Energy Storage TED Talk

Hello Readers,

If you have been following my blog, you may remember that I wrote my graduate dissertation on the GHG emissions impact of adding energy storage technology to the electric utility grid. Of course, at the time I wrote it, there were only a few energy storage technologies that were close to meeting the technological and economic thresholds that would be practical in the real world. Instead, I had to do my calculations on the premise that those thresholds could be reached.

Thus, I was ecstatic to discover a TED Talk by Donald Sadoway, a professor at MIT, about a technological breakthrough in utility-scale energy storage. Please enjoy the TED Talk below.

Cheers,

Sean

Hello Readers,

When I saw this new TED talk about literally growing batteries, I knew I had to share it. This is especially relevant to our ability to manufacture batteries and other energy storage systems (as discussed in my dissertation).

Please enjoy!

~Sean

Dissertation: Discussion

V.    Discussion

This section will start by discussing the implications of the results of this study, will offer some general comments on the concept of adding ancillary ES to the electricity grid, will reflect upon changes that could have improved this study and possibilities for future studies, and will conclude with some final thoughts on the impact of adding ES to the electricity grid will have on attempts to reduce anthropogenic climate change.

The comparison between BL and S1 allows this study to touch base with the reality of ES’s utility as a means of reducing the actual GHG emissions of the contemporary electricity grid in order to impact the causes of anthropogenic climate change. As the results show simply adding bulk ES to the contemporary electric utility grid in the model region would likely contribute to a relatively small impact on GHG emissions. Thus, while adding bulk ES to the utility grid may have a number of other tangible benefits, it will not be a sufficient step in an effort to reduce the impacts of anthropogenic climate change. However, S2, S3, and S4 indicate bulk ES can certainly play a part in such efforts.

The discovery that the net GHG emissions of BL and S1 are virtually identical is of particular interest because it detracts from the notion that the additional GHG emissions caused by inefficiencies in traditional fossil-fuel plants can be overcome through the use of bulk ES. Instead, overcoming such inefficiencies may rely solely upon the use of ancillary ES technologies (as discussed below) or other interventions. However, with the development of a model with a more detailed and sophisticated incorporation of generating inefficiencies (e.g. one that simulates generation on a plant by plant basis) might be needed to definitively conclude the matter.

The results of this study show that it is technically feasible to reduce overall and per unit GHG emissions and meet consumer electricity demand. In particular S2, S3, and S4 show 11%, 83%, and 55% respective reductions in GHG emissions compared to BL and S1. Since each of these scenarios can attribute a non-negligible portion of their GHG emissions to bulk ES capacity, the calculated reductions are likely to be less than those of any comparable study that did not consider the need for bulk ES.

The range of emissions rates in S2, S3, and S4 (with S3 the ‘nuclear future’ scenario being the lowest of them all) suggests that a move towards grid development policies and plans that focus on nuclear-powered generation (in combination with the implementation of new bulk ES technologies) could have the greatest impact on anthropogenic climate change. While this seems indisputable from the results, governments and regulatory authorities would need to take into account the other concerns surrounding nuclear-powered generators before pursuing S3 as a particular course of action.

It should also be noted that the GHG emissions values associated with the renewable generation capacity in the model is based on the values for on-shore wind turbines, which were considered the most viable type of turbine in the model region. According to Weisser (2007) off-shore wind turbines may have a lower GHG emission rate (i.e. 9-19 kgCO₂e/MWh) than that of on-shore turbines (i.e. 8-30 kgCO₂e/MWh). Therefore, the comparison of S3 and S4 may not be entirely straightforward for regions with significant coastal resources. Also, with regard to shorter-term planning (i.e. over the next two decades) decision makers will need to consider whether it is more technically and politically feasible to quickly set up wind and solar farms or nuclear power plants. Similarly, for longer-term planning (i.e. beyond 2030) consideration will need to be taken with regard to the limited nature of nuclear fuel sources and the ultimate disposal of spent fuel; whereas with intermittent renewable resources, neither of these issues are a concern.

