Title: Towards More Automated Engagement of Electrical Appliances in the Smart Grid

Residential and commercial buildings contain a growing number of electrical appliances delivering many of the services that these human shelters offer. Traditionally, information about the past and future electrical power consumption of a large portion of these devices has been undocumented and separated from decisions taken by the building occupants, facility managers and grid operators. However, an increased awareness of the benefits of energy conservation in light of future climate change scenarios, as well as a restructuring of the power infrastructure towards a more interconnected and flexible grid have brought about significant opportunities and use cases for this type of information. Yet, obtaining granular and accurate information about appliance power consumption from existing measurement sources is not straightforward given that our metering infrastructure (and technologies) do not provide full observability. In this talk I will present recent advances my group has made towards providing two pieces of information regarding appliances in buildings, namely: identifying appliance types from measurements of voltage and current at the outlet, and estimating the state of a large population of refrigerators (or any other thermostatically controlled load) from measurements of their aggregate power consumption.

Bio:

Mario Berges is an assistant professor in the Department of Civil and Environmental Engineering at CMU. He is interested in making use of cost-effective sensor systems to automatically create models and generate insights that can be used to improve the behavior of infrastructure systems, prevent failures, and better plan for the future. He has been working on three different approaches related to this: appliance-level energy feedback through minimally intrusive strategies, sharing sensing and actuation resources at Internet scales, and unsupervised sensor fusion for proactive energy management. His current research interests also include machine learning, signal processing, the Internet of Things and the smart grid. He is the faculty co-director of the IBM Smart Infrastructure Analytics Laboratory at CMU, as well as the director of the Intelligent Infrastructure Research Lab (INFERLab). Among recent awards, he received the Outstanding Early Career Researcher award from FIATECH in 2010. He received his B.Sc. in 2004 from the Instituto Tecnologico de Santo Domingo, in the Dominican Republic; and his M.Sc. and Ph.D. in Civil and Environmental Engineering in 2007 and 2010, respectively, both from Carnegie Mellon University.

Title: 11 Insights Made Possible by High Resolution Disaggregated Data on Residential Electricity Use

For the past five years, Pecan Street has operated one of the nation˘s most data-intensive research trials on residential electricity use, behavioral response and field performance of different measurement devices. Using smart meters, gateway devices and (primarily) CT-collar systems, Pecan Street researchers have measured appliance-level electricity use every minute in hundreds of homes in three states, including measuring generation from rooftop solar panels in nearly 200 homes and home charging in over 70 homes with electric vehicles. In late 2013, Pecan Street began making all of this data available for free in an online searchable database to the academic research community. This presentation lays out 11 preliminary insights drawing on original data from these research trials.

Bio:

Brewster McCracken is President and CEO of Pecan Street Inc., a research and commercialization institute focused on the utility industry headquartered at The University of Texas. In July 2014, Time described Pecan Street's work as "the most extensive energy-tracking study in U.S. history. . . . That kind of data is unprecedented in the electricity industry. . . The Pecan Street team is already using it to upend long-held theories about electricity use."

In 2013, Smart Grid Today named him one of the nation's "Smart Grid Pioneers", and GreenBiz.com named him to its VERGE 25 list of 25 U.S. smart grid leaders. He is the author of numerous research reports on customer energy use.

Title: Data Driven Investigation of Faults in HVAC Systems

It is estimated that the complexity of modern HVAC systems leads to device mis-configuration in 40% of buildings, and that HVAC faults waste upto 40% of the energy consumed. Fault detection methods used in practice are error prone, with excessive alarms leading to operator alert fatigue, faults left undetected and energy wastage. While sophisticated fault detection techniques have been developed and demonstrated in the literature, they are seldom used in practice. We investigate this gap by applying various fault detection techniques on real data from a 145,000 sqft, five floor building. We find that, while different algorithms are successful in detecting different classes of faults, none of these algorithms capture control loop configuration faults by design. To address this, we develop a novel algorithm, Model, Cluster and Compare (MCC) that is able to detect anomalies by automatically grouping similar entities in an HVAC system, in an unsupervised manner, and com- paring them. We implemented MCC to detect faults in Variable Air Volume (VAV) boxes in our building testbed, and demonstrate that it successfully detects non-obvious configuration faults.

