No abstract

DEMAND RESPONSE (DR) TRADITIONALLY REFERS TO THE ability to curtail some electrical loads at peak times to alleviate the need for peaking generation sources. Basically, it means being able to turn loads off on command. Progress in communication protocols and technology has been extraordinary in the past decade, making cheap, fast communication widespread. Over the next decade, we expect inexpensive broadband to become ubiquitous. In addition, more and more electrical loads are equipped for communication as well as control. Together, these trends enable a new way of thinking about DR, which we call demand dispatch.Demand dispatch is the capability to aggregate and precisely control (or dispatch) individual loads on command. Unlike traditional DR, demand dispatch is active and deployed all the time, not just at peak times. Demand dispatch represents a qualitatively different approach to balancing generation and load for a power grid. We believe that demand dispatch will be an important enabling technology for incorporating ever higher levels of intermittent renewable generation on the grid.In this article we touch on some background requirements for demand dispatch and how the Internet can be used for communication and control. In addition, we review some of the basics of the operation of the electric power grid. We show how loads that meet the communication and control requirements can be aggregated and dispatched-turned on or off-to help manage the grid. Aggregated loads will be able to perform many of the same ancillary services for the grid that are provided by power plants today. We describe some benefi ts of load-based ancillary services, such as the potential for very fast response, and explain how some characteristics of load-based services differ from power plants. Finally, we give a concrete example of demand dispatch as it can be applied to plug-in electric vehicles: smart charging.

Nearly all existing counting methods are designed for a specific object class. Our work, however, aims to create a counting model able to count any class of object. To achieve this goal, we formulate counting as a matching problem, enabling us to exploit the image selfsimilarity property that naturally exists in object counting problems. We make the following three contributions: first, a Generic Matching Network (GMN) architecture that can potentially count any object in a class-agnostic manner; second, by reformulating the counting problem as one of matching objects, we can take advantage of the abundance of video data labeled for tracking, which contains natural repetitions suitable for training a counting model. Such data enables us to train the GMN. Third, to customize the GMN to different user requirements, an adapter module is used to specialize the model with minimal effort, i.e. using a few labeled examples, and adapting only a small fraction of the trained parameters. This is a form of few-shot learning, which is practical for domains where labels are limited due to requiring expert knowledge (e.g. microbiology). We demonstrate the flexibility of our method on a diverse set of existing counting benchmarks: specifically cells, cars, and human crowds. The model achieves competitive performance on cell and crowd counting datasets, and surpasses the state-of-the-art on the car dataset using only three training images. When training on the entire dataset, the proposed method outperforms all previous methods by a large margin.

No abstract