SGIP and TE30 Examples

TE30 Example

Fig. 22 shows a reduced-order demonstration model that incorporates all three federated co-simulators; GridLAB-D simulating 30 houses, EnergyPlus simulating one large building, and PYPOWER or MATPOWER simulating the bulk system. This model can simulate two days of real time in several minutes of computer time, which is an advantage for demonstrations and early testing of new code. There aren’t enough market participants or diverse loads to produce realistic results at scale. Even so, this model is the recommended starting point for TESP.

_images/TE30system.png

Fig. 22 Demonstration model with 30 houses and a school

The three-phase unresponsive load comes from a GridLAB-D player file on each phase, connected to the feeder primary. The EnergyPlus load connects through a three-phase padmount transformer, while the houses connect through single-phase transformers, ten per phase. Except for transformers, all of the line impedances in this model are negligible. One of the house loads has been shown in more detail. It includes a responsive electric cooling load, lights, and several non-responsive appliances. In addition, each house has a solar panel connected through an inverter to the same meter, which might or might not implement net metering. Storage, vehicle chargers and other appliances (e.g. electric water heater) could be added. For now, each house is assumed to have gas heat and gas water heater.

SGIP Use Cases

TESP will initially respond to four of the Smart Grid Interoperability Panel (SGIP) use cases [12] and an additional use case to illustrate the growth model.

SGIP-1 and SGIP-6. “The grid is severely strained in capacity and requires additional load shedding/shifting or storage resources” [12]. The details confirm that this use case addresses only generation capacity constraints of the type that might be needed after existing demand-response resources become exhausted.

This use case clearly takes place on a day that available resources are inadequate in a warm location like California or Arizona. In the base-case scenario, the system anticipates the event that morning or even earlier. Contracted demand-response resources—predominantly distributed generator sets―are scheduled to actuate during the day at the predicted time of the peak load. While helpful, the demand response proves inadequate. Therefore, each distribution utility must also conduct emergency curtailment, meaning that entire distribution circuits must be intentionally de-energized to reduce system demand. Each utility is allocated a fraction of the total shortfall to correct.

In the test scenario, nearly everything remains the same, except a double-auction transactive market is coordinating battery energy storage and residential space conditioning and electric water heaters. These controllable assets are presumed to not be contracted by and to not participate in conventional demand-response. As the last available resources become dispatched, the costly final resources elevate the transactive price signal, thus causing transactive assets to respond. The demand-response resources are dispatched as for the base case, presuming they were scheduled that morning without consideration of the transactive system’s response. As the peak demand nears, the need for emergency curtailment might be reduced or fully avoided by the actions of the transactive system.

The principal valuation metrics for this use case address the costs and inconvenience of the emergency curtailment. Interesting impacts include changes in the numbers of customers curtailed, the durations of the emergency curtailment, and unserved load.

SGIP-2. “DER are engaged based on economics and location to balance wind resources” [12]. The scenario narrative states that ramping, not balancing or fast regulation, should be the target grid service for this use case.

This use case requires that bulk wind resources are a substantial fraction (40%) of the region’s bulk resource mix. Wind resources are highly correlated across the region. If the wind resource disappears rapidly, then other resources must be rapidly dispatched to replace the wind energy. This challenge is exacerbated if it occurs while other demand is increasing. If, however, wind resource materializes rapidly, other resources must ramp down, and this challenge is amplified if it occurs while other demand is decreasing. The ideal test day includes both the rapid ramping up and down of wind resource.

In the base case, supply is scheduled every hour or half-hour. The system must always allow a margin—ramping reserves―both up and down should these ramping services be needed. The system counteracts rapid changes in wind, both up and down, by controlling hydropower generation and spinning reserves [12]. The cost of doing this is often modest, given that hydropower generation might not even be the marginal resource. But the costs might understate the fact that more expensive resources might be used to provide this margin, and provision of ramping might impact hydropower generation maintenance costs. The cost of reserving resources is incurred regardless whether the system is ramping up or down. These reserves, as well as the costs of providing them, are addressed centrally by the system. The provision of ramping services is not isolated in that the quality of response might excite balancing and regulation services to become engaged.