Title: Algorithmic Decision Theory and the Smart Grid

Today's decision makers in fields ranging from engineering to medicine to homeland security have available to them remarkable new technologies, huge amounts of information, and the ability to share information at unprecedented speeds and quantities. These tools and resources are particularly relevant to problems at the nexus of energy, environment, and the economy that lie at the heart of the development of the smart grid. The tools and resources will enable better decisions if we can surmount concomitant challenges: The massive amounts of data available are often incomplete or unreliable or distributed and there is great uncertainty in them; interoperating/distributed decision makers and decision-making devices need to be coordinated; many sources of data need to be fused into a good decision, often in a remarkably short time; decisions must be made in dynamic environments based on partial information; there is heightened risk due to extreme consequences of poor decisions; decision makers must understand complex, multi-disciplinary problems. In the face of these new opportunities and challenges, the field of Algorithmic Decision Theory (ADT) aims to exploit algorithmic methods to improve the performance of decision makers (human or automated). ADT is extremely relevant to problems of the smart grid, which allows for real-time precision in operations and control previously unobtainable but raises new issues of privacy and vulnerability.

Title: Energy Analysis of HVAC Systems in Commercial Buildings

HVAC systems in modern commercial buildings contribute to 40% of its energy use, and yet its energy consumption is typically monitored using building level meters, providing little insight into its efficiency of operation. We have built an infrastructure of technologies that enables us to get a deeper view into the energy consumption of an HVAC system. We have developed a web service based data storage solution, called BuildingDepot, which helps collect data across different sensors in the building, exposes these data through RESTful APIs, and provides access control. On top of BuildingDepot, we have developed applications which manage and analyze this data both for building occupants and managers. Genie is our occupant management web service, which provides an interface to the occupants through which they can view room temperature, send comfort complaints and energy consumed by HVAC. BDSherlock is our building manager level web service, that provides an overview of building status, sensor data across the buildings, and identifies the faults that cause energy wastage.

Title: Data-driven Building Model Capture for Economic Energy Demand Control

One of the biggest challenges in the design of closed-loop Cyber-Physical Systems (CPS) is in accurately capturing the dynamics of the underlying physical system. In the context of buildings, the modeling difficulty arises due to the fact that each building is designed and used in a different way and therefore, it has to be uniquely modeled (Heterogeneity). Furthermore, each building system consists of a large number of interconnected subsys- tems which interact in a complex manner and are subjected to time varying environmental conditions (Complexity).

Title: Optimal HVAC Building Control with Occupancy Prediction

Buildings account for about 41% of primary energy consumption and 75% of the electricity, 44% more than the transportation sector and 36% more than the industrial sector. Space heating, space cooling, and ventilation are the dominant end uses, accounting for 41% of all energy consumed in the buildings sector. Growing interest in sustainability has resulted in research efforts to reduce energy consumption while providing adequate quality of service to its occupants.

In this work, we present a Model Predictive Control (MPC) framework for optimal HVAC control that minimizes energy consumption while staying within the comfort bounds of the occupants. The novelty of our approach lies in the use of prediction occupancy models derived from data traces and incorporating those models within the MPC framework. We use a Blended Markov Chain (BMC) occupancy prediction model in order to predict thermal load and occpancy of each zone in the building. We test our approach in simulation and compare it with occupancy schedules and control rules currently use in our university buildings. Our preliminary results show that 18% savings in cooling in the summer, and 10\% savings in heating in the winter are achievable when conditioning the building using our MPC/BMC control framework.

Title: Placement and Control of Energy Storage in the Grid: a System Operator's Viewpoint

Storage technologies have many potential applications in the power grid ranging from mitigating intermittency of renewable generation to load shifting. In this talk, we consider storage as a communal asset, i.e., the system operator solves the problem of minimizing total generation cost in the grid subject to transmission constraints of the network, when it has access to distributed energy storage resources. Given a stochastic net-demand (load minus renewable supply) process evolving over a transmission-constrained power network, we show that the expected system benefit of storage derived under the optimal dispatch policy is concave and non-decreasing in both energy and power capacities of storage. Thus, the greatest marginal benefit is derived at small energy capacities of installed storage. For such capacities, we characterize the locational (nodal) marginal value of storage in terms of the spectral properties of the stochastic net-demand process. This allows us to calculate the marginal benefits in terms of strict level crossings of a scalar random process derived from wind and demand data across a power network. We present the results of numerical simulations on some power networks. In a second study, we consider the use of storage in load shifting, i.e., attening generation profile to reduce a convex cost of generation. We show how data driven empirical study leads us to an interesting observation, which we characterize (partially) in a structural result: a generator node that connects to the rest of the network through one capacity-constrained transmission line never needs any storage capacity under optimal allocation. This holds regardless of the demand profile or other network parameters.