From another perspective, the results show that the amount of installed bulk ES necessary to make S2, S3, and S4 viable paints a picture that also favors the nuclear future option represented by S3. The results of S2 show the need for about 36.7 GW of power releasing capability over the course of an hour, and S4 would require the installation of enough ES to release about 57.9 GW of power during an hour.  Meanwhile, S3 (with significantly lower emissions) requires only about 28.7 GW of power absorbing capability from ES per hour. Although, it should be noted that some of the ES requirements could be met with ancillary ES capacity – especially in the most extreme conditions, none of these figures are insignificant in comparison to the overall generation requirements of the system (i.e. nearly 120 GW during the weekly peak).

Since S4 has more than triple the installed renewable generation capacity of S2 but does not require triple the amount of installed ES capability, it can be concluded there is likely to be a diminishing requirement for ES at higher percentages of intermittent generation capacity. However, since S2 and S3 have comparable amounts of intermittent generation capacity, the fact that S2 has only about 72 GW of installed non-intermittent capacity where S3 has about 80 GW of such capacity indicates (not surprisingly) the amount of ES required is inversely proportionate to the amount of non-intermittent generation capacity. However, it is not entirely clear from these results which (if either) factor has a greater impact. Likewise, predicting the correct amount of required ES may be a complex algorithm that accounts for both factors as well as their relation to electricity demand.

Since this model amalgamates the generic use of bulk ES, it only calculates the so-called direct GHG emissions associated with ES. Other GHG emissions (e.g. those associated with construction, maintenance, and demolition) would be highly dependent upon the specific types of bulk ES technology that are installed. Based on Weisser’s (2007) suggestion that LCA GHG emissions could vary by an order of magnitude, the importance of understanding the specific amount of ES required to be installed during any particular grid development strategy becomes more important if attempts are made to include full LCA GHG emission information. Thus, for conscientious planners the difference between needing to install 28.7 GW as opposed to 57.9 GW of ES maybe a pivotal factor in choosing a strategy aimed at decarbonizing an electric utility grid that reaches beyond simple financial considerations.

In the interest of ensuring that this study was applicable to real world situations, the amount of generation capacity by fuel type was closely matched with that of a the PJM Interconnection in the USA (see section III.2. for details). Thus, in order to measure the applicability of this study, the rate of GHG emissions simulated in the model region can be compared to the official rate of GHG emissions declared by the PJM for 2009. According to PJM (2009) the annual average rate of GHG emissions for 2009 was 1,137 lbs. CO₂e/MWh (or 517 kgCO₂e/MWh). Furthermore, the range for the average rate of GHG emissions varied monthly in 2009 from a rate of 484-561 kgCO₂e/MWh. Curiously, this matches up with the lower end of the range of the simulated rate of emissions in BL, which was 557 kgCO₂e/MWh. Of course, the PJM rates are well below the upper end of the range of the simulated rate of emissions in BL, which was 761 kgCO₂e/MWh.

This discrepancy could be related to a number of factors. For instance, it could indicate that PJM has relatively new (and efficient) generation technologies compared to the studies discussed by Weisser (2007). Of course, another explanation is that the cumulative effect of all of the subtle differences between the model region and the real PJM territory (e.g. the difference between the actual consumption patterns and the sinusoidal simulation) have resulted in an over estimation of GHG emissions. However, it seems likely that the discrepancy indicates the figures PJM used to calculate the results displayed in their report do not include full LCA GHG emissions but rather simple direct GHG emissions, which would explain a bias in which the figures in this study should be higher than those in the PJM report. Thus, even if the final figures are not interchangeable on a 1 to 1 basis with the PJM report, the real world results (i.e. the total actual GHG emissions released into the atmosphere) associated with operating the PJM electric utility grid as a whole should follow a trend similar to the results of this study if bulk ES technologies are installed and the installed generation capacity shifts in a manner similar to one of the three scenarios S2, S3, or S4.

As mentioned in section II.2., the installation of ancillary ES (or the use of plug-in EVs) could hypothetically reduce or eliminate the need for traditional generation plants to be operated in spinning reserve or frequency regulation modes. In so doing the actual GHG emissions associated with S2, S3, and S4 (especially those of S2 and S4 which have higher percentages of fossil fuel powered plants in operation) would be lowered below those shown in the results. Furthermore, it is likely that ancillary ES, which tends to have higher roundtrip cycle efficiencies than bulk ES technologies, would not have as many direct GHG emissions associated with its use. The one caveat to such a suggestion being that the use of ancillary ES, which tends to have a much more rapid self-discharge rate than bulk ES technologies, would need to be carefully calibrated and coordinated to ensure that unnecessary losses are not regularly experienced on the system.

