Energy Policy Institute’s Sixth Annual Energy Policy Research Conference
Ecosystem discovery: Measuring clean energy innovation ecosystems through knowledge discovery and mapping techniques

https://doi.org/10.1016/j.tej.2016.09.012Get rights and content

Abstract

While the term ‘innovation ecosystem’ is often utilized, the concept is rarely quantified. Oak Ridge National Lab conducted a ground-breaking application of natural language processing, link analysis and other computational techniques to transform text and numerical data into metrics on clean energy innovation activity and geography for the U.S. Department of Energy. The project demonstrates that a machine-assisted methodology gives the user a replicable method to rapidly identify, quantify and characterize clean energy innovation ecosystems.

EPSA advanced a novel definition for clean energy innovation ecosystem as the overlap of five Ecosystem Components: 1) nascent clean energy indicators, 2) investors, 3) enabling environment, 4) networking assets and 5) large companies. The tool was created with the flexibility to allow the user to choose the weights of each of the five ecosystem components and the subcomponents. This flexibility allows the user to visualize different subsets of data as well as the composite IE rank. In an independent parallel effort, a DOE analyst in EPSA developed a short list of 22 top US clean energy innovation ecosystems; the Ecosystem Discovery tool was able to identify over 90% of the analyst-reported ecosystems. Full validation and calibration remain outstanding tasks.

The tool and the underlying datasets have the potential to address a number of important policy questions. The initial broad list of U.S. clean energy innovation ecosystems, with geographic area, technology focus, and list and types of involved organizations can help describe regional technology activities and capabilities. The implementation of knowledge discovery techniques also revealed both the potential and limitations of an automatic machine extraction methodology to gather ecosystem component data. The project demonstrates that a machine-assisted methodology gives the user a replicable method to rapidly identify, quantify, and characterize clean energy innovation ecosystems.

Introduction

On Nov. 30, 2015, as part of the Paris Conference of Parties 21, the U.S. government along with 19 other countries announced that they would double their R&D budgets for clean energy innovation over the next five years under the umbrella of Mission Innovation. In this context, EPSA required a broad list of current clean energy innovation ecosystems (IEs) in the U.S. and their characteristics. Manual IE compilation can be labor-intensive and is subject to poorly characterized completeness. EPSA and ORNL jointly tested whether knowledge discovery methods would work to find information about clean energy IEs.

While there is much competing literature on the topic of their effectiveness, our intent was not to study whether or not an ecosystem is “effective.” The purpose of this exercise was to demonstrate the feasibility of an automatic data ingest pipeline to perform text analysis, natural language processing, and link analysis to identify clean energy IEs in the U.S. in a replicable manner and, if possible, offer insights into their data characteristics. Far less literature is available with empirical data on the characteristics and composition of IEs, so this work was intended to fill this gap.

Section snippets

Theory

Per the project’s objective, we were looking for clean energy IEs that already exist and that have the ability to share resources within a close physical proximity. In addition to the human capital resource, clean energy innovations face barriers to finance due to the high capital intensity and long lead times compared to other sectors (Howell, 2015), so we deemed it important to include the five most critical components of clean energy innovation in our definition of ecosystems: nascent clean

Complete data analytics workflow

The methodology of the project was divided into three phases:

  • Collect and categorize U.S. clean energy IE data;

  • Visualize the data in a user friendly format; and

  • Create a tool that could analyze and score the data to discover and rank a broad list of IEs for a number of technologies including, but not limited to, solar, storage and wind power.

Results

This section will provide examples of output from the Ecosystem Discovery tool. As a caveat, the results of the output will differ based on the assumptions made for the weightings of the various categories. The goal of the study was to see whether this methodology would work, but not to definitively determine where each ecosystem is in the U.S. Therefore, there is no assumption that the following examples are correct, but are simply used to illustrate the capabilities of the tool (Table 2).

The

Discussion

The tool uses data to determine where ecosystems exist and the underlying scoring algorithm “ranks” the strength of the ecosystem. The user could choose the range of ecosystems that they want to see such as the top 15 in solar (See Fig. 10) or the top 125 (See Fig. 11). The user could also drill down on a certain region and get a more granular view of the surrounding CBSAs (Fig. 12).

From Fig. 10, the top ecosystems are those that would be expected from such an exercise: (1) San Francisco, CA,

Conclusion

The solar case study was based on an expert recommendation of a known ecosystem. The Ecosystem Discovery tool was initially validated by being able to find the known ecosystems, Toledo, OH, for solar; Austin, TX, for storage, and St. Louis, MO, for nuclear. While Austin, TX, and even St. Louis, MO, may have been otherwise detected by traditional means, the Toledo, OH, example best illustrates the importance and use case of this tool. The Ecosystem Discovery tool can detect lesser known

Acknowledgements

The authors thank Eric Hsieh (U.S. DOE), Hugh Chen (U.S. DOE), Hugh Ho (U.S. DOE), John Jennings (U.S. DOE), Jeff Dowd (U.S. DOE), and Jeff Alexander (RTI) for their review and input. This work was financed by the Office of Program and Innovation Policy Analysis of the Energy Policy and System Analysis Office (EPSA-52) of the U.S. Department of Energy.

