What does the next generation of demand response look like? Will technology advances, ubiquitous use of distributed sensing, control and cyber-secure platforms combined with low-cost communication infrastructure lead to autonomous and intelligent building controls?  Will this lead to the widespread integration of buildings with the grid, leading to both benefits to the consumers and the grid operators?  The authors of this article believe it will and that the technologies and platforms that appear on the pages of Automated Buildings every month ( VOLTTRON™, wireless communication, advanced controls) are the seeds for the next generation of demand response. 

Why do we need next generation of demand response technologies?

In the U.S., buildings consume over 75% of the electricity.  A significant portion of the new electricity generation is in the form of distributed renewables, which is variable and not dispatchable.  If the variable generation becomes a significant fraction of the total system supply, it could potentially impact the reliability of the grid.  Distributed thermal and battery storage can mitigate some of the imbalance between supply and demand caused by variable generation assets.  However, business-as-usual manual and semi-automated processes will not scale and will not be cost-effective.  We need fully automated intelligent control algorithms that can make autonomous decisions based on user preferences, and we also need transactional platforms that can be deployed on low-cost distributed computing platforms.

How is demand response deployed in buildings now?

The concept of demand response (DR) has been with us for a long time now, over 4 decades. Early DR deployments were mostly direct load control, which is still practiced by some utilities today. Control companies have developed semi-automated DR strategies, such as demand limiting and duty cycling.  The duty-cycling control strategy has been traditionally used to manage peak demand by controlling the ratio of the on-period to the total cycle time of rooftop units (RTUs) or air-handling units (AHUs).  Although duty-cycling provides relief from electric demand, it does not have the ability to dynamically prioritize the RTUs to be curtailed to manage peak electricity consumption. In this approach, it is generally difficult to identify the RTUs that can be curtailed without affecting zone comfort, and indiscriminate curtailment of RTUs can lead to comfort issues by negatively affecting the zone temperature and humidity conditions. Therefore, a load control strategy is needed that anticipates the future effects of thermal comfort and peak load relief based on current conditions and past historical data.

What is beyond DR?

We need a process that dynamically prioritizes the available loads for curtailment based on quantitative (deviation of zone conditions from set point) and qualitative ( a type of zone: kitchen, corner office, director’s office, etc.) rules set by the building operators. A PNNL-developed technology, Intelligent Load Control (ILC), offers a solution for achieving this balance (Kim et al. 2016, Kim and Katipamula 2017). ILC is an algorithm, or a set of actions, deployed via VOLTTRON™, a software platform, to a building’s control system. The ILC technology can automatically adjust building energy use by coordinating heating and cooling, lights and other building functions while maintaining occupant comfort. In addition to bringing flexibility and responsiveness to building energy consumption, ILC benefits the grid by turning buildings into resources that help balance supply and demand—which could help address some of the challenges of adding intermittent resources such as the wind and solar power to the energy supply mix.

ILC employs the analytic hierarchy process (AHP)—a method for prioritizing actions to achieve the best results. For example, AHP strategies can help determine whether shutting down or turning on heat pumps in a certain sequence will achieve optimum energy reductions.

What are ILC capabilities?

Use case 1, Managing Monthly Peak Demand: When demand for electricity rises across the grid, so do costs. ILC can manage building devices in a way that reduces more expensive peak power use periods.  In the traditional utility rate structure, many utilities charge not only for the energy (kWh) consumed but also for the peak power (kW) consumed.  The peak is generally calculated as a rolling average over any 15-minute or 30-minute period during a billing period (typically 30 days).  To manage the building peak under this scenario, the building peak consumption over the 30-day billing period has to be forecast.  PNNL has developed an agent that will forecast power using historical energy consumption by the building and weather forecasts into the future. Next, a peak demand target (e.g., 85% of forecasted peak) is selected.  Using this approach, ILC can control the operation of multiple controllable loads (e.g., heat pumps serving offices and other work spaces or dimmable lighting loads). ILC has been validated on several real buildings (see Figure 1 and Figure 2 below for sample results). Results had shown that when energy consumption peaked at different times—such as first thing in the morning while the building was being readied for the workday—ILC quickly prioritized operation of heat pumps, shutting down some units while running others, and returning the building to normal operations when the peak demand period passed. The approach successfully reduced demand to meet an established limit, which benefits power grid operations, and no loss of service levels was reported.  

