Navin Sharma

Predicting solar generation from weather forecasts using machine learning

A key goal of smart grid initiatives is significantly increasing the fraction of grid energy contributed by renewables. One challenge with integrating renewables into the grid is that their power generation is intermittent and uncontrollable. Thus, predicting future renewable generation is important, since the grid must dispatch generators to satisfy demand as generation varies. While manually developing sophisticated prediction models may be feasible for large-scale solar farms, developing them for distributed generation at millions of homes throughout the grid is a challenging problem. To address the problem, in this paper, we explore automatically creating site-specific prediction models for solar power generation from National Weather Service (NWS) weather forecasts using machine learning techniques. We compare multiple regression techniques for generating prediction models, including linear least squares and support vector machines using multiple kernel functions. We evaluate the accuracy of each model using historical NWS forecasts and solar intensity readings from a weather station deployment for nearly a year. Our results show that SVM-based prediction models built using seven distinct weather forecast metrics are 27% more accurate for our site than existing forecast-based models.

Blink: managing server clusters on intermittent power

Reducing the energy footprint of data centers continues to receive significant attention due to both its financial and environmental impact. There are numerous methods that limit the impact of both factors, such as expanding the use of renewable energy or participating in automated demand-response programs. To take advantage of these methods, servers and applications must gracefully handle intermittent constraints in their power supply. In this paper, we propose blinking---metered transitions between a high-power active state and a low-power inactive state---as the primary abstraction for conforming to intermittent power constraints. We design Blink, an application-independent hardware-software platform for developing and evaluating blinking applications, and define multiple types of blinking policies. We then use Blink to design BlinkCache, a blinking version of memcached, to demonstrate the effect of blinking on an example application. Our results show that a load-proportional blinking policy combines the advantages of both activation and synchronous blinking for realistic Zipf-like popularity distributions and wind/solar power signals by achieving near optimal hit rates (within 15% of an activation policy), while also providing fairer access to the cache (within 2% of a syn- chronous policy) for equally popular objects.

Cloudy Computing: Leveraging Weather Forecasts in Energy Harvesting Sensor Systems

To sustain perpetual operation, systems that harvest environmental energy must carefully regulate their usage to satisfy their demand. Regulating energy usage is challenging if a systems demands are not elastic and its hardware components are not energy-proportional, since it cannot precisely scale its usage to match its supply. Instead, the system must choose when to satisfy its energy demands based on its current energy reserves and predictions of its future energy supply. In this paper, we explore the use of weather forecasts to improve a systems ability to satisfy demand by improving its predictions. We analyze weather forecast, observational, and energy harvesting data to formulate a model that translates a weather forecast to a wind or solar energy harvesting prediction, and quantify its accuracy. We evaluate our model for both energy sources in the context of two different energy harvesting sensor systems with inelastic demands: a sensor testbed that leases sensors to external users and a lexicographically fair sensor network that maintains steady node sensing rates. We show that using weather forecasts in both wind- and solar-powered sensor systems increases each systems ability to satisfy its demands compared with existing prediction strategies.

The case for efficient renewable energy management in smart homes

Distributed generation (DG) uses many small on-site energy sources deployed at individual buildings to generate electricity. DG has the potential to make generation more efficient by reducing transmission and distribution losses, carbon emissions, and demand peaks. However, since renewables are intermittent and uncontrollable, buildings must still rely, in part, on the electric grid for power. While DG deployments today use net metering to offset costs and balance local supply and demand, scaling net metering for intermittent renewables to many homes is difficult. In this paper, we explore a different approach that combines residential TOU pricing models with on-site renewables and modest energy storage to incentivize DG. We propose a system architecture and control algorithm to efficiently manage the renewable energy and storage to minimize grid power costs at individual buildings. We evaluate our control algorithm by simulation using a collection of real-world data sets. Initial results show that the algorithm decreases grid power costs by 2.7X while nearly eliminating grid demand peaks, demonstrating the promise of our approach.

A weighted center of mass based trilateration approach for locating wireless devices in indoor environment

This paper presents a weighted center of mass based trilateration approach for locating a wireless device based on the signal strength received from the access points at known locations. This approach mainly consists of two phases: (1) The calculation of distance from RSSI values of various access points as received by the mobile device, (2) Determination of the most probable location of wireless device using coordinates of various known or fixed access points and calculated distances of the device from those access points. Kalman filter is also used in both phases in order to remove the measurement noise component and to increase the accuracy of estimation. The proposed algorithm provides a solution for location tracking of mobile devices in indoor environment where the configuration of access points like transmit power etc., is not fixed and the movements in environment affecting attenuation of signal is so unpredictable that any mathematical modeling of indoor RF signal propagation is infeasible.

