There are several areas where the modeling of battery systems and the compliance rules pertaining to their use can be improved.
First, the specification of a battery system used for compliance must be well defined and enforceable in the software. Such specifications must cover the options available to define both the physical battery equipment and the battery system control algorithms.
Second, there are a handful of known simplifications and approximations in CSE, the implications of which should probably be better understood. These approximations gloss over real-world details of battery operation including 1) how performance characteristics change with environmental conditions and with charge/discharge rates 2) how battery performance degrades with age and use, and 3) how the measured battery value is influenced by the time-resolution of the simulation.
Finally, exercising the models across a sample of building types, configurations, and climates will help stakeholders fully understand how including residential battery systems in compliance analysis affects incentives, decision-making, and the larger power system.
If modeled batteries are to be permitted EDR credit, it is important for CBECC-Res to limit simulations to realistic battery systems. It is important that the battery systems modeled in CBECC-Res reliably and accurately represent real battery systems made available to the public. There are three generally accepted approaches to managing user inputs to this end in CBECC-Res:
1. Users may directly input parameters (e.g., capacity, efficiency) that must be verified by the building inspector provided such information is available on a product nameplate. Often these inputs have imposed limits to restrict users from providing less sensible inputs to “game” the compliance analysis. For example, as mentioned above, it is possible at present to set both charge and discharge efficiencies to 1.0 and model a fictional lossless battery. A parameter-checking rubric would limit a modeler’s ability to do such.
2. Users select generic types of battery products that lock in values for model input parameters. These types of products are selected to represent the full range of products available. An example of this approach elsewhere in CBECC-Res is the PV module type dropdown where users can select between generic “Standard” and “Premium” modules.
3. Users select a specific brand and model of battery product from a library of certified/rated products. The model input parameters are pre-defined according to third-party testing and locked in. The model number in the compliance report is then verified by the building inspector. An example of this approach elsewhere in CBECC-Res is Heat Pump Water Heaters where users can select the brand and model number of a NEEA (Northwest Energy Efficiency Alliance) rated product.
Currently, CBECC-Res largely follows the first approach with some of the CSE model input parameters being determined by CBECC-Res rulesets. However, the current approach for selecting a control algorithm follows more of the second approach. Some mixture of these three approaches is generally acceptable.
There is a current need for direct communication with battery manufacturers to determine what levels of inputs they are comfortable with users providing to represent their products. As far as we are aware, there are no official rating standards to determine the efficiency (let alone any thermal impacts on the efficiency) of commercially available residential battery systems. This is an area that requires more attention, and it is recommended that manufacturers be engaged directly to gain insight into any efforts from the industry to standardize battery system performance ratings.
The CBECC-Res software allows the modeler to choose among the four control strategy algorithms described above for compliance runs. The strategies can return a range of EDR credits because of how aggressively each targets TDV savings. The Basic strategy, maximizing self-utilization of on-site generation, is generally understood to be how residential battery system control behaves today. The Tesla Powerwall’s “Self-powered” mode is a Basic strategy (Tesla, Undated).
The control strategies in the CBECC-Res software all assume full cycling of the available capacity. This is in contrast to the user setting in modern consumer battery systems for specifying a fraction of the available capacity to be used solely as backup power. If enabled, the battery only cycles through the top portion of the battery capacity in normal use. CBECC-Res does not currently enable specifying a backup fraction. Doing so would reduce the amount of load-shifting the battery could accomplish and so would reduce the EDR savings from the battery. CSE does not simulate power outages, so it would not use the reserved backup fraction.
It is reasonable to think that the next generation of battery systems will have more sophisticated strategies, designed to reduce the residence's peak load (minimize demand charge), or to shift load away from high time-of-use periods. Some of those strategies might resemble the Time-of-Use algorithm in CBECC-Res today. The two “Advanced Time-Based Control” modes described for the Powerwall’s modes of operation support page do enable time-shifting to account for peak periods, although neither one matches up with the strategies available in CBECC-Res. One of the differences is that neither of the Powerwall’s modes enables full-power discharge to the grid: even at peak, the battery only serves the building’s load.
Looking forward more, distributed batteries may--within a few years--have communication protocols to receive pricing information and/or charge/discharge requests from a service provider or regulatory body. With even a small amount of live information transfer, more sophisticated strategies that include forecasting and optimization become possible.
Whichever control strategy is used for compliance analysis must be able to be verified by a building inspector. Clarifying which of the existing control strategies should qualify for compliance simulations in 2019, and how they will be verified could be an important step for CBECC-Res, as might reviewing what behaviors any next-generation strategies should allow.
In its present configuration, CSE treats charge efficiency, discharge efficiency, and available capacity all as user-input constants. A model operator would likely input the published battery system specifications for efficiency and capacity, leading to optimistic model results.
In real-world operation, charge and discharge efficiencies decline from the ideal levels under a range of conditions including high or low ambient temperatures, high charge/discharge rates, and depth of discharge. For example, the Tesla Powerwall 2 Datasheet specifies the efficiency and capacity with the following footnote: “Values provided for 25°C (77°F), 3.3 kW charge/discharge power.” and notes that ideal operating temperatures are “0°C to 30°C (32°F to 86°F)” (Tesla, 2018).