More specifically – assuming that the transportation system in the USA continues to rely heavily upon the use of personal cars and that the current research, development, and deployment trends for plug-in EVs of the past few years continue – the use of plug-in EVs seems at least superficially favorable compared to typical ancillary ES from perspective of reducing the overall GHG emissions of the USA. Even though the current direct GHG emissions from the electric utility grid may be higher than the current direct emissions from petroleum-fueled vehicles. The opportunity to symbiotically eliminate direct GHG emissions from the transportation sector, reduce additional GHG emissions caused by inefficient electricity generation during ancillary generation, and creating a grid that is better prepared to cope with increased intermittent renewable generation capacity appears to be a win-win-win scenario. However, in order to fully validate this notion a comprehensive LCA study of the GHG emissions (as opposed to a simple direct emission comparison) associated with plug-in EVs must be performed.

Before making any final conclusions, it is prudent to reflect upon the possibility of redoing, improving, or building upon the findings of this study. First and foremost, the attempt to survey utilities and ES manufacturers, which was not fruitful, could have been completely omitted. It was initially attempted to secure some original data. However, even the most professional and tenacious attempts to encourage participants to respond apparently could not overcome the barriers to a high response rate (e.g. a lack of prestige or name recognition, a short timeframe, and a lack of direct incentives). In retrospect, even a 100% response rate would likely not have provided any more insight into the current state of ES technological development (let alone that of the near future) than the supplementary sources upon which this study relied. One suggestion for future studies with similar goals would be to skip the formal survey process and rely more heavily upon academic articles and an investigative strategy (i.e. searching for publically available information such as users manuals or notes from intra-industry conferences). Employing such a strategy would have saved several weeks of time spent on survey creation, distribution, and collection. This time would have been spent more effectively on model development and running an increased number of trial runs.

Specifically, an increased number of trial runs would have allowed a better understanding of how incrementally increasing and decreasing various types of generation technologies and/or incrementally improving the roundtrip cycle efficiency of ES technologies impacts GHG emissions. Also, observing the results pertaining to the percentage of the net GHG emissions attributable to the use of bulk ES technology begs the question: is there a maximum percentage of the GHG emissions that can be attributed to the use of ES technology? An initial hypothesis is that the potential attributable percentage would prove to display an asymptotic nature related to the roundtrip cycle efficiency of ES (i.e. the attributable percentage would not exceed the percentage of roundtrip losses incurred by the ES technology).

Additionally, a more sophisticated model that is based on the aggregated electricity generation of hundreds or thousands of individual plants would have allowed for a better understanding of the nature of spinning reserve and frequency regulation inefficiencies in traditional plants. Thus, a greater level of confidence could be assumed about similarities in the results showing that BL and S1 are so similar. However, to produce such a model the computing power of Microsoft Excel would likely not be sufficient, so another software base for the model would be required.

Finally, based on all of the background research and the results of the computer simulation performed in this study, it has been concluded that the inclusion of bulk and ancillary energy storage on electric utility grids over the next two decades would benefit attempts to mitigate the causes of anthropogenic climate change by reducing the overall greenhouse gas emissions released by electric utility grid operators. However, such an effort to include energy storage on an electric utility grid will only result in significant benefits if accompanied by a shift in generation capacity away from fossil fuels and toward nuclear and/or renewable generating capacity. In so doing, electric utility operators will be able to mitigate the causes of anthropogenic climate change without adversely impacting the habits of contemporary electricity consumers in developed countries.

Previous Post: Results

Table of Contents - References

Dissertation: Results

IV.    Results

The net GHG Emissions (measured in kgCO₂e) resulting for each scenario are shown in Figure 1, and the rate of net GHG emissions compared to the weekly electricity demand that is met (measured in kgCO₂e/MWh_demand) are shown in Figure 2. The results show that emissions and the rates of emissions are virtually identical in BL and S1. The scenarios (listed in order of descending emissions and rates of emissions) are S1, S2, S4, and S3.

Figure 1: This figure displays the net greenhouse gas emissions (in millions of kgCO₂ equivalent) of a simulated electric utility grid during a one-week model run for a baseline scenario (BL), which does not incorporate bulk energy storage technology, and four other scenarios (S1, S2, S3, S4), which incorporate bulk energy storage technologies.