Title: Issues related to Energy Consumption in STEM Education

The use of energy is a global issue and can be addressed via mathematics courses. Equity in the consumption, consumerism and distribution of available energy is an important global issue for the students to understand. It is known worldwide that STEM (Science, Technology, Engineering and Mathematics) Education is critical in an effort to prepare our students and promote innovation, creativity, global competence, democratic principles and social justice, improve economic competitiveness, and provide leadership. It has been stated that the poorest 10% only accounts for 0.5% and the wealthiest 10% accounts for 59% of all consumption. Globally, 20% of the world's population in the higher-income countries accounts for 86% of total private consumptions and the 20% of world's population in the lower-income countries accounts for 1.3% of total private consumption. The goal of this project is to present with a design, create and develop academic content in a course related to energy consumption in accordance with students' level of mathematical understanding, and then implement and assess the outcomes

Title: Minimizing Data Storage for High Granularity Energy Data Collection

To reduce peak demand, many utility companies are transitioning from fixed rate pricing plans to real-time pricing plans. To apply real-time pricing plans, it is crucial to collect accurate energy data with high granularity from individual homes. Thus, smart meters that measure energy data in high time granularity have been rolled out in many countries. However, energy data with high granularity needs large data storage and would generate significant communication overhead for utility companies to collect all the data from large amount of homes. In this paper, we present a frame work for utility companies to collect energy data with minimum data storage, computing resources and communication overhead. We first make use of a polynomial curve to sketch the approximate curve of power consumption. Then based on analyses of power consumption patterns, we propose two different types of methods to record the differences between real power consumption curve and the polynomial curve with minimum storage. We conducted extensive system evaluations with 30 homes' power consumption data every second for more than 2 months. Results indicate i) our design can reduce data storage significantly by 70% with only 0.02kW error on average from real data in a single home, and ii) even though our design is based on energy consumption data, it also works well with other energy data (e.g., voltage and frequency).

Title: SolarCast: An Open Service to Predict Solar Power in the Black Box Setting

With reducing costs, homeowners are encouraged to install rooftop solar generation units. But solar power is characteristically intermittent and inconsistent due to the diurnal nature of sun and varying weather conditions at any given place. Thus, predicting solar power is non-trivial. Presence of a solar power prediction service would help in both supply and demand side load management by integrating with smart devices, sunny load scheduling, predicting grid level variable aggregated demand, designing incentives to customers with distributed power generations capacity etc. Past work, has primarily focused on white-box prediction schemes that require detailed site and panel specific information. A homeowner may not always have such information at his disposal. Further, in case of rooftop installations there is an additional challenge of shadows caused by local topology. We present SolarCast, an open service to make black box solar power predictions. It only requires a geographic location of the site and a minimal amount of historical generation data. We evaluate SolarCast's accuracy on a geographically diverse dataset of solar deployments across the different states in US. We also expose RESTful APIs for developers to integrate our predictions into their smart energy devices and other related apps. With increased adoption of our service, we intend to evolve more powerful models that would improve our prediction accuracy.

Title: Net load forecasting for communities with distributed and centralized solar generation

The power grid is undergoing a change. The growth of solar power utilization has been staggering: global solar capacity has increased from 1.76 GW to 102.15 GW over the past decade (2001-2012). Due to the stochastic nature of solar power, these increasing levels of solar penetration result in several planning and operational challenges. The impact of an increasing solar penetration on load forecast capability was studied for a community with centralized solar generation and 33% annual solar penetration. It was found that the load forecasting skill drops by 9% for the hour-ahead load forecast because of additional variability and uncertainty introduced by solar power. Analysis of the error distribution as a function of daily solar penetration for different levels of variability revealed that the solar power variability drives the forecast error magnitude whereas increasing penetration level has a much smaller contribution. Therefore, net load forecasting i.e., the integration of demand load and renewable generation forecast techniques is needed to replace existing grid load management systems. Net load forecasting models for communities with centralized and distributed generation are being developed. Preliminary results show that the accuracy of net load forecast is limited by the solar forecasting capabilities. For systems with centralized solar generation, the ramps in net load are proportional to solar irradiance whereas for distributed system, ramps are related to weighted average of solar power. Lastly, for systems with distributed solar generation and high solar penetration (>5%), the error magnitude is a linear function of solar variability. Thus, while allocating resources and operating reserves for grids with high solar penetration, the forecasted solar variability of the next day should also be considered.