Regarding cycling depth, as the battery approaches full charge, the terminal voltage increases, requiring more power to charge. On discharge, the terminal voltage declines along with the charge level, so a given current extracts less and less power (DiOrio et al., 2015). Colloquially, the last handful of ions are harder to insert into the anode on charging and also don’t come back out as easily on discharge. That means cycling efficiency varies inversely with discharge level.
As implied for the Powerwall, capacity is also variable with external conditions. This is especially true of cold temperatures as has been discussed in the context of electric vehicles having reduced range in cold conditions (see, for example, Smart, 2015).
To mitigate temperature dependence, some battery systems include onboard thermal management systems: pumps and/or fans to maintain optimal operating temperatures. Such systems can keep the efficiencies and capacities close to optimal conditions, but they use electricity to do so. Furthermore, those auxiliary loads can vary dramatically for different technologies and applications.
A clear path to improving the fidelity of CSE would be to incorporate parameterizations of performance curves for either general or specific residential battery systems. The project would aim for a thorough and holistic treatment of variable operational performance, touching how each of capacity, efficiency, and power are impacted by conditions and characteristics of use. These performance curves would want to have options for systems both with and without active thermal management, such that an actively managed system would be less sensitive to ambient conditions but would experience a parasitic load under certain conditions.
Another part of this development project would be to provide CSE with options for battery systems to be assigned to zones of the residence. The thermal conditions a battery system experiences will vary substantially depending on whether the battery is installed in a conditioned space, garage, basement, or outdoors. The inefficiencies in the charging and discharging process will manifest as heat generated in the ambient environment, impacting the thermal loads.
Battery storage capacity declines over time and with use. Numerous factors can impact the aging process. Factors known to accelerate degradation include deep charging cycles, operation at high or low temperatures, and fast charging/discharging (Smith et al., 2017). Conversely, strategies to extend the lifetime of a battery system would include some combination of: preferring shallow cycles to deep discharge, maintaining the battery at moderate temperatures, and charging/discharging at well below maximum power.
Commercially-available residential battery systems are equipped with battery management systems to orchestrate state-of-charge and charge/discharge rates and thermal management systems to maintain healthy battery operating temperatures. Those systems help avoid the extreme situations that are big hits to long-term performance.
On one hand, even with these systems in place, battery system lifetime performance is likely to vary from one installation to another depending on how the battery is used, the ambient conditions it experiences, and other factors. On the other hand, as of 2018, commercially available residential battery systems typically come with warranties that guarantee a fraction of initial capacity remains available at a given cumulative throughput or number of cycles.
The Tesla Powerwall 2 warranty, for example, states that 70% of the initial 13.5 kWh usable capacity will remain after a) ten years of unlimited cycles if the system is used for PV self-consumption or b) ten years or 37.8 MWh of aggregate throughput if used for other applications (Tesla, 2017). Referencing the section above about how CBECC-Res sets the available capacity, those figures allow an argument that the 15% aging derate applied in CBECC-Res should be applied to the initial useful capacity rather than to the total capacity.
For the questions being considered in CBECC-Res simulations, a static, lifetime-average assumption for capacity degradation is a reasonable approach. While research projects that investigate novel--and especially aggressive--battery use strategies would want to consider how a given battery strategy impact degradation, the CBECC-Res 2019 simulations use pre-defined operational strategies comfortably within the intended use of today’s battery storage systems. If there becomes interest in using CBECC-Res or CSE for analyses that include battery aging and degradation dynamics, that would be motivation to add a battery-aging module to CBECC-Res.
CBECC-Res runs CSE simulations at 3-minute timesteps. However, solar incidence, loads, PV production, and battery charge and discharge decisions currently have hourly resolution in CSE.
A recent study evaluated how timestep resolution affects the value residential batteries provide (Abdullah et al., 2017). The researchers gathered 1-minute irradiance and load data, populated a model with a PV system, a battery, and a self-consumption-maximizing battery strategy (the Basic strategy in CBECC-Res). They ran simulations at 1-min resolution and also aggregated the input data to 2, 5, 10, 30, and 60 minutes and ran simulations at those coarser resolutions. They performed the set of simulations across a number of different household load profiles.
When the researchers
compared the value the battery provided at the various resolution levels, they
found that value (defined as cost savings in a time-of-use market)
declined as resolution decreased. In the histogram from the study reproduced in
Figure 6, the 60-minute simulations have markedly lower cost savings than the
1-minute simulations but also have a wide spread.
The cause of the reduced value at coarser timesteps can be explained with a basic example (paraphrased from Abdullah et al., 2017): For a 30-minute interval during which a residence’s demand and PV output match (on average over the interval), a simulation carried out on 30-minute resolution data would determine that the battery would not be operated, and so would not deliver value during that interval. However, if during this interval PV output was fluctuating significantly (for example due to passing clouds, or repeated inverter drop-outs as a result of local voltage rise) a simulation at a higher temporal resolution would result in the battery being charged and discharged repeatedly and delivering some value to the household.