Figure 2: This figure displays the rate of net greenhouse gas emissions (in kgCO₂ equivalent per MWh of demand met) of a simulated electric utility grid during a one-week model run for a baseline scenario (BL), which does not incorporate bulk energy storage technology, and four other scenarios (S1, S2, S3, S4), which incorporate bulk energy storage technologies.

Interestingly, in BL and S1approximately 3.1% and 2.8% of the net GHG emissions respectively could have been avoided through the use of ES (or transferred to a later period of time in the case of S1). In S2, S3, and S4 approximately 10.2%, 1.1%, and 22.8% of the net GHG emissions respectively are attributable to the use of ES technology. See Table 4 for some of the pertinent output data associated with each of the scenarios.

Table 4: This table lists some of the output data from one-week trial runs of each scenario that is referred to throughout the study. For a complete list of the output data please see Appendix F.

	BL	S1	S2	S3	S4
Net GHG Emissions (Lower Bound)	8,554,897,158 kgCO2e	8,554,886,391 kgCO2e	7,616,839,503 kgCO2e	1,471,886,100 kgCO2e	3,825,223,442 kgCO2e
Net GHG Emissions (Upper Bound)	11,700,791,870 kgCO2e	11,699,165,045 kgCO2e	10,329,233,555 kgCO2e	2,307,909,136 kgCO2e	5,483,357,498 kgCO2e
GHG Emissions Rate (Lower Bound)	557 kgCO2e/ MWh	557 kgCO2e/ MWh	496 kgCO2e/ MWh	96 kgCO2e/ MWh	249 kgCO2e/ MWh
GHG Emissions Rate (Upper Bound)	761  kgCO2e/ MWh	761 kgCO2e/ MWh	672 kgCO2e/ MWh	150 kgCO2e/ MWh	357 kgCO2e/ MWh
Maximum ES Storage	N/A	13,617 MW/hour	17,489 MW/hour	23,489 MW/hour	12,849 MW/hour
Maximum ES Release	N/A	3,183 MW/hour	36,659 MW/hour	28,659 MW/hour	57,935 MW/hour
GHG Emissions Associated with ES	3.1%*	2.8%	10.2%	1.1%	22.8%

*Denotes a potential for GHG emissions to be avoided.

Previous Post: Methodology (8 of 8) - Other Limitations (2)

Next Post: Discussion

Table of Contents - References

Dissertation: Methodology (8 of 8) - Other Limitations (2)

With the previous to caveats being understood, the weekly load factor, which is the total electricity generated during the week divided by the amount of energy that would have been generated if electricity demand was constantly at the peak load for the week, was calculated to be nearly 0.77 during all scenarios. As a comparison to the real world, the simulated weekly load factor was lower than the PJM annual load factor of 0.58 during 2009. This indicates that the model results are representative of a week that is less variable than the year as a whole. Also as an error-checking mechanism, the consistency of the load factor results suggests that in all cases electricity generation was meeting demand in the model. A deviation from that value would mean that demand was not being met.

As noted in section III.3.a., in all scenarios this study assume that the average roundtrip cycle efficiency of all bulk ES technologies in use on the electricity grid is 75%. However, this value was simply chosen due to the fact that several ES technologies are purported to have a range of roundtrip efficiencies that include 75%. Since it is unlikely that any particular technology will monopolize the ES market over the period of this study, this was a safe value to assume. However, the extent to which these technologies improve (or the extent to which these technologies fail to match their promised efficiencies) during the studied time period will be the extent to which the results of this study are an over-(or under-)estimation. Thus, the results of this study should be understood as a guideline rather than a strict prediction of the future.

Previous Post: Methodology (7 of 8) - Other Limitations (1)

Next Post: Results

Table of Contents - References

Dissertation: Methodology (7 of 8) - Other Limitations (1)

D.    Other Limitations

This section highlights some caveats to the results of this study that are not explicitly stated elsewhere in this paper. These caveats are points either within the model or the assumptions made during this study that may cause the results to imperfectly align with those found in analogous situations in the real world.

While efforts were made to approximate the shape and levels of consumer demand for electricity (see section III.3.b. and Equation 1), these approximations do not directly match up with the actual diurnal and weekly fluctuations in demand during any period of time. However, it should also be noted that no two weeks in the real world will have demand curves that precisely align with one another. Thus, readers must be cautioned that there is no reason to expect that any specific week in a year will yield the exact results of this study.