Figure 1: The top graph shows the operation of the building without ILC controlling the devices during heating season. Note 1: Peak demand for this day is 175 kW, which was set at around 7 a.m. Note that the rest of the day the demand never exceeded 145 kW. Also, note that the duration of the peak demand was less than 15 minutes.

The bottom graph shows the same building with a target peak of 145 kW. Note 1: Peak demand for this day never exceeds the target of 150 kW, which was set at around 7 a.m. Note 2: Some end-use loads were turned off to manage the load shape. If ILC were operational the previous day, the building could have avoided approximately $125 of peak charges at $5/kW or $500 at $20/kW.

Figure 2: The top graph shows the operation of the building without ILC controlling the devices during the cooling season. Note 1: Peak demand for this day is 145 kW, which occurred between 2 p.m. and 4 p.m. Also, note that the peak demand occurred for significantly longer duration than originally anticipated.

The bottom graph shows the same building with a target peak of 135 kW. Note 1: Peak demand for this day never exceeds the target of 135 kW, which was set several times between 12:30 p.m. and 4 p.m. Note 2: Some loads were turned off to manage the load shape. If ILC were operational the previous day, the building could have avoided approximately $50 of peak charges at $5/kW or $200 at $20/kW.

Use Case 2, Supporting Capacity Bidding: Capacity bidding is a method in which buildings agree to use less energy over certain periods, providing relief to the power grid at times when electricity demand is high. Under this approach, building management would reach a monthly agreement on a certain amount of capacity, or “relief” the building can give up, if requested to do so by the grid. The capacity amount can be constant or cover specific blocks of time. In return, electricity suppliers would provide incentives for the relief.  ILC has been tested and validated to support capacity bidding programs, and the results are represented in the chart below (Figure 3). The red line denotes the temperature adjusted baseline or the amount of power that would be used by the building if nothing changes. The green line is the capacity bidding target, and the blue is actual usage. Once the call comes from the grid for relief, ILC automatically begins coordinating device operation to achieve consumption levels below the red line—while concurrently ensuring building safety and comfort levels. In testing, ILC readily achieved objectives, reducing electricity use for the correct time intervals.  The entire process is fully automated, including generation of baseline that is used to quantify the amount of relief the building has provided (Figure 3).  In addition, the building owners/operators can see how well the building is actually performing in real time.  Finally, the measurement and verification (M&V) process are also fully automated.  As soon as the event ends, the M&V report is automatically generated and can be emailed to relevant parties.

Figure 3: Shows the results of testing the capacity bidding approach on a commercial building.  The light green line is the 10-day baseline, the red line is the 10-day temperature-adjusted baseline, the dark green line is the target based on the capacity relief the building was going to provide, and the blue line is the actual consumption during the event.  Note that the actual consumption (blue line) was mostly below the target value (green line) during the event.

Use Case 3, Supporting Transactive Control and Coordination: ILC has also demonstrated its ability to support the transactive control concept, which in the future is expected to revolutionize operation of the nation’s energy system (Corbin et al. 2016, Hao et al. 2017). Transactive control enables energy-related devices to continuously exchange electricity consumption and price information with suppliers and negotiate and respond to that information to optimize energy use in real time.

Under ILC, a price-capacity curve is established for individual buildings. The curve represents how much energy a building is willing to use based on the price at a given time, and is communicated and proposed to the power market (Figure 3). Once the market accepts, ILC begins managing devices to the capacity limit that corresponds to the cleared price, while maintaining safety and comfort. If the energy is low, the building can buy more power and perhaps perform tasks in advance, such as pre-cooling. It’s envisioned the transactions between buildings and the power market will occur in short time intervals (e.g., every 5 minutes), enabling improved grid planning, flexibility, and efficiency.

Figure 3: The graph shows expected building consumption (blue), the target consumption (price-based), and the actual (green). Note that the green line stays below the orange for most of the day. 

How is ILC deployed in the field?

The ILC algorithm is deployed from VOLTTRON, an open-source distributed control and sensing software platform also developed at PNNL (Katipamula et al. 2016a, 2016b, 2016c). VOLTTRON provides an environment for applications, or “agents,” like ILC and serves as a single point of contact for interfacing with devices (rooftop units, lighting, other building loads and systems, meters, etc.). VOLTTRON’s ability to manage data, devices, and decision-making has provided new options for coordinating the integration of buildings, the grid and clean energy sources.

