Machine Learning for the Digital Utility

It is hard to escape the hype surrounding artificial intelligence (AI). Blockchain is possibly the only other technology to command as many headlines as AI. Yet, there is a tangible difference between what is written about the two technologies. On the one hand, despite plenty of issues with blockchain, many in the media still effusively claim it will revolutionize the world in the same way as the internet. AI receives similarly gushing praise from some quarters—typically in technology and trade press, which extol the virtue of an increasingly automated future—but can be heavily criticized in mainstream press as a threat to employment or civilization itself.

Machine learning is an AI technology that is rapidly moving into the mainstream and is high on the agenda of many utilities. This report discusses machine learning in the context of other AI and analytics technologies, provides a brief description of how it works, and examines some recent high profile achievements. While machine learning is not new in parts of the utility value chain, despite some inherent limitations to the technology, various drivers will bring it out of pockets of excellence and thrust it into many other areas of the business.

This Navigant Research report describes several use cases for machine learning and examines why machine learning has an advantage over existing analytics techniques, including customer segmentation, pricing forecasts, anomaly detection, fraud detection, and predictive maintenance. Future requirements for machine learning—specifically for distributed energy resources (DER) management and transactive energy—are also discussed, as are several recommendations for utilities developing their machine learning strategies.

Key Questions Addressed:
  • What is machine learning?
  • How do machine learning and artificial intelligence (AI) differ?
  • Why does machine learning get so much good and bad press?
  • Which use cases are most suited to machine learning?
  • Where could machine learning be applied in the future?
Who needs this report?
  • Hardware vendors
  • Software vendors
  • Analytics vendors
  • Utilities
  • Regulators
  • Trade unions
  • Investor community

Table of Contents

1. Executive Summary

2. Machine Learning in the Age of AI

2.1   Is Machine Learning AI? More Importantly, Does It Matter?

2.2   Historic Definitions of Machine Learning

2.3   Machine Learning Relies on High Quality Data

2.4   Training Sets Machine Learning Apart from Other Analytics

2.5   Machine Learning Will Rapidly Become Deep Learning

3. Machine Learning Is Coming of Age in the Energy Industry

3.1   Technology Advancements Make the Case for Machine Learning

3.2   The Market Becomes More Open to Machine Learning

3.3   Machine Learning Is Not, and Never Will Be, a Panacea

4. The Many Use Cases for Machine Learning in Energy

4.1   Clustering

4.2   Regression

4.3   Classification

4.4   Future Energy Markets Will Increasingly Rely on Machine Learning

5. Recommendations

5.1   Manage Employees’ Antipathy to AI

5.2   Remember that Machine Learning Has Limitations

5.3   Place Machine Learning in the Context of Other Analytics Tools

5.4   Procure Machine Learning as Part of a Wider Strategy

5.5   Bake Data Management and Security into Analytics Strategies from the Outset

List of Charts and Figures

  • Analytics Maturity
  • New Products and Services Have Increasing Demand for Analytics

{"userID":"","pageName":"Machine Learning for the Digital Utility","path":"\/research\/machine-learning-for-the-digital-utility","date":"1\/18\/2018"}