Title: Optimizing Demand Response for Smart Grid

Demand response is widely recognized as a crucial tool for incorporating renewables into the grid. This talk presents my research efforts towards a rigorous mathematical foundation for optimizing demand response for smart grid. This is an interdisciplinary challenge for both control algorithm design for local customers and market design for utility companies and society. For local control algorithm design, I will first focus on data centers, which represents the fastest growing sectors in energy usage and greenhouse gas pollution. I have designed algorithms with theoretically provable guarantees to deal with information uncertainties and the need of distributed control. Together with Southern California Edison, we quantified the potential of data center demand response and showed that it's worth billions of dollars worldwide. Moving from theory to practice, I helped HP design and implement the industry's first Net-Zero Energy Data Center. I will then briefly discuss other demand response opportunities in buildings and electric vehicles, based on my recent work at Lawrence Berkeley National Lab. In order to make sure the demand response provided by local customers aligned well with the global objective of utility companies and society, we need to design electricity market in the right way to incentivize customers. I will present our preliminary results along this line.

Title: Fair Load-side Control for Frequency Regulation in Smart Grids

Frequency control rebalances supply and demand while maintaining the network state within operational margins. It is implemented using fast ramping reserves that are expensive and non-renewable, and which are expected to grow with the increasing penetration of renewables. The most promising solution to this problem is the use of demand response, i.e. load participation in frequency control. Yet it is still unclear how to efficiently integrate load participation without introducing instabilities and violating operational constraints.

In this talk we present a comprehensive load-side frequency control mechanism that can maintain the grid within operational constraints. Our controllers can rebal- ance supply and demand after disturbances, restore the frequency to its nominal value and preserve inter-area power ows. Moreover, our controllers are distributed (unlike generation-side), fair among participating loads, and can further maintain line ows within thermal limits. We prove that such a distributed load-side control is globally asymptotically stable and illustrate its convergence with simulation

Title: Agents Vote for the Environment: Designing Energy-Efficient Architecture

In this research-in-progress paper we present a new real world domain for studying the aggregation of different opinions: early stage architectural design of buildings. This is an important real world application, not only because building design and construction is one of the world's largest industries measured by global expenditures, but also because the early stage design decision making has a significant impact on the energy consumption of buildings. We present a mapping between the domain of architecture and engineering research and that of the agent models present in the literature. We study the importance of forming diverse teams when aggregating the opinions of different agents for architectural design, and also the effect of having agents optimizing for different factors of a multi-objective optimization design problem. We find that a diverse team of agents is able to provide a higher number of top ranked solutions for the early stage designer to choose from. Finally, we present the next steps for a deeper exploration of our questions.

Title: Integration Strategies for High Penetration of Photovoltaics Using Solar Forecasting, Control, and Optimization

As solar energy penetration in the power grid increases, utilities face great challenges coping with the variable output of solar power, especially from photovoltaics (PV) panels. To increase solar energy penetration without compromising grid operating cost and compliance with voltage standards and to maximize economic benefits of PV systems, a complex system design and control of PV inverters, energy storage systems, and existing power line components such as voltage regulators is required. The impact of solar variability on distribution feeders with high solar energy penetration is examined via quasi steady-state simulations, coupled with solar now casting using high resolution sky imager, at different solar penetration levels and cloud conditions. The adverse impacts of variable solar power include voltage fluctuations, high frequency and magnitude power ramps, and increasing tap operations. Several control schemes utilizing PV inverters, storage systems, and solar forecasting are investigated to mitigate these impacts and maximize the benefits of solar energy. Some specific approaches to be studied are: 1) local control of single inverter unit (in PV and/or ESS) for local voltage stabilization and ramp smoothing using active and reactive power compensation with solar forecasting in sub-minute time scale; 2) distributed model predictive control of real and reactive power from PV inverters and storage systems for line loss minimization, cost reduction of power provided by utilities, and ramp mitigation; 3) centralized control of real and reactive power from multiple distributed resources for global ramp rate mitigation, global voltage stabilization, reduction of tap operations and overall generation cost minimization. PV and load hosting capacity of the distribution feeder will be improved when these controls are applied appropriately. The tools used for this study are OpenDSS for quasi-steady state simulation and Matlab for data processing, control scheme design, and optimization.

