Flywheels Texas University-gcep Program.pdf

Sizing Advanced Flywheel Energy Storage Clay S. Hearn *, Michael C. Lewis, and Dr. Robert E. Hebner Center for Electromechanics The University of Texas at Austin www.utexas.edu/research/cem August, 2012

Introduction The Center for Electromechanics (CEM) at the University of Texas at Austin is working under a program sponsored by Stanford University through Global Climate & Energy Project (GCEP) to develop the next generation of flywheel energy storage for the grid. The overall goal of this program is to develop technologies to store half of the grid’s energy in 50 years. CEM believes advanced flywheel energy storage can be a significant part of the solution. Given that nearly all electricity in the world is produced from generators and about 60% of the world’s power is used in motors, it makes sense to look at rotational energy storage. Flywheels store energy mechanically by spinning high strength composite rotors at high speeds. The primary limiting factor to energy storage potential for flywheels is due to the maximum allowable material stress that current materials can achieve. The theoretical maximum energy density for potential flywheel materials is a simple ratio between maximum allowable material stress and density. Current high strength graphite composites can achieve up to 600 kJ per kg (or 170 Wh/kg). One key to developing flywheels with higher energy densities is to develop lighter materials with increased strength. Another key aspect of flywheel energy storage, which separates it from other devices such as batteries or ultracapacitors, is that the rated power capability is not directly tied to energy storage capacity. Energy transfer, to charge and discharge a flywheel, is provided by motor-generators. Therefore, energy storage capacity and power capability can be tailored to meet specific grid requirements. Flywheels can also be implemented in modular fashions to meet net storage and power requirements.

*

Contact info - email:

[email protected], //www.utexas.edu/research/cem/

1

The University of Texas at Austin 1 University Station Austin, TX, USA 78712

The Center for Electromechanics 10100 Burnet Rd., EME 133 Austin, TX, USA 78758

Several past flywheels from CEM are noted in Figure 1, which was adapted from the Electricity Storage Association. Figure 1 compares power and discharge time at rated power of various storage technologies. Typically flywheels are considered to be pulse power devices which compete well with ultracapacitors or high power lithium ion batteries. However, CEM has also developed flywheels which can provide longer term energy storage needs. The first example is shown as item number 3, which was a 130 kWh flywheel to provide energy storage for an advanced hybrid locomotive train. Second, item number 4 was a flywheel system developed for NASA that was designed to provide energy storage to the International Space Station. This latter system would charge and discharge over the 90 minute day-night cycle that the station experiences in orbit.

No.

4

3

Project

1

CHPS

2

ATB Bus

3

ALPS Locomotive

4

NASA FESS System

2 1

Figure 1.

Comparisons of energy storage technologies and CEM flywheels (1)

This document will present a methodology for evaluating energy storage requirements at different levels in the grid. By using real world home usage data provided by Pecan Street, Inc., flywheel energy storage was sized at the home, transformer, and community level of the Mueller development in Austin TX. The presented technique utilizes parametric sweeps through an optimal control law to develop sizing curves which show trade-offs between energy storage size and power requirements.

2

The University of Texas at Austin 1 University Station Austin, TX, USA 78712

The Center for Electromechanics 10100 Burnet Rd., EME 133 Austin, TX, USA 78758

Sizing with an Optimal Controller For the utility grid, storage will act as a buffer between the grid supply and the load demand. This buffer gives the system an extra degree of freedom and level of controllability to shift either load demand, or power generation, which did not exist before. The total demand can be represented as a combination of electricity loads and renewable generation, Figure 2.

Figure 2.

Expressed control loop for flywheel energy storage

To properly size energy storage for a given load demand, or power generation source, a controller should be selected which will determine the real time grid power requirements to maintain the stored energy, 𝑄𝑓𝑤 , or state-of-charge of the storage device. For flywheel energy storage, the change in stored energy with respect to time,

𝑑𝑄𝑓𝑤 𝑑𝑡

, will equal the grid power into the flywheel, 𝑃𝑔 , minus the

load demand, 𝐷𝐿 , and minus losses which may come from windage or bearings. For flywheels, the losses can be estimated by using a linear time constant, 𝜏𝑓𝑤 , as shown in [eqn. 1]. 3