Towards continuous policy-driven demand response in data centers

Demand response (DR) is a technique for balancing electricity supply and demand by regulating power consumption instead of generation. DR is a key technology for emerging smart electric grids that aim to increase grid efficiency, while incorporating significant amounts of clean renewable energy sources. In todays grid, DR is a rare event that only occurs when actual peak demands exceed the expected peak. In contrast, smart electric grids incentivize consumers to engage in continuous policy-driven DR to 1) optimize power consumption for time-of-use pricing and 2) deal with power variations from non-dispatchable renewable energy sources. While data centers are well-positioned to exploit DR, applications must cope with significant, frequent, and unpredictable changes in available power by regulating their energy footprint. The problem is challenging since data centers often use distributed storage systems that co-locate computation and storage, and serve as a foundation for a variety of stateful distributed applications. As a result, existing approaches that deactivate servers as power decreases do not translate well to DR, since important application-level state may become completely unavailable. In this paper, we propose a DR-compatible storage system that uses staggered node blinking patterns combined with a balanced data layout and popularity-based replication to optimize I/O throughput, data availability, and energy-efficiency as power varies. Initial simulation results show the promise of our approach, which increases I/O throughput by at least 25% compared to an activation approach when adjusting to real-world wind and price fluctuations.

Towards a Virtualized Sensing Environment

While deploying a sensor network is necessary for proof-of-concept experimentation, it is a time-consuming and tedious task that dramatically slows innovation. Treating sensor networks as shared testbeds and integrating them into a federated testbed infrastructure, such as FIRE, GENI, AKARI, or CNGI, enables a broad user community to benefit from time-consuming deployment exercises. In this paper, we outline the challenges with integrating sensor networks into federated testbeds in the context of ViSE, a sensor network testbed we have integrated with GENI, and describe our initial deployment experiences. ViSE differs from typical embedded sensor networks in its focus on high-bandwidth steerable sensors.

MultiSense: fine-grained multiplexing for steerable camera sensor networks

Steerable sensors, such as pan-tilt-zoom video cameras, expose programmable actuators to applications, which steer them in different directions based on their goals. Despite being expensive to deploy and maintain, existing steerable sensor networks allow only a single application to control them due to the slow speed of their mechanical actuators. To address the problem, we design MultiSense to enable fine-grained multiplexing by (i) exposing a virtual sensor to each application and (ii) optimizing the time to context-switch between virtual sensors and satisfy requests. We implement MultiSense in Xen and explore how well proportional share scheduling, along with extensions for state restoration and request batching, satisfies the unique requirements of steerable sensors in the form of pan-tilt-zoom video cameras. We present experiments that show MultiSense efficiently isolates the performance of virtual cameras, allowing concurrent applications to satisfy conflicting goals. As one example, we enable a tracking application to photograph an object moving at nearly 3 mph every 23 ft along its trajectory at a distance of 300 ft, while supporting a security application that photographs a fixed point every 3 seconds

A distributed file system for intermittent power

Designing server clusters for intermittent power introduces new possibilities to make them cheaper, greener, and more reliable, including leveraging variable electricity prices to buy more power when it is cheap, increasing the use of clean renewable energy, and capping power at low levels to extend UPS lifetime during blackouts. However, regulating power usage to take advantage of these possibilities is challenging, since applications often access persistent distributed state, where power fluctuations impact I/O performance and data availability. To address the problem, we design and implement BlinkFS, which combines a blinking abstraction with a power-balanced data layout and popularity-based replication/reclamation to optimize I/O throughput and latency as power varies. Our experiments show that BlinkFS outperforms existing approaches, particularly at low steady power levels and high levels of intermittency. As one example of our results, we show that BlinkFS reduces completion time for MapReduce-style jobs by 42% at 50% full power compared to an existing energy-proportional DFS.

SRCP: Simple Remote Control for Perpetual High-Power Sensor Networks

Remote management is essential for wireless sensor networks (WSNs) designed to run perpetually using harvested energy. A natural division of function for managing WSNs is to employ both an in-band data plane to sense, store, process, and forward data, and an out-of-band management plane to remotely control each node and its sensors. This paper presents SRCP , a Simple Remote Control Protocol that forms the core of an out-of-band management plane for WSNs. SRCP is motivated by our target environment: a perpetual deployment of high-power, aggressively duty-cycled nodes capable of handling high-bandwidth sensor data from multiple sensors. The protocol runs on low-power always-on control processors using harvested energy, distills an essential set of primitives, and uses them to control a suite of existing management functions on more powerful main nodes. We demonstrate SRCPs utility by presenting a case study that (i) uses it to control a broad spectrum of management functions and (ii) quantifies its efficacy and performance.