The Olympic Peninsula Project consisted of a field demonstration and test of advanced price signal-based control of distributed energy resources (DERs). Sponsored by the U.S. Department of Energy (DOE) and led by the Pacific Northwest National Laboratory, the project was part of the Pacific Northwest Grid- Wise Testbed Demonstration.

Other participating organizations included the Bonneville Power Administration, Public Utility District (PUD) #1 of Clallam County, the City of Port Angeles, Portland General Electric, IBM’s T.J. Watson Research Center, Whirlpool and Invensys Controls. The main objective of the project was to convert normally passive loads and idle distributed generation into actively participating resources optimally coordinated in near real time to reduce stress on the local distribution system.

Planning began in late 2004, and the bulk of the development work took place in 2005. By late 2005, equipment installations had begun, and by spring 2006, the experiment was fully operational, remaining so for one full year.

The motivating theme of the project was based on the GridWise concept that inserting intelligence into electric grid components at every point in the supply chain – from generation through end-use – will significantly improve both the electrical and economic efficiency of the power system. In this case, information technology and communications were used to create a real-time energy market system that could control demand response automation and distributed generation dispatch. Optimal use of the DER assets was achieved through the market, which was designed to manage the flow of power through a constrained distribution feeder circuit.

The project also illustrated the value of interoperability in several ways, as defined by the DOE’s GridWise Architecture Council (GWAC). First, a highly heterogeneous set of energy assets, associated automation controls and business processes was composed into a single solution integrating a purely economic or business function (the market-clearing system) with purely physical or operational functions (thermostatic control of space heating and water heating). This demonstrated interoperability at the technical and informational levels of the GWAC Interoperability Framework (www.gridwiseac.org/about/publications.aspx), providing an ideal example of a cyber-physical-business system. In addition, it represents an important class of solutions that will emerge as part of the transition to smart grids.

Second, the objectives of the various asset owners participating in the market were continuously balanced to maintain the optimal solution at any point in time. This included the residential demand response customers; the commercial and municipal entities with both demand response and distributed generation; and the utilities, which demonstrated interoperability at the organizational level of the framework.

PROJECT RESOURCES

The following energy assets were configured to respond to market price signals:

  • Residential demand response for electric space and water heating in 112 single-family homes using gateways connected by DSL or cable modem to provide two-way communication. The residential demand response system allowed the current market price of electricity to be presented to customers. Consumers could also configure their demand response automation preferences. The residential consumers were evenly divided among three contract types (fixed, time of use and real time) and a fourth control group. All electricity consumption was metered, but only the loads in price-responsive homes were controlled by the project (approximately 75 KW).
  • Two distributed generation units (175 KW and 600 KW) at a commercial site served the facility’s load when the feeder supply was not sufficient. These units were not connected in parallel to the grid, so they were bid into the market as a demand response asset equal to the total load of the facility (approximately 170 KW). When the bid was satisfied, the facility disconnected from the grid and shifted its load to the distributed generation units.
  • One distributed microturbine (30 KW) that was connected in parallel to the grid. This unit was bid into the market as a generation asset based on the actual fixed and variable expenses of running the unit.
  • Five 40-horsepower (HP) water pumps distributed between two municipal water-pumping stations (approximately 150 KW of total nameplate load). The demand response load from these pumps was incrementally bid into the market based on the water level in the pumped storage reservoir, effectively converting the top few feet of the reservoir capacity into a demand response asset on the electrical grid.

Monitoring was performed for all of these resources, and in cases of price-responsive contracts, automated control of demand response was also provided. All consumers who employed automated control were able to temporarily disable or override project control of their loads or generation units. In the residential realtime price demand response homes, consumers were given a simple configuration choice for their space heating and water heating that involved selecting an ideal set point and a degree of trade-off between comfort and price responsiveness.

For real-time price contracts, the space heater demand response involved automated bidding into the market by the space heating system. Since the programmable thermostats deployed in the project didn’t support real-time market bidding, IBM Research implemented virtual thermostats in software using an event-based distributed programming prototype called Internet- Scale Control Systems (iCS). The iCS prototype is designed to support distributed control applications that span virtually any underlying device or business process through the definition of software sensor, actuator and control objects connected by an asynchronous event programming model that can be deployed on a wide range of underlying communication and runtime environments. For this project, virtual thermostats were defined that conceptually wrapped the real thermostats and incorporated all of their functionality while at the same time providing the additional functionality needed to implement the real-time bidding. These virtual thermostats received
the actual temperature of the house as well as information about the real-time market average price and price distribution and the consumer’s preferences for set point and comfort/economy trade-off setting. This allowed the virtual thermostats to calculate the appropriate bid every five minutes based on the changing temperature and market price of energy.