The University of Texas at Austin 1 University Station Austin, TX, USA 78712

The Center for Electromechanics 10100 Burnet Rd., EME 133 Austin, TX, USA 78758

There are many different options available for designing and selecting an appropriate controller, which can range from simple heuristic control laws to more complicated control techniques. 𝑑𝑄𝑓𝑤 1 = 𝑃𝑔 − 𝐷𝐿 − 𝑄 𝑑𝑡 𝜏𝑓𝑤 𝑓𝑤

[eqn. 1]

One possible control technique is to select an Optimal Controller which will minimize a given cost function for a known load demand profile. This technique provides a consistent controller which is easy to implement and solve for a known load profile. The cost function in [eqn. 2] is used for the minimization routine. The first term in the cost function is an end constraint which requires the flywheel stored energy at the end of the simulation, 𝑄fw (T) to be equal to the initial amount of flywheel stored energy, 𝑄0 . The integral portion of the cost function seeks to minimize the sum of the grid power, 𝑃𝑔 , and the deviation of flywheel stored energy, 𝑄𝑓𝑤 (t), from the initial stored energy, 𝑄0 , subject to weighting factors 𝑎 and 𝑏. Given the cost function in [eqn. 2], parameter sweeps of weighting variables, 𝑎 and 𝑏 can be performed to assess trade-offs between energy storage capacity, flywheel power capability, and end grid power performance. 𝐽(𝑡0 ) =

1 1 2 2 𝑆𝑞 �𝑄𝑓𝑤 (𝑇) − 𝑄0 � + � �𝑎�𝑄𝑓𝑤 (𝑡) − 𝑄0 � + 𝑏𝑃𝑔2 (𝑡) � 𝑑𝑡 2 2

[eqn. 2]

Flywheel Sizing Analysis To understand how location affects energy storage sizing, home and solar generation data provided by Pecan Street, Inc. was used to study flywheel sizing at different levels throughout the grid. Pecan Street is an active R&D effort in Austin TX to implement and evaluate smart grid technology in the new Mueller Development. There are 741 homes in this development and 25% of these homes have solar installations, which translates into 1 MW of generation capacity. A thorough analysis was performed which studied locating flywheel energy storage throughout different locations in the Mueller community. These locations include flywheels at the individual home level, at the local transformer level (supports approximately eight homes), and the community level (all 741 homes). Sample load data for an individual home, transformer level, and the entire community are shown in Figure 3,Figure 4, and Figure 5, respectively. The data corresponds to an arbitrary day in the summer of 2011. These plots show the net power consumption throughout the day (red), the solar power generated (green), and the net grid power required (black) which is the difference between the net power consumed and the power generated by the PV arrays. 4

The University of Texas at Austin 1 University Station Austin, TX, USA 78712

The Center for Electromechanics 10100 Burnet Rd., EME 133 Austin, TX, USA 78758

At the individual home, power usage is characterized by high pulse loading of the air-conditioning system, as shown in Figure 3. The load profiles at the home level vary immensely due to individual customer use and preferences.

Figure 3.

Example power profile for individual home (2)

At the transformer and community level, Figure 4 and Figure 5 respectively, aggregation of the individual home profiles starts to replicate the traditional grid energy use profile, with peak loads occurring in the evening. Further aggregation and smoothing of the power profile is achieved at the community level for all 741 homes, with a peak power usage of 2.5 MW.

Figure 4.

Power profile for transformer (2)

5

The University of Texas at Austin 1 University Station Austin, TX, USA 78712

Figure 5.

The Center for Electromechanics 10100 Burnet Rd., EME 133 Austin, TX, USA 78758

Power profile for 741 home community (2)

Results of Sizing at Single Residence An initial analysis evaluated flywheel energy sizing for 13 individual homes in the Mueller community. An optimal controller was defined to minimize the cost function presented earlier in [eqn. 2]. A sweep of weighting values, 𝑎 and 𝑏, in the cost function was performed to generate sizing curves which show tradeoffs between flywheel storage capacity and peak power from the grid. This procedure is performed by holding the weight value of 𝑏 equal to one and increasing the value of 𝑎, starting from zero. At 𝑎 = 0, the flywheel will be sized for diurnal energy storage to average the net home load throughout the day. As 𝑎 is increased in value, the flywheel energy storage size is reduced, and transitions to smoothing and peak shaving the net load demand. Tradeoffs between flywheel size and decrease in peak grid power consumption are shown in Figure 6 assuming a flywheel with a time constant of 50 hours. For each individual sizing curve, the end point on the right represents diurnal energy storage requirements to average the total load over the 24-hour period. Flywheels would need to deliver between 10 and 25 kWh of energy, with power levels up to 9 kW.