In the test case, a transactive system is in operation, but the system otherwise operates the same.

We do not expect the double-auction transactive system to be particularly helpful for this use case. The dispatch algorithm generates the equivalent of a locational marginal price, which is responsive to the locational cost of marginal resource, efficiency, and transport constraints. While there will be some benefit caused by the transactive period being shorter than the scheduling interval, the transactive system here will respond to the marginal cost, which does not reflect ramping service costs. So, as wind ramps up and down, there will be a corresponding helpful reduction and increase in the transactive price signal. However, the transactive signal is not designed to align with the scheduling intervals and the corresponding needs for ramping services that result within each scheduling interval.

Primary impacts will address ramping reserves and their costs under the alternative scenarios.

SGIP-3. “High-penetration of rooftop solar PV causes swings in voltage on distribution grid” [12]. Solar generation capacity is stated to be up to 120% of load. Reversals of power flow can occur. Solar power intermittency creates corresponding voltage power quality issues. We choose to focus on the voltage management challenge, given that flow reversal is not itself a problem if it makes sense for system economics.

In the base case, this condition might today be disallowed at the planning stage because of the challenges that reversed power flow might induce in protection schemes. Presuming such high penetration and reversed flows are allowed, the distribution feeder must use its existing resources—capacitors, reactors, regulating transformers—to keep voltage in its acceptable range. Solar power inverters mostly correct to unity power factor today. Voltage tends to increase, if uncorrected, at times that solar power is injected into the distribution system. It is likely that this feeder will encounter voltage violations and flicker because of the high penetration and intermittency of the PV generation.

In the test case, the double-auction transactive system is operating on the high-solar-penetration feeder. Voltage management is not directly targeted by transactive mechanisms today, but the behaviors of the mechanisms can affect voltage management.

The primary impacts will be changes in the occurrences of voltage range violations, power quality events, and operations of voltage controls (e.g., tap changes) on the feeder.

SGIP-6. “A sudden transmission system constraint results in emergency load reductions” [12]. A distribution system network operator with a system having 150 MW peak winter load is given 15-minutes advance notice by his transmission supplier to curtail 40 MW. The curtailment is to last 2 hours. The distribution system network operator has no generation resources of his own to use. Business as usual mitigation is to conduct rolling blackouts. Alternatives exist if some or all of the emergency curtailment can be satisfied by DER [12]. Alternatively, the event might be naturally exercised by emulating contingency and maintenance outages. These events would then be stochastic in their occurrences.

SGIP-6 is very similar to SGIP-1, but it is caused by a system constraint rather than inadequate supply resources. It can be emulated by reducing the capacity of transmission or distribution that supply the test feeders. Refer to our discussion of SGIP-1 for the remedial actions, including conventional demand response, emergency curtailment, and double-auction transactive system that will be used in the base case and test scenarios. The valuation metrics and impacts are expected to be the same.

SGIP 1 Model Overview

Fig. 23 shows the types of assets and stakeholders considered for the use cases in this version. The active market participants include a double-auction market at the substation level, the bulk transmission and generation system, a large commercial building with one-way responsive HVAC thermostat, and single-family residences that have a two-way responsive HVAC thermostat. Transactive message flows and key attributes are indicated in orange.

In addition, the model includes PV and storage resources at some of the houses, and waterheaters at many houses. These resources can be transactive, but are not in this version because the corresponding separate TEAgents have not been implemented yet. Likewise, the planned new TEAgent that implements load shedding from the substation has not yet been implemented.

_images/SGIP1system.png

Fig. 23 SGIP-1 system configuration with partial PV and storage adoption

The Circuit Model

Fig. 24 shows the bulk system model in PYPOWER. It is a small system with three generating units and three load buses that comes with PYPOWER, to which we added a high-cost peaking unit to assure convergence of the optimal power flow in all cases. In SGIP-1 simulations, generating unit 2 was taken offline on the second day to simulate a contingency. The GridLAB-D model was connected to Bus 7, and scaled up to represent multiple feeders. In this way, prices, loads and resources on transmission and distribution systems can impact each other.

