Machine Learning Use Cases

August 5th, 2016 by Matthew Ross

Over the past three weeks, Energyworx has taken its readers on a journey to better understand Machine Learning and what it means for the Energy Industry. Part 1 provided the basics, Part 2 covered Cloud Machine Learning and Part 3 discusses some example use cases.

The Energy Transition is bringing new distributed energy resources onto the grid, increasing digitalization and creating a deep desire to automate tasks. These new technologies provide the foundation for new revenue generating and cost cutting models, allowing utilities to shift from a commodity driven business model to one based on value-add services powered by data.

Energyworx full company pitchdeck (MASTER)

Machine Learning is poised to play a significant role in automation and new business model creation. Datasets are growing exponentially within utility and energy companies, providing the fuel for new model creation and training. Let’s take a closer look at three viable use cases.

Predictive Maintenance

Predictive MaintenanceDespite all the grid modernization, a large portion of the grid’s infrastructure is aging and functioning beyond its intended service life. Failing equipment that is not replaced in time can lead to dangerous malfunctions and downtime. On the other hand, replacing equipment too early leads to unnecessary out-of-pocket costs.

Analytical models have been used for years to better predict when to service equipment. Transformers, for example, can fail catastrophically if they are overloaded. Analysis of historic and real-time loading data can inform utilities to redistribute load or replace and resize the transformer.

Machine Learning offers three main benefits for predictive maintenance:

  1. Utilizing historical data to identify anomalies and patterns that lead to failure
  2. Model training using utility and open datasets: historical (event) data, consumer/substation loads, weather, etc.
  3. Automate maintenance schedules and integration with workflow management

Customer Load Profiling

Customer ProfilesFor decades, load profiles were pretty similar for residences and saw predictable variation in the C&I space. With the introduction of DERs and new energy efficiency programs (e.g. demand response), load profiles are changing dramatically. More precision is needed for Energy Retailers to make competitive offerings.

Just as we learned in Part 1 how Spotify uses Clustering to create sub-genres like “Deep Discofox” for better music recommendations, Energy Retailers can use similar techniques to create new customer load profiles like “EV + Solar” to provide more competitive rate structures.

Load profiling is quickly becoming the most accurate method for load forecasting – whether for trading or planning purposes. Machine Learning can build on this best practice in follow ways:

  1. Utilize clustering and classification techniques to create precise profiles
  2. Create short- and long-term energy and demand forecasts
  3. Predict how new strategies and (third party) energy efficiency measures impact load
  4. Provide customers with more competitive rate structures

Microgrid Coordination

MicrogridOften viewed as the holy grail of DERs integration and grid reliability, microgrids have mostly been implemented as pilot projects or in remote regions. This is mainly due to the cost and complexity of implementing microgrids but there is also uncertainty in how DERs will interact in a microgrid, how microgrids will interact with other microgrids and how all this impacts the “macro” grid.

These questions need to be answered (with technology) before microgrids will become prevalent. Each resource – both generation and load – within the microgrid has its own capabilities and communication methods which results in great opportunity met with a complex data problem. The increasing frequency and diversity of data needs to be interpreted and acted upon in real-time to reap all the benefits promised by microgrids.

This is an ideal use case for Cloud Based Machine Learning to provide more effective microgrid resource coordination through the following methods:

  1. Define the most important parameters (e.g. cost of energy, reliability, load shape, etc.)
  2. Coordinate the response of DERs based on desired parameters at that moment in time
  3. Automate customer load response and distribute energy cost effectively
  4. Securely connect the data with third parties for energy efficiency and specialized grid services

Curious how Machine Learning could impact your organization? Send us a message and we’ll take a closer look at your situation and explore some relevant use cases.

Matthew Ross
Director Business Support USA

Subscribe to our newsletter

Let’s keep in touch and allow us to send you updates once in a while. Subscribe to our newsletter.

Subscribe!

Related