This study also does not attempt to account for annual (a.k.a. seasonal) fluctuations in consumer electricity demand. While other studies have attempted to do this, it did not seem prudent in this study since all of the scenarios (except BL) represent times at some point in the future that will likely have consumer demands that differ from those of today. Thus, as this study may yield results that could be considered representative of a week in July 2010. The results may just as easily be representative of a week in December 2029 due to relatively unpredictable increases (or decreases) in consumer demand. Therefore, this study does not attempt to claim that these results are particular to any time of year, and it cannot be determined with certainty that the results represent an upper- or lower-bound of the impact of ES during any particular year in the future.

Previous Post: Methodology (6 of 8) - Parameters

Next Post: Methodology (8 of 8) - Other Limitations (2)

Table of Contents - References

Dissertation: Methodology (6 of 8) - Parameters

C.    Parameters

It was outside of the purview of this study to perform a full LCA of all of the types of generation technology. However, Weisser (2007) reviews numerous LCA studies that analyze the GHG emissions associated with the various types of electricity generation simulated in this study. The review offers a range of GHG emissions for each type of generation based on different LCA techniques applied to different generation sites throughout the developed world, and all of the values included in the review are directly attributable to a particular, recent study with original data (Weisser 2007). As a spot check on the Weisser (2007) article, the results of the Jaramillo et al (2007) article, which performed a comprehensive analysis of the LCA GHG emissions associated energy generated from combusting coal and natural gas (including synthetic and liquid natural gas) in North America, were also taken into consideration. Upon translating the results of both articles into similar units, it was found that the results were relatively similar. Please see Appendix D for a list of the original LCA results in both articles.

Table 3: This table displays the greenhouse gas emissions input parameters for all generation fuel types used in the model in all scenarios. Note that all values are from Weisser (2007).

Generation Fuel Type	Lower Bound Emissions	Upper Bound Emissions
Nuclear (LWR)	2.8 kgCO2e/MWh	24 kgCO2e/MWh
Coal	950 kgCO2e/MWh	1250 kgCO2e/MWh
Natural Gas	440 kgCO2e/MWh	780 kgCO2e/MWh
Oil	500 kgCO2e/MWh	1200 kgCO2e/MWh
Hydro (without ES)	1 kgCO2e/MWh	34 kgCO2e/MWh
Wind (On-Shore)	8 kgCO2e/MWh	30 kgCO2e/MWh
Solar (All PV types)	43 kgCO2e/MWh	73 kgCO2e/MWh

Thus, this study uses the values suggested by Weisser (2007) as the guiding parameters for GHG emissions rates (see Table 3). As such, all scenarios were run using the upper- and lower- bound of the ranges suggested for each generation fuel type. This safely allows an assumption to be made that the ‘true’ GHG emissions for each scenario would likely lie in between the values generated in the upper- and lower-bound trial runs. Unfortunately, due to the discrepancies inherent in LCA analysis, it is not possible to be any more certain about the results produced by the model.

It should be noted that the values for nuclear generation displayed in table three are only derived from studies about so-called ‘light water reactors’ (or LWR), which are currently a common type of nuclear reactor in use (Weisser 2007); however, as the grid is modernized newer types of reactors may be brought into service, so these values can only serve as a guideline. It should also be noted that the range of GHG emissions for solar generation include both types of photovoltaic panels (e.g. mono- and poly-crystalline) and that monocrystalline panels have a range of 43-62 kgCO₂e/MWh whereas polycrystalline panels have a range of 50-73 kgCO₂e/MWh (Weisser 2007).

Weisser (2007) also offers some LCA GHG emission values associated with some types of ES technologies; however, since this study is based on determining the direct GHG emissions (i.e. a portion of the full LCA value) associated with ES technologies, it did not seem prudent to attempt to integrate these values into the model. These omitted values can be found in Appendix D.

Previous Post: Methodology (5 of 8) - Formulae (2)
Next Post: Methodology (7 of 8) - Other Limitations (1)
Table of Contents - References

Dissertation: Methodology (5 of 8) - Formulae (2)

  and otherwise (Equation 5)

In the case of simulating solar-powered electricity generation, there were two levels of potential obstruction (i.e. clouds and lack of daylight). In the model, the user is able to manipulate both of these by selecting the percentage of cloudiness and by adjusting sunset and sunrise. Initially, the model checks to see if a cloud is covering the solar farm. If so, zero electricity is generated. If not, the model is dependent upon the time of day as described below:

Before noon: , and after noon: .

