How AI For Good is changing the energy landscape

Co-founder and CEO of BidgelyAdvancing energy analytics for utilities with the power of data and artificial intelligence.

Artificial intelligence has quickly become one of the most disruptive technologies in the world, sparking both excitement and concern about what AI means for societal advancement.

On the one hand, AI is improving productivity in unimaginable ways, from automatically weeding and sorting produce on farms to cancer screening. On the other hand, abuse of AI has enabled threats like cybersecurity breaches, and just recently Italy temporarily banned ChatGPT for privacy reasons. As AI becomes more sophisticated and integrated into our everyday lives, it is imperative that we draw a line on how AI could – and should – be used.

If AI is used for selfish purposes, the consequences could be dire. But when used for good, AI can solve the world’s most pressing challenges and ensure that people and the planet thrive in harmony. By examining where AI can bring the greatest benefit, we can prioritize enabling greater efficiency and accessibility to critical resources. Energy, including how it is obtained and how we use it on a daily basis, offers numerous possibilities.

In recent years I have written extensively about the advances in the energy industry. From increasing support for underrepresented customer groups, to improving grid management in the face of DERs, to identifying EV ownership and reducing energy theft, AI is helping to optimize operations across the utility. Recently, we are seeing AI being used to achieve global decarbonization goals.

While this range of use cases is wide, each represents an opportunity where AI continues to prove progress “forever.” They also share a common component: energy algorithms — and more specifically, energy algorithms applied to customer data.

What is an energy algorithm?

Basically, AI-powered energy algorithms take large amounts of data from a utility meter and turn it into actionable insights, breaking down key energy usage data by device by device or by category. The algorithm then determines which devices and energy habits contribute to the total consumption of a household or company.

Energy algorithms can use a combination of energy usage data, weather data, and housing data to identify and track actual customer behavior based on actual daily device usage. For example, when a customer goes on vacation or buys an electric vehicle, their usage behavior changes dramatically. AI can detect changes in usage patterns and adjust energy listings accordingly.

Imagine that every appliance usage has its own unique fingerprint. AI extracts these fingerprints to create an energy profile for each individual home or business. The level of detail of these fingerprints depends on the sophistication of the algorithm.

No detail too small

The true value of energy algorithms is in the detail. The most sophisticated AI can identify what appliances a customer has, their energy consumption, how long they are used, what appliance size and fuel type (electric or gas), and which appliance type uses the same fuel type (e.g. central air conditioning or room Air conditioner).

This granularity allows for a more complete understanding of energy usage patterns, enabling consumers and utilities to make more informed decisions.

How smart is your AI?

For utilities looking to use energy AI to gain new insights into their customers’ usage, there are a few things to consider when choosing the right AI model for your business.

First, you want to compare which AI models have been trained the longest. As with any AI, the longer it’s in use, the smarter the output. You may also want to ask how many data points are collected each day. Second, consider the advertised accuracy and effectiveness of the energy algorithms, as these can vary significantly depending on the approach used.

For example, if you want to determine if a device is present in a home and there are 30 use cases for that device, by detecting five of them you can conclude with high confidence that the home has that device, resulting in the following Result: Nearly 100% home-level recognition accuracy. However, if you define the detection accuracy to detect each occurrence of the device individually, the detection would only be 16 percent accurate five times out of 30.

You can also verify accuracy by asking AI vendors for their sample size. Measuring accuracy in small sample sets, say five or ten houses, seems easy and accurate because algorithms can be tailored for these specific cases. However, scaling to hundreds of thousands of households results in different data scenarios that can impact accuracy.

Depending on your specific needs, you should also consider the frequency of insights. Some AI algorithms provide data points on a monthly basis, while others can provide yearly, monthly, weekday, and weekend trends.

It can also be worth looking for providers who have worked with similar utilities to yours, whether in terms of area size, number of customers, or even similar end goals.

Utility AI forever

AI has proven its importance in today’s world and the value of using it for good. As the energy landscape has evolved, with consumers moving from static tariff payers to more active participants, AI is giving utilities a deeper understanding of how consumers use energy every day in order to better serve them. With an improved ability to address this new paradigm of supply and demand, energy providers are creating positive relationships with customers that result in greater energy efficiency, less strain on the grid and lower carbon emissions.

As we continue to learn about the benefits and limitations of AI in energy, we see utilities tackling major energy roadblocks—from grid planning and load shifting to energy efficiency and decarbonization—with ease, and most importantly, from a future-ready perspective.

The Forbes Technology Council is an invitation-only community for top-notch CIOs, CTOs, and technology executives. Am I Qualified?

Leave a Comment