Automated Critical Peak Pricing Field Tests: 2006 Pilot Program Description and Results
Introduction During 2006 Lawrence Berkeley National Laboratory (LBNL) and the Demand Response Research Center (DRRC) performed a technology evaluation for the Pacific Gas and Electric Company (PG&E) Emerging Technologies Programs. This report summarizes the design, deployment, and results from the 2006 Automated Critical Peak Pricing Program (Auto-CPP). The program was designed to evaluate the feasibility of deploying automation systems that allow customers to participate in critical peak pricing (CPP) with a fully-automated response. The 2006 program was in operation during the entire six-month CPP period from May through October. Methodology The methodology for this field study included site recruitment, control strategy development, automation system deployment, and evaluation of sites’ participation in actual CPP events through the summer of 2006. LBNL recruited sites in PG&E’s territory in northern California through contacts from PG&E account managers, conferences, and industry meetings. Each site contact signed a memorandum of understanding with LBNL that outlined the activities needed to participate in the Auto-CPP program. Each facility worked with LBNL to select and implement control strategies for demand response and developed automation system designs based on existing Internet connectivity and building control systems. Once the automation systems were installed, LBNL conducted communications tests to ensure that the Demand Response Automation Server (DRAS) correctly provided and logged the continuous communications of the CPP signals with the energy management and control system (EMCS) for each site. LBNL also observed and evaluated Demand Response (DR) shed strategies to ensure proper commissioning of controls. The communication system allowed sites to receive day-ahead as well as day-of signals for pre-cooling, a DR strategy used at a few sites. Measurement of demand response was conducted using two different baseline models for estimating peak load savings. One was the CPP baseline model, which is based on the site electricity consumption from noon to 6 p.m. for the three days with highest consumption of the previous ten non-weekend days; it is not normalized for weather. The second model, the LBNL adjusted outside air temperature (OAT) regression baseline model, is based on OAT data and site electricity consumption from the previous ten days, and it is adjusted using weather regressions from the fifteen-minute electric load data during each event day. These baseline models were used to evaluate the demand reduction during each DR event for each site. The aggregated response from all sites for each event was also estimated using both baseline models. The evaluation research also included surveying the facility managers regarding any problems or issues that arose during the DR events. Questions covered occupant comfort, controls issues, and other potential problems. This 2006 Auto-CPP study included an assessment of the CPP economics for each site. This consisted of summing all of the credits on non-CPP days and subtracting the charges on CPP days. Estimates of the CPP economics without the demand response control strategies were also developed. Results • Twenty-four facilities participated in the Auto-CPP program. These facilities were a diverse set of building types, including office buildings, retail chain stores, schools, museums, laboratory buildings, a museum, and a bakery. • Thirteen sites participated in the majority of summer CPP events. There were nine CPP events in Zone 1 and eleven in Zone 2 in 2006. Among the Auto-CPP sites, site responses to 125 events were fully automated and evaluated in this study. Their average peak demand reduction was 14% of the whole-facility load based on the three-hour high-price period. The average peak demand reduction was 87 kW per facility, based on the OAT regression baseline model. The savings using a CPP baseline without weather normalization were less than half of the savings using the OAT regression baseline. • The program delivered an aggregated three-hour peak demand reduction of 1.2 MW on June 26, 2006 during an actual CPP event. • Even more potential was available as additional facilities came into the program in fall 2006. If all the sites that participated in 2006 provided their maximum six-hour peak demand reduction on the same day, the program could provide 1.7 MW of load reduction. If all sites provided the maximum three-hour peak demand reduction on the same day, the program could provide 2.0 MW. • During the severe heat wave of July 2006, all of the Auto-CPP sites continued to participate in DR at a time when it was needed most. None of facilities opted out. Internal temperatures in the buildings did rise above normal conditions, with some increase in occupant complaints, but not to the point of disrupting activities in the buildings or causing facilities personnel to disable the automation. • Full automation is technically feasible and provides value to CPP customers. One key aspect of the automation tests is that the facilities continue to participate after many years. Automation improves participation in demand response programs. Recommendations and Future Directions The 2006 Auto-CPP study showed that automating demand response is technically feasible. Planning for a scaled-up Auto-DR program for 2007, which includes other automated programs in addition to CPP, was initiated during 2006 Discussions have been underway with the three California investor-owned utilities (IOUs) to use a common Auto-DR infrastructure. The Demand Response Research Center (DRRC) will continue to support research to help understand the strengths and weaknesses of the current Auto-DR platforms and assist in identifying improvements. Specific examples of future research issues are listed below: • Explore Auto-DR for small commercial and large industrial sites. One of the long-term strategies of automating DR is to utilize customer relationships with current controls and communications technology vendors, informing and educating them on Auto-DR systems. Technically this project showed that most buildings with EMCS could participate in Auto-DR. Further work is needed to explore how to connect the DRAS with smaller buildings that do not have centralized EMCS. Further work is also needed to evaluate the readiness of industrial process control systems for automation. • Develop common peak demand savings evaluation methods. While the automation systems were shown to provide continuous, reliable communications of the DR program signals, more work is needed to understand end-use control strategies. Perhaps the most critical need is to engage the engineering community and auditors who evaluate DR strategies and estimate peak demand savings to develop common methods for savings calculations. While there are decades of experience with energy savings analysis methods and techniques, methods to estimate peak demand savings for short durations are relatively new. Such analysis methods are more complex than historical “bin” methods for energy efficiency analysis that simplify weather data into heating and cooling degree-days. Rather, new dynamic models are needed, based on knowledge of weather data, peak load shapes, and HVAC system and controls, combined in practical ways to provide simple, yet robust concepts for peak demand savings estimates. • Improve communication on the CPP tariff. PG&E’s CPP tariff is complex. The July 2006 heat storm resulted in one month with seven CPP events. This caused an average increase in commercial sector summer bills of fifteen percent. Many of the participating sites were concerned with their high mid-summer utility bill following the heat wave. Improvements in communication by utilities with customers about bills are needed to explain the charges and credits each site is expected to collect for the entire summer if it enrolls in CPP. • Provide better information on the state benefits of DR. Demand response is a confusing term and DR programs are confusing. More effort is needed to communicate the concepts of DR. Automating DR may help improve the reliability of the resource, but there is a hurdle in marketing these programs because of limited understanding. • Consider alternative weather-adjusted baseline models. The Auto-CPP project showed that the CPP baseline was lower than hot peak day loads prior to CPP events. When the CPP baseline is lower than the load shape, there are no estimated DR savings. Weather-sensitive loads need weather-adjusted baseline models.