Before sunrise:   and after sunrise: .

Before sunset: and after sunset: .

Essentially, this generates an equation where the solar farm generates electricity in a sinusoidal pattern – based on the time of day t_day compared to sunrise T_sunrise and the length of daylight L_daylight (as determined by the time difference between sunrise and sunset) – during the day and nothing at night.

Unlike intermittent renewable generation sources, traditional generation sources are operated in a manner designed to meet consumer demand for electricity. This greatly simplifies the equations required to simulate them. However, traditional generation sources can also be run in different modes, which can result in different levels of generation from different types of plants. The two primary modes of plant operation are ‘base load’ and ‘peak load’. As the name implies base load plants form the base of generation needs and run more or less constantly. Whereas, peak load plants are only called into operation during times of peak demand. Since the model is organized based on the fuel type of the generation capacity rather than on individual plants, a third mode of operation was created, which was called ‘mixed load’ and denoted that some plants of the fuel type were operating in base load while others were operating in peak load.

The user interface made use of drop-down field forms. Thus, the actual equations required a significant amount of logical coding in order to match each type of generation with the correct mode of operation. In summary: a generation type in base load constantly operated at its nominal capacity, a generation type in peak load only operated if all other generation options were exhausted, and a generation type operating in mixed load operated only to fill the gap between the base load generation and consumer demand. Also, if multiple types of generation were operating in mixed or peak load, the amount of generation required was distributed evenly among those types but did not exceed the nominal capacity of any particular type.

The user interface offered a choice between having traditional generation account for renewable generation or not. In BL traditional generation needed to take into consideration the electricity generated by renewable sources or else electricity generation did not match demand. In all other scenarios ES rather than traditional generation moderated the renewable generation.

Due to the distinction in GHG emissions, spinning reserve generation was delineated from normal operation. Only generation types operating in mixed load were called into service to operate in spinning reserve. Further, the total generation (i.e. the combination of normal and spinning reserve generation) of each type never exceeded the nominal capacity of that type. It should be noted that this did mean that at least one type of traditional generation needed to be in mixed load for the model to operate properly; however, this is not inconsistent with the real world.

The outputs of the model can be delineated into two groups. The first group of outputs is the diagnostic outputs, which were useful for comparing the model results to the real world and indicating potential errors in the model (see section III.3.d.). The second group of outputs is those highlighted in the results section (e.g. the net GHG emissions and the GHG emissions attributed to ES). The final paragraphs in this section will briefly describe the manner in which the second group of values was calculated.

GHG emissions were calculated in essentially two parts. First, the GHG emissions associated directly with electricity generation was calculated by attributing emissions to each MWh of electricity generation segregated by each type of generation based on the GHG emission rates indicated on the user interface. Also, any spinning reserve generation would have an increased emissions rate. The user interface allowed the proportionate increase in emissions created by fossil-fuel powered plants in spinning reserve to be chosen (non-fossil-fuel powered plants are exempt from this increase because their GHG emissions are associated with construction and decommissioning of the plant rather than plant operation; see section II.2.a. for further explanation). For all scenarios, the value of 0.5 (or a 50% increase in emissions) was applied to plants in spinning reserve mode. Finally, all of these values were summed over all generation types during all intervals to determine the ‘non-ES-related emissions’ and divided by the total electricity generation (measured in MWh) to determine the average rate of emissions for the electricity grid.

Second, the GHG emissions associated with the use of ES were determined by comparing the energy put into storage with the energy withdrawn from storage over the entire week. After accounting for energy losses from the roundtrip cycle efficiency, the model determined how much excess stored energy remained or how much additional energy needed to be produced to make use of the ES. Then, this value was multiplied by the average rate of emissions for the electricity grid as described in the previous paragraph. A positive value of ES-related emissions would indicate that additional emissions (and electricity generation) would be necessary to maintain the use of ES. On the other hand, a negative value would indicate either that electricity generation could have been reduced during the week or depending on the type of ES technology in use that the excess energy could be saved for use during a later time period. Thus, the net GHG emissions are simply the non-ES emissions plus the ES emissions.

