What is Machine Learning?

July 22nd, 2016 by

Part 1: Exploring the Basics

Over the next three weeks, Energyworx will take its readers on a journey to better understand Machine Learning and what it means for the Energy Industry. Part 1 will dive into the basics and look at some best practices.


At Energyworx, we are not immune to using the latest buzzwords – whether it’s Big Data, Cloud or SaaS –  these terms are impossible to avoid as an industry innovator. Machine Learning is no exception and is why we are taking the time to explain the term and how we’re approaching the field with a little help from our friends at Google.

Quick Summary

Machine Learning (ML) is defined as the field of study that gives computers the ability to learn without being explicitly programmed. This has evolved into many sub-fields since Arthur Samuel wrote that in 1959 but the core theme remains the same and has seen growing interest from the Energy Sector during the Energy Transition. With a more complex and data-rich grid, the opportunity to successfully deploy Machine Learning algorithms has never been more promising.

First, it is helpful to understand the two types of Machine Learning:

  1. Supervised Learning – where the algorithm is given sample data and told which answer is “correct”. The main goal here is to produce additional “correct” answers (e.g. Regression or Classification).
  2. Unsupervised Learning – where no labels are given to the datasets and the program/algorithm is tasked with identifying structures and relationships for the data (e.g. Clustering or Anomaly Detection).

Popular Examples

After learning the basics, it’s very clear that simple types of Machine Learning are prevalent everywhere, whether it’s predicting the market value of your home or grouping internet search results. Let’s dive a bit deeper into two examples that demonstrate Machine Learning techniques.

spotifyMost recommendation systems incorporate Machine Learning to improve their suggestions. Spotify is a good example as their recommendation system has evolved substantially since the “thumbs up” button – a form of supervised learning. Their now famous Discovery Weekly Playlist recommends new music based on how your music overlaps with playlists from other users. In addition, it uses Clustering techniques to create very specific genres such as “Deep Discofox”. For more on the techniques behind Discover Weekly, read this great article by Quartz.


Siri_icon.svg“Hey Siri, give me a Machine Learning example.” Speech Recognition has been around for decades now but the latest smartphone models have certainly increased its popularity. Speech Recognition uses several algorithms to correctly interpret spoken word. It’s surprisingly complex but follows many of the basics of Machine Learning such as pattern matching (1 on 1 word matching) and statistical analysis (grammar rules).

What does this have to do with energy?

Many of us can relate to the above examples and begin to infer how these techniques could be applied to challenges faced by utilities and energy companies. We’ll go into more detail in Part 3 of this blog series but here’s a few ML techniques that are highly relevant for the Energy Sector.

  1. Anomaly detection – pinpointing grid disturbances and locating fraud (theft)
  2. Classification – creating more precise consumer load profiles for better forecasting and new service offerings
  3. Regression – analysis of historical and real-time data for predictive maintenance and system protection for grid equipment

Next week we’ll explore Machine Learning at scale by taking a closer look at the Google Machine Learning Products and how they’re being applied in the real world. I’ll also be giving a presentation on Machine Learning for the Internet of Energy Virtual Summit 2016. Be sure to register to hear our approach and some example use cases for the Energy Sector.