6

The University of Texas at Austin 1 University Station Austin, TX, USA 78712

The Center for Electromechanics 10100 Burnet Rd., EME 133 Austin, TX, USA 78758

The curves in Figure 6 show that for 2.0 kWh of delivered energy, the flywheel can reduce the peak grid power load between 30 to 60% for the majority of the home cases. The flywheel peak power for the power smoothing case is 6.5 kW.

Flywheel Energy Storage Sizing for Individual Homes Flywheel Energy Delivered [kWh]

25

Home 1412 Home 1420 Home 1421 Home 1422 Home 1423 Home 1424 Home 1425 Home 1426 Home 1427 Home 1439 Home 1446 Home 1458 Home 1463

20 15 10 5 0 0%

20%

40%

60%

80%

100%

Percent Decrease in Peak Grid Power Figure 6.

Flywheel sizing curves for individual homes

An example of the diurnal power smoothing is shown in Figure 7 for data from Home 1424. For this case, a flywheel with 10.5 kWh delivered energy and 6 kW power capability is required. The optimal grid power covers the average net load demand and flywheel losses. The power fluctuations throughout the day are handled by the flywheel. For the same home data, a smaller flywheel will provide a net power smoothing capability as shown in Figure 8. This flywheel delivers a net energy of 2.4 kWh and power capability of 6.2 kW. Again, this smaller flywheel still handles the daily power fluctuations, but the smaller size equates to lower losses for an equivalent time constant. This reduced size flywheel saves 2.6 kWh of energy consumption over the entire day compared to the larger flywheel for diurnal energy storage.

7

The University of Texas at Austin 1 University Station Austin, TX, USA 78712

Figure 7.

The Center for Electromechanics 10100 Burnet Rd., EME 133 Austin, TX, USA 78758

Flywheel energy storage sized for diurnal load averaging of an individual home (2)

Figure 8.

Flywheel energy storage sized for power smoothing of individual home (2)

Results of Sizing at Transformer and Community A similar study was performed for sizing flywheel energy storage at the transformer level, which services eight homes on average, and for the entire 741 home community. This analysis also investigated the impact of flywheel losses on grid performance by evaluating loss time constants of 50 and 200 hours. Flywheel sizing curves for the local transformer level are shown in Figure 9 for loss time constants of 50 and 200 hours. The first point on the far left of the curves represents peak grid power and energy storage requirements for the 24-hour diurnal energy storage case. 8

The University of Texas at Austin 1 University Station Austin, TX, USA 78712

The Center for Electromechanics 10100 Burnet Rd., EME 133 Austin, TX, USA 78758

The impact of the higher losses can be clearly viewed in this case with a 24% higher grid load power requirement to cover the losses. As energy storage size is decreased, the performance difference with the lower loss flywheel becomes less pronounced.

FW Delivered Energy [kWh]

Transformer Energy Storage vs. Peak Grid Power 100 10 1

Time Constant 200 Hrs Time Constant 50 Hrs

0.1 10.0

15.0

Figure 9.

20.0 25.0 Peak Grid Power [kW]

30.0

Sizing of flywheel energy storage at the transformer

For daily power averaging at the transformer location, the flywheel must deliver 70 kWh of energy with a power capability of 16.3 kW. For the flywheel with a 200-hour time constant, this results in a 52% decrease in peak grid power requirement with a 2% increase in net grid energy. For power smoothing operating, a flywheel with 6 kWh delivered energy and 12 kW power capability could reduce peak grid power by 19%, as shown in Figure 10. The increase grid energy requirement for this case falls to 0.14% for the 200-hour flywheel loss and 0.6% for the 50-hour flywheel loss.

Figure 10.