Title: Beyond Binary Occupancy: Non-Intrusive Sensing of Occupancy Level and Occupant Identities

Information about the occupancy status in residential and commercial buildings is a key enabler for a number of energy efficiency and automation applications. Current systems for occupancy inference usually require the installation of dedicated sensors such as motion sensors, acoustic sensors, or cameras. In this paper, we investigate the feasibility of using digital electricity meters as an opportunistic sensor for non-intrusive sensing of fine-grained occupancy status. Our intuition is that since different occupants usually operate on different devices with their own power signatures, different patterns in the aggregate electricity consumption will reflect the number and identities of current occupants. We have collected data in a three-person family home and a twelve-person university lab, and performed an in-depth analysis of the correlation between the electricity consumption pattern and occupancy status. Our experimental results demonstrate that by leveraging the electricity consumption data, we can infer not only whether a space is occupied, but also the number and identities of occupants with sufficient accuracy, significantly better than a naive strategy that does not use such sensory data.

Co-authored with Kevin Ting and Mani Srivastava.

Title: Bioenergy Sustainability Assessment: Mathematical approaches to protocol developmen

Given the diversity of production pathways and as well as the interest in bioenergy as a renewable energy resource on a global scale, bioenergy sustainability assessment methodologies must be developed that are adaptable for assessing diverse production methods and flexible to support the range of analyses that researchers and policymakers may seek to utilize with them. These assessment goals must also be reached while maintaining mathematical rigor with respect to aggregation, normalization, and quantifying data uncertainty. This talk presents indicators for bioenergy sustainability and discusses how the study of aggregation functions can be used to address challenges in the development of bioenergy sustainability assessment protocols.

Title: Exploiting Flexible Loads for Renewable Integration

Vast and deep integration of renewable energy resources into the existing power grid is essential in achieving the envisioned sustainable energy future. Many countries around the globe as well as many states in the U.S. have set up aggressive Renewable Portfolio Standards (RPSs). The state of California, as an example, has targeted a 33% RPS by 2020. Volatility, stochasticity, and intermittency characteristics of renewable energies, however, present a challenge for integrating these resources into the existing grid in a large scale as the proper functioning of an electric grid requires an instantaneous and continuous power balance between supply and demand. In this talk, we see how the demand-side exibility can be used to match random supply. In particular, we show that the thermal storage potential of a collection of Thermostatically Controlled Loads (TCLs) (such as residential air conditioners) can be modeled as a battery and can be leveraged for enabling a deep penetration of renewable energy resources.

Title: SmartShift: Expanded Load Shifting Incentive Mechanism for Risk-averse Consumers

Peak demand for electricity continues to surge around the world. The supply-demand imbalance manifests itself in the form of rolling brownouts in California and power cuts in India. Exposing consumers to real-time pricing, in an effort to incentivize them to shift their usage pattern, is often suggested as an approach to mitigate this problem. This is akin to increasing tolls at peak commute times. We show that risk-averse con- sumers of electricity react to price fluctuations by scaling back on their total demand, not just their peak demand leading to the unintended con- sequence of an overall decrease in production/consumption and reduced economic efficiency. We propose a new scheme that allows homes to move their demands from peak hours in exchange for greater electricity con- sumption in non-peak hours. This is akin to how airlines incentivize a passenger to move from an over-booked flight in exchange for, say, two tickets in the future. We present a formal framework for the incentive model that is applicable to different forms of the electricity market. We show that our scheme not only enables increased consumption and con- sumer social welfare but also allows the distribution company to increase prots. This is achieved by allowing load to be shifted while insulating consumers from real-time price fluctuations. This win-win is important if these methods are to be incorporated in practice.