_images/PYPOWERsystem.png

Fig. 24 Bulk System Model with Maximum Generator Real Power Output Capacities

Fig. 25 shows the topology of a 12.47-kV feeder based on the western region of PNNL’s taxonomy of typical distribution feeders [13]. We use a MATLAB feeder generator script that produces these models from a typical feeder, including random placement of houses and load appliances of different sizes appropriate to the region. The model generator can also produce small commercial buildings, but these were not used here in favor of a detailed large building modeled in EnergyPlus. The resulting feeder model included 1594 houses, 755 of which had air conditioning, and approximately 4.8 MW peak load at the substation. We used a typical weather file for Arizona, and ran the simulation for two days, beginning midnight on July 1, 2013, which was a weekday. A normal day was simulated in order for the auction market history to stabilize, and on the second day, a bulk generation outage was simulated. See the code repository for more details.

Fig. 26 shows the building envelope for an elementary school model that was connected to the GridLAB-D feeder model at a 480-volt, three-phase transformer secondary. The total electric load varied from 48 kW to about 115 kW, depending on the hour of day. The EnergyPlus agent program collected metrics from the building model, and adjusted the thermostat setpoints based on real-time price, which is a form of passive response.

_images/FeederR1_1.png

Fig. 25 Distribution Feeder Model (http://emac.berkeley.edu/gridlabd/taxonomy_graphs/)

_images/School.png

Fig. 26 Elementary School Model

The Growth Model

This version of the growth model has been implemented for yearly increases in PV adoption, storage adoption, new (greenfield) houses, and load growth in existing houses. For SGIP-1, only the PV and storage growth has actually been used. A planned near-term extension will cover automatic transformer upgrades, making use of load growth more robust and practical.

Table 1 summarizes the growth model used in this report for SGIP-1. In row 1, with no (significant) transactive mechanism, one HVAC controller and one auction market agent were still used to transmit PYPOWER’s LMP down to the EnergyPlus model, which still responded to real-time prices. In this version, only the HVAC controllers were transactive. PV systems would operate autonomously at full output, and storage systems would operate autonomously in load-following mode.

Table 1 Growth Model for SGIP-1 Simulations

Case

Houses

HVAC Controllers

Waterheaters

PV Systems

Storage Systems

No TE

1594

1

1151

0

0

Year 0

1594

755

1151

0

0

Year 1

1594

755

1151

159

82

Year 2

1594

755

1151

311

170

Year 3

1594

755

1151

464

253

Insights and Lessons Learned

A public demonstration and rollout of TESP was included in a workshop on April 27, 2017, in Northern Virginia. That workshop marked the end of TESP’s first six-month release cycle. The main accomplishment, under our simulation task, is that all of the essential TESP components are working over the FNCS framework and on multiple operating systems. This has established the foundation for adding many more features and use case simulations over the next couple of release cycles, as described in Section 3. Many of these developments will be incremental, while others are more forward-looking.

Two significant lessons have been learned in this release cycle, meaning those two things need to be done differently going forward. The first lesson relates to MATPOWER. It has been difficult to deploy compiled versions of MATPOWER on all three operating systems, and it will be inconvenient for users to manage different versions of the required MATLAB runtime. This is true even for users who might already have a full version of MATLAB. Furthermore, we would need to modify MATPOWER source code in order to detect non-convergence and summarize transmission system losses. This led us to replace MATPOWER with PYPOWER [14] for the public releases of TESP. During 2019, TESP will be able to use AMES for day-ahead markets and unit commitment [15].

The second lesson relates to EnergyPlus modeling, which is a completely different domain than power system modeling. We were able to get help from other PNNL staff to make small corrections in the EnergyPlus model depicted in Fig. 26, but it’s clear we will need more building model experts on the team going forward. This will be especially true as we integrate VOLTTRON-based agents into TESP.