Semi-Supervised Novelty Detection with Adaptive Eigenbases, and Application to Radio Transients

Energy production, distribution, and consumption play a critical role in the sustainability of the planet and its natural resources. Electric power systems have been going through major changes that are aimed to make the energy infrastructure “smarter”, scalable, and more efficient. These new generation of smart energy grids need novel computational algorithms for supporting generation of power from wide range of sources, efficient energy distribution, and sustainable consumption. This paper argues that a fundamentally distributed approach with more local flexibility is a lot more sustainable methodology compared to the traditional centralized frameworks for analyzing and processing data. It considers the problem of predicting power generation and consumption trends over a distributed smart grid. Since power generation from solar, wind, geothermal and other renewable sources are likely to be part of many households in near future, both power generation and consumption data will be generated over a wide area network. Moreover, a good part of the communication links between the household data sources and the central server are likely to be over the wireless networks with low bandwidth and high data-plan cost. Analyzing such data (some of it privacy sensitive) in a centralized is not scalable, sometimes not privacy-preserving, and often not practical because of cost-sensitivity of the applications. This paper presents a more sustainable distributed asynchronous algorithm for constructing energy demand prediction models in a smart grid by multivariate linear regression. The paper offers the algorithm, analysis, and experimental results.