Smart Grid News and Analysis

IoT World Forum 2016
IoT World Forum 2016

AutoGrid Systems Sets World Record for Big Data and the Grid: Over One Million Power Forecasts Every 10 Minutes


AutoGrid Systems, a leader in big data analytics and cloud computing solutions for the energy industry, has established an industry defining benchmark for forecasting power consumption across the grid with predictive analytics.

AutoGrid’s Energy Data Platform (EDP) can accurately forecast the power consumption of over one million endpoints simultaneously every 10 minutes on a medium-size cluster running on commodity hardware servers. Endpoints include households, commercial buildings, individually metered office suites, factories and other locations or devices linked to the grid and communication networks. The performance scales linearly, without any compromises in accuracy, by simply adding more servers to a cluster.

The benchmark was established under a grant awarded to AutoGrid by the Department of Energy’s Advanced Projects Research Agency-Energy (ARPA-E) under the Green Electricity Network Integration (GENI) program to demonstrate how predictive analytics, hosted on the cloud and built around open standards, can dramatically reduce the cost of implementing demand response programs and provide grid operators greater flexibility and granularity while allowing them to offer more ways for electricity end-users to participate in these programs to reduce electricity cost. Lawrence Berkeley Laboratory and Columbia University also participated in the project.

“Forecasting is one of the fundamental challenges that grid operators face when trying to balance supply and demand in real-time,” said ARPA-E Program Director, Dr. Tim Heidel. “The ability to perform accurate, bottoms-up forecasts for individual customers and aggregate them at every transformer, feeder, substation point, or every node on the transmission grid has the potential to transform how the power grid is managed while dramatically lowering the cost of operating expenses across the entire electricity supply chain.”

To put the forecasting benchmark in perspective:
• A single DROMS cluster can generate more than 144 million forward-looking forecasts regarding energy consumption per day and over 52.5 billion forecasts in a year.
• Five DROMS clusters would be capable of monitoring and managing power at every commercial and industrial site in the U.S. equipped with smart meters.
• DROMS can scale clusters linearly to handle exponentially increasing amounts of data, generating forecasts and implementing controls in real-time.

AutoGrid’s forecasting technology is available as a cloud-based API to any application using AutoGrid’s Energy Data Platform (EDP). AutoGrid’s Demand Response Optimization and Management System uses the technology to provide accurate forecasts on the amount of load shed available for a given demand response event based on variables such as weather and price incentives. The forecasts become more and more accurate by using machine learning techniques to learn from past user behavior. DR program administrators (utilities or aggregators) can use these forecasts to accurately bid available DR resources into energy markets and to optimize DR event dispatch across multiple programs during a DR season to improve overall yield. Given the scale and speed of these forecasts, DR can be employed to manage spinning reserves, frequency regulation, voltage and reactive power optimization to improve grid reliability in the presence of increasing penetration of intermittent wind and solar generation resources. These programs have a much more stringent dispatch and response time requirements than traditional demand response programs that target peak capacity and can be dispatched days or several hours in advance.

“We are proud of our achievements under the ARPA-E project and its successful completion. The ARPA-E program gave us the tools to innovate and fundamentally re-think how predictive analytics and Internet-scale computing can reshape the future of the electric power industry,” said Dr. Amit Narayan, founder and CEO of AutoGrid. “Not only can cloud-based solutions be deployed more economically and more rapidly than traditional systems, big data analytics provide more insight and control to systems operators to improve performance, security and reliability.”
Cost-effective forecasting systems are critical to reducing energy spikes in consumption during peak hours, as well as recovering from natural disasters and outages. The Federal Energy Management Commission (FERC) estimates that demand response programs could potentially reduce peak power demand in the U.S. by up to 20 percent. DROMS cost 90 percent less than conventional, hardware-based demand response systems to implement and can yield up to 30 percent more energy per event.

As part of the project, AutoGrid collaborated with Lawrence Berkeley National Labs and other organizations within the OpenADR Alliance to show how open standards-based systems can enable new technologies to come online seamlessly and in a cost-effective manner. With the industry’s first and only OpenADR 2.0b certified, fully backward-compatible Virtual Top Node server, AutoGrid can signal millions of devices from hundreds of manufacturers, ranging from residential thermostats to Building Management Systems (BMS), as well as incorporate emerging technologies, such as LED lighting, battery and electric vehicle charging stations, into grid management programs. In addition to significantly reducing costs and providing more flexibility to utilities and their end-customers in program design, the benefits of AutoGrid DROMS include a much faster speed-to-implementation: Projects can be deployed in days as opposed to months using the cloud infrastructure and can seamlessly scale from small initial deployments to full-scale programs involving millions of endpoints.