated at 1.2 MW/7.2 MWhr (6 h of storage), ideally suited to substation upgrade deferral.

Of the faster response systems, A123 has demonstrated a modular Li-ion system rated at 2 MW/0.5 MWhr (15 min storage). This multi-purpose system can address a range of applications such as regulation services. Beacon Power has constructed a 20 MW/5 MWhr (15-min storage) flywheel demonstration in New York State designed for regulation services (Fig. 8).

Fig. 8

Beacon power 20 MW/5 MWhr (15 min storage) flywheel demonstration in New York State

East Penn Manufacturing Co. through its subsidiary Ecoult has constructed a 3 MW/3 MWh (1 h storage) system utilising UltraBatteries (modified lead-acid) for regulation services. This technology will be discussed further below. Ecoult has also supplied storage systems to the Prosperity Solar Energy Storage Project, New Mexico (Fig. 9), simultaneously providing voltage smoothing and peak shifting of power from the 500 kW PV plant. The 500 kW/500 kWh (1 h) smoothing system uses the East Penn UltraBattery and the 250 kW/1,000 kWh (4 h) shifting system uses a more conventional high-performance lead-acid battery (East Penn Unigy II).

Fig. 9

Prosperity solar energy storage project, New Mexico featuring both voltage smoothing and peak shifting

3 A Case Study: Wind Farm Smoothing

This case study describes the energy storage trial implemented at Hampton Wind Farm in NSW, Australia. The objective of the trial at Hampton is to smooth the ramp rate of the wind farm before presenting it to the grid. In turn the impact objective is to achieve higher penetration of wind and renewable energy in grid systems. While the Hampton system smooths the energy produced “at the source” on the wind farm, it is an objective of the work that the system and learning are transferable wherever the benefit of reducing renewable energy variability exists, for example at grid nodes (or substations) or via the provision of ancillary services generally.

The implementation a wind smoothing system followed a path of progression from laboratory trials, through the attachment of larger scale systems to a Vestas V47-660 kW wind turbine (Fig. 10). Stage 1, commissioned in mid 2010, consisted of a custom built system rated at 144 kW/240 kWh, and which featured four battery banks of different lead-acid battery types, including prototype UltraBatteries from Furukawa. Stage 2, commissioned in mid-2011, consisted of commercial modular building blocks rated at 1 MW (capped 660 kW)/500 kWh utilising East Penn UltraBatteries (Fig. 11).

Fig. 10

1 MW UltraBattery storage system at Hampton Wind Farm

Fig. 11

UltaBattery cell bank at Hampton Wind Farm

The UltraBattery is a hybrid energy storage device that integrates a supercapacitor with a lead-acid battery in one unit cell, without the need for extra electronic control (Fig. 12). This unique design, harnessing the best of both technologies, produces a battery which can provide high power discharge and charge with a long, low-cost life (Lam and Louey 2006).

Fig. 12

The principle of UltraBattery technology

Developed in Australia by the CSIRO Energy Transformed Flagship research program, the UltraBattery already serves applications for use in hybrid electric vehicles (HEVs) with further variants aimed at resolving issues of intermittency in capturing energy produced from renewable sources. It is manufactured in various forms by Furukawa in Japan and East Penn in the USA and is available in production quantities. Testing by Sandia National Laboratories Single Cell Testing under a regulation services profile has shown it to have several times the life of conventional lead-acid batteries (Hund et al. 2012).

Initial results from the first stage of the system show that with a simple proportional-integral (PI), fixed-parameter algorithm, significant reductions in rates of change of power output (ramp rates) can be achieved. Figure 13 shows results from 1 day with a variety of wind conditions. The lower traces show the raw turbine input and the smoothed output when combined with the storage system. The upper traces show the reduction in 5-min ramp rate which averages a factor of 7. The 1-min ramp rate reduction achieved by the system is a factor of 10.

Fig. 13

Smoothing of Wind output and ramp rate reduction with fixed-parameter controller algorithm

4 A More Advanced Algorithm

A more advanced algorithm system is now being developed. The system, shown in Fig. 14, works as an adaptive scheme which allows the smoothing parameters to be continuously changed. An offline optimising scheme is used to design the functions used in real-time. The optimisation takes into account a number of objectives (goals) and costs while being aware of system electrical and physical constraints. The system can be re-optimised for each installation.

Fig. 14

Schematic of Wind smoothing algorithm development system

The offline learning mechanism uses a quantum particle swarm optimisation algorithm. This is a meta-heuristic search method that copes well with highly non-linear objective functions and lends itself to parallelisation. This allows the optimiser to use of all 32 cores on a CSIRO multicore-computing server. A 21-day training data set composed of a combination of challenging periods is used to evaluate the quality of adaptive parameter settings.

The learning engine is used to create functions for the execution engine. This engine generates PI variables in real-time based on current conditions and derived from the learning that are in turn provided to the storage controller and used to generate charge and discharge commands to the batteries.

5 Advanced Algorithm Results

Initial trials show a significant improvement is possible using this approach. Figure 15 shows the result of a simulated comparison where the standard fixed parameter PI algorithm is run against the same wind data sample as the new adaptive algorithm. In this case, the adaptive mechanism has achieved a result where it has achieved a superior reduction in the 5-min ramp rate over the standard fixed parameter algorithm and significantly reduced the high frequency noise in the signal. As a result of the training, the adaptive system recognised that wind conditions had been comparatively low for a period and anticipated the storm front that moved through by moving the state of charge of the energy store lower to where it had headroom to react more efficiently, reducing sudden output changes.

Fig. 15

Comparison of standard fixed parameter PI algorithm and adaptive PI algorithm

The architecture approach with the adaptive mechanism allows for many variables to be considered during the learning and for performance to be optimised against a number of targets. For example, minimisation of battery use has now been combined into the parameters optimised by the Hampton algorithm and CSIRO testing has shown that using the methodology the amount of energy passed through the energy store to achieve the ramp rate objectives can be significantly reduced while maintaining system performance. Further developments will test simultaneous energy shifting and smoothing by utilising the continuously variable control parameters. There is also potential to translate this simultaneous approach to solar PV.

6 Conclusions

Proven and emerging storage technologies have a wide variety of roles to play on the electricity grid, particularly in integrating renewable energy. The large range of storage timescales and capacities required give opportunities for all technologies, including performing multiple functions, simultaneously. The ability to extract value is currently limited and further regulatory changes are needed before the full potential of storage can be realised.