The real-time market in the project was implemented as a shadow market – that is, rather than change the actual utility billing structure, the project implemented a parallel billing system and a real-time market. Consumers still received their normal utility bill each month, but in addition they received an online bill from the shadow market. This additional bill was paid from a debit account that used funds seeded by the project based on historical energy consumption information for the consumer.

The objective was to provide an economic incentive to consumers to be more price responsive. This was accomplished by allowing the consumers to keep the remaining balance in the debit account at the end of each quarter. Those consumers who were most responsive were estimated to receive about $150 at the end of the quarter.

The market in the project cleared every five minutes, having received demand response bids, distributed generation bids and a base supply bid based on the supply capacity and wholesale price of energy in the Mid-Columbia system operated by Bonneville Power Administration. (This was accomplished through a Dow Jones feed of the Mid-Columbia price and other information sources for capacity.) The market operation required project assets to submit bids every five minutes into the market, and then respond to the cleared price at the end of the five-minute market cycle. In the case of residential space heating in real-time price contract homes, the virtual thermostats adjusted the temperature set point every five minutes; however, in most cases the adjustment was negligible (for example, one-tenth of a degree) if the price was stable.

KEY FINDINGS

Distribution constraint management. As one of the primary objectives of the experiment, distribution constraint management was successfully accomplished. The distribution feeder-imported capacity was managed through demand response automation to a cap of 750 KW for all but one five-minute market cycle during the project year. In addition, distributed generation was dispatched as needed during the project, up to a peak of about 350 KW.

During one period of about 40 hours that took place from Oct. 30, 2006, to Nov. 1, 2006, the system successfully constrained the feeder import capacity at its limit and dispatched distributed generation several times, as shown in Figure 1. In this figure, actual demand under real-time price control is shown in red, while the blue line depicts what demand would have been without real-time price control. It should be noted that the red demand line steps up and down above the feeder capacity line several times during the event – this is the result of distributed generation units being dispatched and removed as their bid prices are met or not.

Market-based control demonstrated. The project controlled both heating and cooling loads, which showed a surprisingly significant shift in energy consumption. Space conditioning loads in real-time price contract homes demonstrated a significant shift to early morning hours – a shift that occurred during both constrained and unconstrained feeder conditions but was more pronounced during constrained periods. This is similar to what one would expect in preheating or precooling systems, but neither the real nor the virtual thermostats in the project had any explicit prediction capability. The analysis showed that the diurnal shape of the price curve itself caused the effect.

Peak load reduced. The project’s realtime price control system both deferred and shifted peak load very effectively. Unlike the time-of-use system, the realtime price control system operated at a fine level of precision, responding only when constraints were present and resulting in a precise and proportionally appropriate level of response. The time-of-use system, on the other hand, was much coarser in its response and responded regardless of conditions on the grid, since it was only responding to preconfiured time schedules or manually initiated critical peak price signals.

Internet-based control demonstrated. Bids and control of the distributed energy resources in the project were implemented over Internet connections. As an example, the residential thermostats modified their operation through a combination of local and central control communicated as asynchronous events over the Internet. Even in situations of intermittent communication failure, resources typically performed well in default mode until communications could be re-established. This example of the resilience of a well-designed, loosely coupled distributed control application schema is an important aspect of what the project demonstrated.

Distributed generation served as a valuable resource. The project was highly effective in using the distributed generation units, dispatching them many times over the duration of the experiment. Since the diesel generators were restricted by environmental licensing regulations to operate no more than 100 hours per year, the bid calculation factored in a sliding scale price premium such that bids would become higher as the cumulative runtime for the generators increased toward 100 hours.

CONCLUSION

The Olympic Peninsula Project was unique in many ways. It clearly demonstrated the value of the GridWise concepts of leveraging information technology and incorporating market constructs to manage distributed energy resources. Local marginal price signals as implemented through the market clearing process, and the overall event-based software integration framework successfully managed the bidding and dispatch of loads and balanced the issues of wholesale costs, distribution congestion and customer needs in a very natural fashion.

The final report (as well as background material) on the project is available at www.gridwise.pnl.gov. The report expands on the remarks in this article and provides detailed coverage of a number of important assertions supported by the project, including:

  • Market-based control was shown to be a viable and effective tool for managing price-based responses from single-family premises.
  • Peak load reduction was successfully accomplished.
  • Automation was extremely important in obtaining consistent responses from both supply and demand resources.
  • The project demonstrated that demand response programs could be designed by establishing debit account incentives without changing the actual energy prices offered by energy providers.

Although technological challenges were identified and noted, the project found no fundamental obstacles to implementing similar systems at a much larger scale. Thus, it’s hoped that an opportunity to do so will present itself at some point in the near future.