Smarter utilities through machine learning

Artificial intelligence (AI) and machine learning (ML) are rapidly transforming many elements of our lives, including the utility industry.

Utilities, such as energy, gas, water, and waste management, already rely on smart devices for optimization of infrastructure and the supply-demand balance. Now smarter utilities – a whole ecosystem of technology-driven, sophisticated marketplaces – are emerging. They have a lot to gain from the use of ML. The energy sector and smart power grids, in particular, will benefit enormously from recent advances in ML and AI. 

The energy sector and the infrastructure it relies on is incredibly complex. As a result, it’s often plagued by maintenance issues, system or equipment failures, and management challenges which can be caused by a variety of factors including inclement weather, surges in demand, and misallocation of resources. In fact, it has been estimated that 61 to 86% of energy on the U.S. grid is wasted due to overloading grids and creating congestion.

The Data Conundrum

Energy grids provide a treasure trove of data, much of which can help operators triage issues as they arise. However, collecting and aggregating this data is a significant challenge given the high volume of data constantly passing through the grid. For instance, think about the signals coming from billions of different pieces of equipment and from millions of sources across the grid. As a result, it’s an incredibly daunting task for operators to keep up with this flood of data, which can often lead to missed insights that may cause a malfunction, or worse, an outage.

Properly collecting this data is only half the challenge. Once it’s collected and organized, making use of this data is a consistent headache for data scientists. A diverse group of algorithms must be built to uncover the insights needed to ensure grids run efficiently. From there, they must be constantly maintained to guarantee accuracy, which requires a significant investment in time and resources for those involved. 

How Automation Can Help

Like many other business applications, harnessing the power of ML to automate processes within data management can provide significant benefits for the energy sector. Some of the most applicable applications include:

• Predicting failures - With the right algorithms in place, operators can better predict grid failures before they reach the customer. As a result, energy companies can avoid customer dissatisfaction and the corresponding financial losses that come with it.
• Energy management - Surges in energy usage can cause major problems to the grid. By properly allocating energy based on demand, operators can enable “load management” across the grid, saving resources when needed and ultimately leading to greener, more efficient practices.
• Maintenance - Energy grids cover massive sections of the country and can often be hit with multiple different weather scenarios at the same time such as high winds in one area, lightning strikes in another, and heavy rain in an entirely different region. Being able to automate the intake of maintenance signals and predict where maintenance may be needed enables operators to prioritize work, save money, and reduce downtime. 

Putting Machine Learning in Practice

The energy sector is rapidly adopting ML capabilities to automate the way the grid is operated, creating new demands on development teams. To accomplish these goals and to stay on pace, developers need fast and easy access to ML capabilities. They can’t afford to sink weeks and weeks and into building the code and architectures required to make automation possible. 

Fortunately, there are solutions available to make this possible, enabling developers to quickly test ML-powered applications seamlessly and without system interruption. By putting ML within the reach of application developers, teams can get to value faster. Embedding ML within data management platforms is a way of enabling developers, and it ultimately empowers data science teams to spend more time innovating and less time building and maintaining. As the sector becomes more advanced, so will the ML operations coming into play, which will lead to a smarter grid, more efficient operators, and happier customers. 

This article was contributed by Jeff Fried, Director of Product Management, InterSystems