Jessica Lin is a Senior ORISE Fellow to the U.S. Department of Energy in the Innovation Department of the Office of Energy Policy and System Analysis. Prior to her current role, she worked on international energy projects with the World Bank Group, primarily focusing on what role the bank could play to commercialize clean energy technologies. She started her career in the private sector, working in finance and strategic management consulting, before leading a wireless mesh technology social

References (12)

  • T. Corsatea

    Localized knowledge, local priorities and regional innovation activity for renewable energy technologies: evidence from Italy

    Reg. Sci.

    (2014)
  • Economist

    Immigration and America’s High Tech Industry The Jobs Machine

    (2013)
  • D. Google

    Developer’s Guide: Google Maps Geocoding API

    (2016)
  • D. Google

    Google Places API

    (2016)
  • Harvard Business School

    Institute for Strategy and Competitiveness

    (2016)
  • S. Howell

    Financing Constraints as Barriers to Innovation: Evidence from R&D Grants to Energy Startups

    (2015)
There are more references available in the full text version of this article.

Cited by (3)

  • Coupling mechanism and development prospect of innovative ecosystem of clean energy in smart agriculture based on blockchain

    2021, Journal of Cleaner Production
    Citation Excerpt :

    Even if all block information is obtained, the real identity of the participating nodes cannot be determined. Ecosystem refers to the overall system formed by living things and the environment, with many types, which are generally divided into semi-artificial, artificial and natural ecosystems according to their origin (Lin et al., 2016). The natural ecosystem is formed by the endogenous evolution of nature or migration over time, without any artificial development, and can be seen everywhere in nature.

  • Modeling and simulation of enterprise innovation investment decision based on system dynamics

    2018, Xitong Gongcheng Lilun yu Shijian/System Engineering Theory and Practice

Jessica Lin is a Senior ORISE Fellow to the U.S. Department of Energy in the Innovation Department of the Office of Energy Policy and System Analysis. Prior to her current role, she worked on international energy projects with the World Bank Group, primarily focusing on what role the bank could play to commercialize clean energy technologies. She started her career in the private sector, working in finance and strategic management consulting, before leading a wireless mesh technology social enterprise startup. She continues to serve as a thought leader on issues of early-stage finance and clean energy commercialization for the World Bank’s Climate Technology Program. She received her Master’s in Public Policy from the Harvard Kennedy School and her Bachelors of Science in Electrical Engineering and Computer Science from MIT.

Supriya Chinthavali has been working as a computer scientist in ORNL since March 2010. She has multiple years of experience working on U.S. DOE- and DOD-sponsored projects like VERDE (Visualizing Energy Resources Dynamically on Earth), SPIDERS, PFMT, etc., in the Energy domain at ORNL. Prior to ORN, she worked for Delphi Automotive systems as an Advanced Software Engineer and developed embedded software systems for Toyota and GM. Her research interests include data analysis and information visualization of electric grid and complex infrastructure systems and machine learning. She graduated with a B.S. in Electronics and Communication, M.S. in Automotive Embedded Systems from India (2008), and a M.S. in CSE from Georgia Institute of Technology (2015).

Chelsey Dunivan Stahl graduated from Morningside College in 2013 with a double major in Computer Science and Business Administration. She is currently a Data Analytics Software Engineer for the Computational Data Analytics Group at Oak Ridge National Laboratory focusing primarily in data mining and knowledge discovery.

Christopher Stahl graduated from Florida Southern College with a Bachelor of Science in 2011 and has been working with the Computational Data Analytics group at Oak Ridge National Laboratory since then. He currently is a Data Analytics Software Engineer who works on multiple projects with a focus on data mining, data analytics, and knowledge discovery.

Sangkeun Matt Lee received a Ph.D. degree in computer science and engineering from Seoul National University in 2012. He joined Oak Ridge National Laboratory as a post-doctoral research associate in 2013 and became a research staff member in 2015. His research interests include large-scale graph mining and analytics, big data systems and architectures, information retrieval and recommender systems, and their applications. He has experience of working on ORNL’s Lab-directed Research & Development (LDRD), U.S. DOE- and NIH-sponsored projects and developing various software tools like gm-sparql (Graph Mining Using SPARQL) and ORiGAMI (Oak Ridge Graph Analytics for Medical Innovation).

Dr. Mallikarjun (Arjun) Shankar’s research focuses on the interdisciplinary bridge between computer science and the basic and applied sciences. He is the Director of ORNL’s Compute and Data Environment for Science (CADES) institute which hosts joint scientific initiatives between the computing directorate and ORNL’s science focus areas including environmental and biological sciences, materials and neutron science, nuclear science, and the physical sciences. Arjun received his Ph.D. in computer science from the University of Illinois, Urbana-Champaign, following which he worked in industry designing and building next-generation data management infrastructures. His research in the national laboratory setting has involved designing national scale sensor networking systems, energy infrastructure monitoring and control frameworks, and mechanisms to perform integrated data analytics and modeling and simulation. He is a member of the ACM and the IEEE Computer Society.

View full text