学术文献库

Our nation's infrastructure for generating, transmitting, and distributing electricity - "The Grid" - is a relic based in many respects on century-old technology. It consists of expensive, centralized generation via large plants, and a massive transmission and distribution system. It strives to deliver high-quality power to all subscribers simultaneously - no matter what their demand - and must therefore be sized to the peak aggregate demand at each distribution point. Ultimately, the system demands end-to-end synchronization, and it lacks a mechanism for storing ("buffering") energy, thus complicating sharing among grids or independent operation during an "upstream" outage. Recent blackouts demonstrate the existing grid's problems - failures are rare but spectacular. Moreover, the structure cannot accommodate the highly variable nature of renewable energy sources such as solar and wind. Many people are pinning their hopes on the "smart grid" - i.e., a more distributed, adaptive, and market-based infrastructure for the generation, distribution, and consumption of electrical energy. This new approach is designed to yield greater efficiency and resilience, while reducing environmental impact, compared to the existing electricity distribution system. Initial plans for the smart grid suggest it will make extensive use of existing information technology. In particular, recent advances in data analytics - i.e., data mining, machine learning, etc. - have the potential to greatly enhance the smart grid and, ultimately, amplify its impact, by helping us make sense of an increasing wealth of data about how we use energy and the kinds of demands that we are placing upon the current energy grid. Here we describe what the electricity grid could look like in 10 years, and specifically how Federal investment in data analytics approaches are critical to realizing this vision.