Previous Post: Methodology (4 of 8) - Formulae (1)

Next Post: Methodology (6 of 8) - Parameters

Table of Contents - References

Dissertation: Methodology (4 of 8) - Formulae (1)

B.    Formulae

This section outlines the formulae used to calculate the inputs (i.e. electricity demand, renewable generation patterns, traditional generation modes, and spinning reserve) and the model outputs (i.e. net GHG emissions, and GHG emissions attributed to ES). Please note that several of the equations used in this model make use of the random number generator in Microsoft Excel, which produces numbers between -1 and 1. In order to simplify the equations listed below the variable R_x will be used to denote any randomly generated number; however, the use of multiple R_x variables in one equation should be understood to denote multiple random numbers.

Based on the assumptions about electricity demand listed in the section above, the simulated electricity demand D was created using a combination of diurnal D_daily and weekly D_weekly fluctuations as well as fluctuations that represent a frequency shift D_shift similar to that described by Kempton et al (2008). Please see Equation 1 for a detailed view of demand simulation.

  (Equation 1)

Where:

;

with t being the 15-minute time interval;

on weekends or on weekdays;

And with M_day being the maximum demand for the day without including D_shift.

Interestingly, it was discovered that the random number generator created a set of numbers that were not evenly distributed about zero. Whereas in most cases this was not necessarily an issue, in the case of frequency shift, which Kempton et al (2008) explains should contain roughly as much demand for regulation up as regulation down, there was a need for a relatively even distribution. Thus, the bias, which was calculated to be proportionate to 234 for the particular set of random numbers in the model, was subtracted from the random numbers to provide an even distribution in the frequency shift.

By configuring D_curve as it is in the equation above, the daily demand fluctuations reach a minimum at the interval representing 5:00-5:15 a.m. and a maximum at the interval representing 5:00-5:15 p.m. This also means that daily demand fluctuations are about halfway up and halfway down at the interval of 11:00-11:15 a.m. and p.m. respectively. Meanwhile, D_daily is set up to offer a range of variability of up to 10% with regard to the daily maxima and minima on weekdays theoretically centered around 95% of what the inputs for maximum and minimum demands for the week (D_max and D_min respectively) are set to on the model’s user interface. Finally, D_weekly leaves the weekdays unchanged and reduces the demand on weekdays by approximately 10%, which is roughly consistent with the assumptions made in the previous section.

The amount of renewable energy generated on the simulated utility grid is essentially a function of two factors primary: the amount of generating capacity installed and the amount of natural energy (wind or solar rays) available to the system at any particular time. Østergaard (2008) indicates that for renewable energy it is important to consider the effects of the geographic aggregation of all renewable resources on a system. In order to approximate the nuances of a geographically dispersed system, the available renewable generating capacity (Wind_CAP and Solar_CAP) selected on the model’s user interface was divided evenly among 12 different energy ‘farms’ for both wind and solar. Equations 2 and 3 show that the total wind and solar generation (G_w and G_s respectively) are simply the sum of the electricity generated at each of the 12 farms.

  (Equation 2)

(Equation 3)

This segmentation allows for a level of overall wind and solar generation that is more stable than the generation at any particular farm. The individual wind and solar farms (represented as G_wi and G_si respectively) are each respectively governed by Equations 4 and 5 below.

Or

(Equation 4)

Where:

  and  ;

And or .

In the case of simulating wind-powered electricity generation, the user interface of the model allows the user to choose the average availability factor A_f (the proportion of the time that the wind turbines are available for electricity generation; i.e. when the turbines are not broken or under maintenance) and the average capacity factor C_f (the amount of electricity that is typically generated compared to the nominal capacity of the turbines) of the wind turbines connected to the grid.

The choice between G_wA and G_wB simply ensures that the model does not allow a wind farm to generate more than its nominal capacity during any particular interval. While G_wB allows for a random variation of electricity generation centered on the average amount of generation determined by the C_f, A_f, and Wind_CAP settings on the user interface. Meanwhile G_wC means that approximately 5% of the time the wind farm will generate no electricity, which represents intervals with a wind speed below the minimum speed required to turn the blades on the turbines.