Title: Using Behavioral Science to Make an Impact with Energy Data in the Field

Smart Grid systems across the world are generating an abundance of energy data that can be leveraged at many levels to make more informed decisions about grid operations. End-users tend to be overlooked, but they are key players in the grid, and engaging them with technologies that make use of this data is critical. Hence, an optimally functioning grid couples new technology with changes in human behavior. Although grid technologies themselves vary considerably, they generally share common goals: to reduce energy consumption via efficiency and/or curtailment, to shift use to off-peak times of day, and to enable distributed storage and generation options. End-users are integral to unlocking these potentials for impact. In this presentation, we highlight several ways in which behavioral science can be applied to better understand and engage customers in smart grid systems.

Bio:

Nicole Sintov holds a Ph.D. in Psychology from USC. She is a Behavioral Scientist at the USC Information Sciences Institute and teaches courses in the USC Department of Psychology and USC Environmental Studies Program. Dr. Sintov's research focuses the development of technologies that can be used to modify human behavior in order to promote environmentally beneficial outcomes, particularly related to reducing energy consumption among end-users. Incorporating behavioral science principles such as persuasion, social influence, and goal theory, her work aims to increase adoption of smart grid technologies and achieve energy conservation and load-shifting through field studies. She has deployed numerous such community-based field programs that directly engage end-users by combining energy data with tools of behavior change. She is the recipient of a grant to investigate Customer-Focused Smart Grid Technologies from the Los Angeles Department of Water and Power and also is Co-PI of Dept. of Defense MURI award for Scalable, Stochastic and Spatiotemporal Game Theory for Real-World Human Adversarial Behavior.

Title: AMI data for Advanced Distribution Automation

More than 60% precent of Smart Grid Investment Grant (SGIG) projects involve to install the Advanced Metering Infrastructure (AMI). The main purposes for AMI installation are to support the billing procedure, customer services and to improve the operational efficiencies. AMI gathers load data at customers level, i.e. household level.

Customer load profiles show more stochastic behavior by decreasing the time scale as well as by integration of rooftop PV panels. By considering the scale of data variations in time and space, i.e. spatio-temporal data, households, data-driven methods are needed to provide the actionable information for power system operation at utility scale.

This presentation will focus on developing statistical solutions to construct the load models at distribution feeders to improve systems operations and load estimation at distribution transformers for advanced distribution automation programs, e.g. Conservation Voltage Reduction (CVR).

Title: Competition and Coalition Formation of Renewable Power Producers

We investigate group formations and strategic behaviors of renewable power producers in electricity markets. These producers currently bid into the day-ahead market in a conservative fashion because of the real-time risk associated with not meeting their bid amount. It has been suggested in the literature that producers would bid less conservatively if they can form large groups to take advantages of spatial diversity to reduce the uncertainty in their aggregate output. Using real world bid data from system operators, we show that large groups of renewable producers would act strategically to lower the aggregate output because of market power. To maximize renewable power production, we characterize the trade-off between market power and generation uncertainty as a function of the size of the groups. We quantify a sweet spot in the sense that there exists groups that are large enough to achieve the uncertainty reduction of the grand coalition, but are small enough such that they have no significant market power. We consider both independent and correlated forecast errors under a fixed real-time penalty. We also consider a real-time market where both selling and buying of energy are allowed. We validate our claims using PJM and NREL data.

Demo Abstracts:

Title: Short Term Solar Forecasting Using Sky Imagery and Its Potential Application in Control and Optimization for a Smart Grid

The variability and uncertainty of solar generation compared to more conventional fossil power sources creates a unique challenge with respect to the integration of solar resources, requiring larger regulation and reserve capacities to meet ancillary service requirements. Of particular interest to the energy industry are sudden and widespread changes in irradiance typically caused by large clouds or widespread changes in cloud cover. Reduction in the uncertainty of solar generation through accurate solar forecasting reduces solar integration costs. This demo presents a short-term solar irradiance forecast technique using a whole sky imager developed at UC San Diego. First, the forecast procedure will be presented and the performance of the system will be validated against distributed ground-based GHI data. Following that, a historical forecast and a real-time operational forecast will be demonstrated. Finally, a potential application of the forecast for control and optimization of a smart grid is shown.