The simulations in the Abdullah et al. study resemble the PV-battery-load-grid interactions in CBECC-Res from which the TDV-based savings are calculated. Therefore, it is likely that the CBECC-Res estimates of battery-related TDV reduction are underestimates. We can investigate addressing this underestimation either through an indirect approach (e.g., a simple post-process multiplier on the annual battery TDV) or by explicitly simulating variations at sub-hourly frequencies. Accounting for this effect explicitly in CSE would require a substantial effort to obtain or derive higher frequency solar and load data.
This work plan describes the tasks necessary for us to exercise models across a sample of building types, configurations, and climates. Analysis of the results will help stakeholders fully understand how including residential battery systems in compliance analysis affects incentives, decision-making, and the larger power system.
The analysis will cover several dimensions of battery system performance in California, including:
•Climate zone,
•Type of residential construction (defined by prototype buildings representing both single family and multifamily buildings)
•Relative sizes of PV systems and batteries (including runs with PV but no battery),
•Different battery properties (e.g., efficiency, maximum charge rate), and
•Different battery control strategies
The result from this task will be a matrix of proposed simulation runs representing good coverage of the overall problem space. The matrix will describe the range and values used for each of these dimensions.
We will define the base prototype models used for the analysis. These will likely be based on the standard prototype models developed by the CBECC-Res team. Additional prototypes may be added depending on the needs of the study defined in Task 1. Each prototype model will be identical to the standard (code-minimum) model except for the addition of the PV and battery systems.
In previous work, we have automated CBECC-Res parametric analysis using Big Ladder’s Modelkit/Params parametric framework. Big Ladder has created a spreadsheet interface to this framework that can be used to define all permutations of within the parametric space. The resulting compliance margins relative to the standard model for each permutation will be automatically populated in a single results spreadsheet. Further analysis of these results will help answer the question of how much credit batteries will receive in the current version of CBECC-Res 2019. Subsequent analyses can easily be performed following any changes to the CBECC-Res software or to explore other scenarios that would affect the compliance margins related to battery systems.
This analysis will give a sense of the magnitude of the compliance credit related to battery usage. It can be the basis for future economic evaluations of batteries in contrast to other energy efficient, renewable energy, and/or demand response technologies.
Abdulla, Khalid, Kent Steer, Andrew Wirth, Julian de Hoog, and Saman Halgamuge. “The importance of temporal resolution in evaluating residential energy storage.” In 2017 IEEE PES General Meeting (to appear), pp. 16-20. 2017.
DiOrio, Nicholas, Aron Dobos, Steven Janzou, Austin Nelson, and Blake Lundstrom. “Technoeconomic modeling of battery energy storage in SAM.” No. NREL/TP-6A20-64641. National Renewable Energy Lab. (NREL), Golden, CO (United States), 2015.
Energy and Environmental Economics, Inc. “TDV PV+Storage Tool.” Undated.
Ferris, Todd, Larry Froess, Dee Anne Ross. “2019 Residential Alternative Calculation Method Reference Manual; For the 2019 Building Energy Efficiency Standards.” California Energy Commission. Expected July 2018.
Ming, Zachary, et al. “Time Dependent Valuation of Energy for Developing Building Efficiency Standards: 2019 Time Dependent Valuation (TDV) Data Sources and Inputs.” 2017. Accessed April 16, 2018. http://docketpublic.energy.ca.gov/PublicDocuments/16-BSTD-06/TN216062_20170216T113300_2019_TDV_Methodology_Report_21517.pdf.
Smart, John. “Advanced Vehicle Testing Activity Cold Weather On-road Testing of the Chevrolet Volt.” No. INL/EXT--14-34030. Idaho National Laboratory (INL), Idaho Falls, ID (United States), 2015. https://inldigitallibrary.inl.gov/sites/sti/sti/6372861.pdf.
Smith, Kandler, Aron Saxon, Matthew Keyser, Blake Lundstrom, Ziwei Cao, and Albert Roc. “Life prediction model for grid-connected Li-ion battery energy storage system.” In American Control Conference (ACC), 2017, pp. 4062-4068. IEEE, 2017.
Tesla, Inc. “TESLA POWERWALL LIMITED WARRANTY (USA): April 19, 2017.” 2017. Accessed April 16, 2018. https://www.tesla.com/sites/default/files/pdfs/powerwall/powerwall_2_ac_warranty_us_1-4.pdf.
Tesla, Inc. “Tesla Powerwall 2 AC Datasheet.” https://www.tesla.com/sites/default/files/pdfs/powerwall/Powerwall%202_AC_Datasheet_en_northamerica.pdf. January 30, 2018. Accessed May 4, 2018.
Tesla, Inc “Support: Powerwall Modes of Operation with Solar” https://www.tesla.com/support/energy/own/powerwall/modes-of-operationwithsolar.html. Undated. Accessed May, 16, 2018.
Wilcox, Bruce. “CBECC-Res 2019.0.4 Research Version – June 15, 2017.” 2017. Accessed April 16, 2018. http://www.bwilcox.com/BEES/docs/CBECC-Res%202019-0-4_Quick_Tour.pdf.