Applied Statistics Fuel the Energy Optimization Revolution

Features

  • Author: Lillian Pierson
  • Date: 05 Aug 2013
  • Copyright: Image appears courtesy of iStock Photo

In more recent years, Department of Defense (DoD) programs, like the U.S. Navy’s Incentivized Shipboard Energy Conservation (I-ENCON) program, have been making groundbreaking progress towards decreasing both their overall carbon footprints and the amount of environmental damage that naturally accompanies oil and gas consumption. This move towards improved environmental good-stewardship is justified by the cost savings that more fuel-efficient operations afford. The recent progress of these DoD energy efficiency programs is well demonstrated by I-ENCON; a program that started in 1993, but only recently has begun to realize savings upward of $45M per quarter due to ship fuel cost avoidance.

thumbnail image: Applied Statistics Fuel the Energy Optimization Revolution

There are data scientists and applied statisticians who believe big data sources can be the key to even more fuel savings. Dr. Melinda Thielbar is one of these revolutionary data scientists. Dr. Thielbar is a Ph.D. statistician with 15 years of experience in the software industry. She is currently working at IAVO Research Scientific on a project where she uses applied statistics to perform high-level power system optimization and to guide IAVO’s software engineers in the design of an automated control system for maximum energy-optimization.

Big Data and Data Science for Power System Optimization

Dr. Thielbar analyzes operational data sets in order to generate custom recommendations for power system operations that will increase the energy efficiency of U.S. Navy ships. She utilizes a unique type of “big data” in order to perform her analyses. Generally, when people refer to big data, they are referring to data that is generated from computer or internet-based transactions and events. In this implementation, however, Thielbar is utilizing big data that is generated from mechanical systems in the form of machine data “signals”.

In this implementation, however, Thielbar is utilizing big data that is generated from mechanical systems in the form of machine data “signals”.

U.S. Navy vessels support hundreds of people. Ship power plants contain multiple generators, and their propulsion plants contain multiple engines. All of these components require fuel, and they produce a huge variety, volume, and velocity of unstructured signals of non-constant correlation. Dr. Thielbar uses this signal data to generate recommendations for power system reconfiguration and energy optimization.

Statistical Models for Power System Monitoring and Energy Optimization

In this implementation, Dr. Thielbar is working with two different types of statistical models. One of the model types is designed to help her develop rules for the power system operation. The other is designed to predict and monitor the actual operations of the ship’s power system.

Statistical Models for the Development of Power System Rules

The models used to generate power system rules are able to describe the optimal number of generators needed to power the ship given the flux of operational requirements over time (i.e.; optimal power system design). These proprietary statistical models rely on standard time-series analysis to develop the basic set of power system rules. These rules must then be tested and revised repeatedly during operational cycles.

Statistical Models for Predicting and Monitoring Power System Optimization

Both linear and non-linear models are used to monitor and detect shifts in power usage of different mechanical systems across the ship. These models work together to detect differences between model-predicted values for machine signals and the real-time monitored operational values of machine signals. Significant differences between these values indicate a “state-change”, which thereby prompts the system to change the model being used to monitor the ship operations. All of these models utilize time-series analysis to auto-update parameter estimates on a continuous basis.

Predictive models used to detect state-changes in the ship’s operations are built out of the collaborative knowledge of Dr. Thielbar and experts in electrical engineering and ship operations. The design processes of these predictive models employ both linear and non-linear time series analyses, including Autoregressive Integrated Moving Average (ARIMA) models for prediction and nonlinear models for state detection.

For each type of model, Dr. Thielbar analyzes a sample of machine data to help her choose the most appropriate form of model. She then specifies error terms and how error will enter the model (included, additive, etc.). Subsequently, she performs model testing against the signal test data and makes any needed adjustments. Lastly, she tests this model against new operational data to make sure that the results of its predictive algorithms closely match real-time operational data.

Although still “cutting-edge”, the practice of utilizing sensor and control automation “big data” to optimize energy systems is becoming a more common practice across the entire Department of Defense organization.

The Next Generation Department of Defense

Melinda Thielbar’s work with applied statistics is supported by a Small Business Innovation Research (SBIR) project, and it is part of an exciting, overarching trend in the United States Department of Defense at large. U.S. Navy, Army, and Air Force departments are all investing in technological improvements that will decrease the amount of energy required to support their operations. Most of this technology is data-driven. Although still “cutting-edge”, the practice of utilizing sensor and control automation “big data” to optimize energy systems is becoming a more common practice across the entire Department of Defense organization. Dr. Thielbar’s work is another excellent example of scientists and engineers who are finding ways to protect the environment by economically incentivizing the adoption of environmentally beneficial technologies and solutions.

Related Topics

Related Publications

Related Content

Site Footer

Address:

This website is provided by John Wiley & Sons Limited, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ (Company No: 00641132, VAT No: 376766987)

Published features on StatisticsViews.com are checked for statistical accuracy by a panel from the European Network for Business and Industrial Statistics (ENBIS)   to whom Wiley and StatisticsViews.com express their gratitude. This panel are: Ron Kenett, David Steinberg, Shirley Coleman, Irena Ograjenšek, Fabrizio Ruggeri, Rainer Göb, Philippe Castagliola, Xavier Tort-Martorell, Bart De Ketelaere, Antonio Pievatolo, Martina Vandebroek, Lance Mitchell, Gilbert Saporta, Helmut Waldl and Stelios Psarakis.