Previous Post: Methodology (3 of 8) - Modeling
Next Post: Methodology (5 of 8) - Formulae (2)
Table of Contents - References

Dissertation: Methodology (3 of 8) - Modeling

3.   Modeling

In addition to the general need to use a simulation tool, Voorspools and D’haeseleer (2000) also highlight the need for “instantaneous” rather than “linear” emission approximations (i.e. not using daily or annual average figures) to accurately determine GHG emissions under varying demand-supply scenarios (Voorspools and D’haeseleer 2000). Despite the need for a sufficiently resolved time-scale, the duration need not be exceptionally long. For example, Verhaegh et al (2010) elected to focus on one-week periods during different seasons (e.g. winter and summer), which suggests that this is a reasonable approach to avoid simulating an entire year’s worth of data.

This study included developing a model in Microsoft Excel, which allowed for the inclusion of a user-friendly interface as well as the ability to make alterations to the model as new types of data were included to increase the accuracy of the model. As a drawback, the Excel-based model could only handle a limited set of calculations before overwhelming the available computing facilities. In order to keep the size of the model within the limitations of the computing facilities available, this study allowed for a time resolution of 15 minute intervals but restricted the simulated time frame to a one-week period, which allowed for sufficient temporal resolution (i.e. as great or greater than that of models described in articles listed in the Reference section) and the inclusion of lower demand periods during the weekend.

In order to develop the model some basic assumptions (see section III.3.a.) were made based on the data gathered regarding ES technologies and electricity generating technologies and electricity consumption patterns in the model region. In some cases these assumptions were used to develop specific formulae that guided the manner in which electricity demand and generation and storage were represented (see section III.3.b.). In other cases these assumptions provided the parameters with which GHG emissions factors and other variables were determined (see section III.3.c.). Naturally, using such assumptions (and the modeling process in general) necessarily limits the accuracy and applicability of the results to real-life purposes (see section III.3.d.).

a.    Assumptions

The first major assumption necessary to build an accurate grid simulation is that the simulated demand for electricity is consistent with actual electricity demand curves. A more accurate demand curve allows for a better understanding of the potential for ES deployment. Officially, PJM recognizes that its ‘on-peak period is all non-holiday weekdays from 7 a.m. until 11 pm and the off-peak is comprised of all other hours’ (PJM 2009). Thus, roughly speaking the demand should be relatively higher between 7 a.m. and 11 p.m. on weekdays than at other times. To simplify matters, this study assumes there all weekdays are non-holidays.

In order to determine the appropriate shape of the diurnal fluctuations and the relative difference between demand on a weekend and on a weekday, historical real-time data was observed on the PJM website (i.e. http://edata2007.pjm.com/eData/index.html). The data observed showed that during the week demand fluctuated between ~70 GW at night and ~120 GW during the day, and the curve appeared roughly sinusoidal reaching about ~90 GW at around roughly 8:30 a.m. and 11:00 p.m. each day. On the weekends, the demand curve showed a roughly similar shape during the night and day; however, the range of the demand was ~62 GW to ~110 GW.

It must be noted that the observation period was not exhaustive (covering only a few days in July). Thus, it is possible that the days observed were not especially ‘normal’; however, this use of real data means the example is not outside the realm of possibility. Furthermore, the observation period does not conflict with any of the other data or reports presented by PJM. As such, the observed data was used as a basis for simulating electric demand (see section III.3.b. for the exact formula used).

The second major assumption made in this study is that over the next two decades the use of bulk ES in the PJM territory is more likely to include a plethora of ES technologies rather than a single type or brand of ES. Therefore, it did not seem prudent to specifically restrict the model to match the exact characteristics of any particular type of ES. Instead, the model was set up to measure the potential for ES use as an output rather than an input. However, to make any useful calculations regarding the impact of ES on GHG emissions it is necessary to make an assumption about the average roundtrip cycle efficiency (i.e. the ratio of usable electricity retrieved divided by the electricity consumed during the storage process) of all the bulk ES in use on the grid.

Rather than speculating about the various proportions of the bulk ES technologies that may be installed over the next two decades, it was simply assumed that for all scenarios the average roundtrip cycle efficiency was 75%. This ratio was chosen because it fell safely within the range of roundtrip cycle efficiencies of all of the bulk ES technologies described in section II.2.a.

Previous Post: Methodology (2 of 8) - Model Region and Scenario Selection
Next Post: Methodology (4 of 8) - Formulae (1)
Table of Contents - References