[an error occurred while processing this directive] When will the path to the beyond demand response take shape?

With the availability of the open-source distributed sensing, control and cyber-secure platforms like VOLTTRON and agents like intelligent load control that are deployed on it, the future is not far away.  The U.S. Department of Energy is engaged in a number of efforts to validate these technologies in both individual buildings and also on a campus scale. Several utilities have also shown interest to go beyond the traditional DR to support grid reliability and low cost to their customers.  Coupled with these developments and the availability of low-cost distributed computer platforms (e.g., Raspberry Pi, Beagle Bone, Intel NUC, etc.) and communications will make beyond DR a reality sooner rather than later!
_____________________________________

References:

Corbin CD, A Makhmalbaf, G Liu, S Huang, VV Mendon, M Zhao, S Somasundaram, H Ngo, and S Katipamula. 2016. “Transactive Control of Commercial Building HVAC Systems.” PNNL-26083, Pacific Northwest National Laboratory, Richland, WA.

Hao H, DC Charles, K Kalsi, and RG Pratt. 2017. "Transactive Control of Commercial Buildings for Demand Response," IEEE Transactions on Power Systems, 32(1), 774–783, January 2017

Katipamula S, J. Haack, G. Hernandez, B. Akyol and J. Hagerman. 2016a. "VOLTTRON: An Open-Source Software Platform of the Future," IEEE Electrification Magazine, vol. 4, no. 4, pp. 15-22, Dec. 2016. DOI: 10.1109/MELE.2016.2614178. URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7725895&isnumber=7725795

Katipamula S, RG Lutes, G Hernandez, JN Haack, and BA Akyol.  2016b. "Transactional Network: Improving Efficiency and Enabling Grid Services for Buildings."  Science and Technology for the Built Environment (2016), 22(6), pp 643-654 DOI:10.1080/23744731.2016.1171628

Katipamula S, K Gowri, and G Hernandez. 2016c. “An Open-source Automated Continuous Condition-based Maintenance Platform for Commercial Buildings.” Science and Technology for the Built Environment (2016) 00, 1–11 doi: 10.1080/23744731.2016.1218236

Kim W, S Katipamula, RG Lutes, and RM Underhill. 2016. "Behind the Meter Grid Services: Intelligent Load Control” PNNL-26034, Pacific Northwest National Laboratory, Richland, WA.

Kim W, and S Katipamula. 2017. "Development and Validation of an Intelligent Load Control Algorithm." Energy and Buildings, 135 (2016), pp 62-73. http://dx.doi.org/10.1016/j.enbuild.2016.11.040

About the Authors

 is a Staff Scientist at Pacific Northwest National Laboratory (PNNL). For over 25 years, Katipamula’s career has focused on improving the operating efficiency of commercial buildings. He has extensive technical experience in the evaluation of advanced design concepts for heating, ventilation and air-conditioning systems (HVAC), development of automated fault detection and diagnostic techniques, self-correcting controls, building, and energy system simulations, analysis and evaluation of new energy efficient technologies, and development and use of analytical modeling techniques. At PNNL, Dr. Katipamula leads a number of research projects in advanced building controls, automated fault detection, and diagnostics, building operations, and building-grid integration. He is a fellow of the American Society of Heating, Refrigeration and Air Conditioning Engineers, the American Society of Mechanical Engineers (ASME) and former associated editor of ASME Journal of Energy Resources Technology.

George Hernandez joined PNNL in 2009 and works in the Electricity Infrastructure and Buildings Division. Mr. Hernandez is the Principal Technical Advisor for Buildings to Grid Integration and Buildings Controls Research Program Manager.  While on detail at DOE’s Building Technologies Office, he created the Sensors and Controls program and co-authored the High-Performance RTU Challenge, the Buildings Performance Database, the Low-Cost Wireless Metering Challenge, Energy Information Handbook, the Portable Sensor Suitcase, Open Source Small Building Control System, and the Transactional Network project.  Most recently, he has been guiding the development of VOLTTRON, an open source agent based software platform.  Mr. Hernandez has extensive knowledge, skills, and capabilities derived from a substantial career in demand side utility management across a wide variety of commercial and industrial sectors and utilities as both a corporate employee and an independent consultant. Mr. Hernandez received his BS in Mechanical Engineering from California State University and his Masters in Mechanical Engineering from The University of California at Berkeley. He is a Licensed Professional Engineer (PE) by the State of California.