Flywheel energy storage sized for power smoothing transformer loads (2)

9

The University of Texas at Austin 1 University Station Austin, TX, USA 78712

The Center for Electromechanics 10100 Burnet Rd., EME 133 Austin, TX, USA 78758

Similar results are achieved with flywheel sizing for the community level of 741 homes. The sizing curves for the entire community are shown in Figure 11. For diurnal energy storage of the entire community, a flywheel system with 5.9 MWh of delivered energy and power capability of 1200 kW is required. A reduced-size flywheel for power smoothing would need to deliver 450 kWh with a power rating of 300 kW, as indicated in Figure 12. Compared to the transformer and home locations, the natural aggregation of the community reduces the power-to-energy storage ratio of the flywheel for power smoothing. The power smoothing flywheel reduced peak grid power by 8%.

FW Delivered Energy [kWh]

Community Energy Storage vs. Peak Power 10000 1000 100 Time Constant 200 Hrs

10

Time Constant 50 Hrs

1 1000 Figure 11.

Figure 12.

1500 2000 2500 Peak Grid Power [kW]

3000

Sizing of flywheel energy storage for 741 home community

Flywheel energy storage for smoothing power at the community level

10

The University of Texas at Austin 1 University Station Austin, TX, USA 78712

The Center for Electromechanics 10100 Burnet Rd., EME 133 Austin, TX, USA 78758

A summary of the flywheel sizing options is shown in Table 1 for different levels of storage throughout the community. As the flywheel is placed at higher levels throughout the community, the effective C-rate, or ratio between power and energy, decreases for both diurnal storage and power smoothing operations. For diurnal storage, the flywheel should be designed between 0.5 to 0.2 C-rate, whereas for power smoothing, the flywheel should be capable of 3 to 1 C-rate. Table 1.

Location Single Home Transformer (8 homes) Community (741 homes)

Peak Demand Power 5.7 KW (average)

Summary of storage results

Diurnal Storage

Power Smoothing

15 kWh / 6 kW

2 kWh / 6.2 kW

30.7 kW

70 kWh / 16.3 kW

6 kWh /12 kW

2500 kW

5.9 MWh /1200 kW

450 kWh / 300 kW

Another observation from Table 1 is that the aggregation of flywheels located in the homes would oversize the energy storage needs at the transformer by 71% and 150% for diurnal and power smoothing, respectively. A reduced amount of oversizing, 10% and 23%, is observed between the transformer and community level, respectively.

Summary This sizing analysis shows how flywheel energy storage can be sized for grid level energy storage considering a wide variety of applications. This methodology applies an optimal control law to the load profile and flywheel model. The advantage of this method is that it removes uncertainties due to selection of the energy storage controller, and important dynamics, such as spinning loss rates, can be included by a time constant in the analysis model. A parametric sweep of weighting factors within the cost function can be performed to view tradeoffs between energy storage size, power capability, and resultant grid power consumption in the form of sizing curves. Based on data provided by Pecan Street, Inc. for an active smart grid in Austin, TX, flywheel energy storage was sized for different levels in the community. For diurnal energy storage, this analysis suggests flywheels should have power to energy storage ratios of 0.5 to 0.2. For power smoothing, higher power flywheels with power to energy storage ratios of 1 to 3 would be required. Overall, as the flywheel energy storage is moved to higher levels in the grid, the power to energy storage ratio decreases for all applications.

11

The University of Texas at Austin 1 University Station Austin, TX, USA 78712

The Center for Electromechanics 10100 Burnet Rd., EME 133 Austin, TX, USA 78758

A key point in this study is the impact of spinning losses on flywheel sizing and performance at long-term energy storage levels. Large flywheel capacities will require reduction in frictional spinning losses to be viable options for grid energy storage. Development of low loss bearing using high temperature superconductors may be a key technology to achieving this goal. For power smoothing and short-term storage applications, reduced size flywheels are less sensitive to spinning losses. The high cycle life and power requirements for these applications make flywheels an attractive option.

References 1. Electricity Storage Association. Storage Technologies: Technology Comparison. Electricity Storage Association Web Site. [Online] [Cited: August 20, 2012.] http://www.electricitystorage.org/technology/storage_technologies/technology_comparison. 2. Pecan Street Inc. Use, Grid and Solar Data: Pecan Street, Inc. Pecan Street Inc. Web Site. [Online] [Cited: August 20, 2012.] http://www.pecanstreet.org.

12