Original article was published on Artificial Intelligence on Medium
MACHINE LEARNING IN ENERGY — PART III
Machine Learning for Energy Distribution
9 use cases for energy distribution companies
Distributed production, the rise of renewables, the move to a smarter grid and competitive marketing is changing the energy distribution market and putting pressure on the profit margins of utilities.
There is an increased need to make smarter decisions on a scale. And these decisions need to be made fast to remain competitive. Machine learning is becoming the most important tool to help you make better decisions around pricing and create better relationships with your customers. With machine learning, you can:
- Predict prices and demand,
- Optimize retail prices,
- Make better offers, win more customers, reduce churn and predict customer lifetime value,
- Predict the merit order and optimize consumption
Accurately Predict Energy Prices
As personal power generation (using solar or wind power) gets easier and cheaper, consumers and businesses are increasingly producing their own power.
Personal power generation allows people to make, consume, and store their own energy. Depending on where they live, they may even be able to sell surplus power back to the local power utility.
Machine learning can help find the best time to produce, store, or sell this energy. Ideally, energy should be consumed or stored when prices are low and sold back to the grid when prices are high.
When we use machine learning models to look at historical data, usage trends, and weather forecasts, we can make much more accurate predictions on an hourly basis. This helps people with personal and business energy generation systems make strategic decisions about what to do with their energy.
For example, Adaptive Neural Fuzzy Inference System (ANFIS) has been used to predict short-term wind patterns for power generation. This allows producers to maximize energy production, and sell it back to the grid when prices are at their peak.
Accurately Predict Energy Demand
Accurately predicting customers’ energy needs is critical for any utility provider. To date, there is no adequate solution for bulk energy storage, which means energy needs to be transmitted and consumed almost as soon as it’s produced.
We’re using machine learning to increase the accuracy of these predictions. Historical energy use data, weather forecasts, and the types of businesses or buildings operating on a given day all play a role in determining how much energy is used.
For example, a hot summer day mid-week means more energy usage because office buildings run air conditioning at a high capacity. Weather forecasts and historical data can help identify those patterns in time to prevent rolling blackouts caused by air conditioners in the summer.
Machine learning finds complicated patterns in the various influencing factors (such as day of the week, time, predicted wind and solar radiation, major sports events, past demand, mean demand, air temperature, moisture and pressure, and wind direction) to explain developments in demand. Because machine learning finds more intricate patterns than humans can, its predictions are more accurate. This means you can increase efficiency and decrease costs when you buy energy — without having to make expensive adjustments.
Optimize Prices Through Better Trading
Providing the best prices for a commodity like electricity is key to surviving in an open energy market. When consumers can choose who provides their electricity, price comparison is inevitable.
To stay competitive, utility providers use machine learning to determine the best times to buy energy, based on when the price is lowest.
When it comes to commodity tracing, there are thousands of factors that affect energy prices — everything from the time of day to the weather. We can use machine learning to analyze tiny changes in these factors, which helps utility providers make more informed decisions about when to buy and sell energy.
But we can go further, for one of our clients we combined gradient boosting, general additive and deep learning models to more accurately predict day-ahead prices. We also used machine learning to help a B2B energy distributor pick trading strategies to buy futures cheaper.
In some markets, this kind of analysis has led to multiple price reductions for consumers in a single year.
Reduce Customer Churn
In open energy markets, where customers have a choice of utility providers, understanding which customers are going to churn out can be critical. In the energy sector, churn rates — the percentage of customers who stop using a service in a given year — can be as high as 25%. Accurately predicting and preventing churn is essential to survival.
We can use machine learning to help utility owners anticipate when a customer is getting ready to churn out.
By using techniques such as Cross-Industry Standard Process for Data Mining (CRISP-DM), AdaBoost, and Support Vector Machines, as well as historical usage data, utility providers can identify key indicators that predict whether a customer is going to churn.
These include things like customer satisfaction, employment status, energy consumption, home ownership, or rental status. A change in any of these factors can point to a customer who’s getting ready to terminate their service.
When we identify these indicators far enough in advance, you can prevent churn by working with customers to solve any problems they’re experiencing.
Predict Customer Lifetime Value
In an open utilities market, utility owners and providers have to pay more attention to metrics like customer lifetime value (CLV). This helps them understand how much any given customer is going to spend over the term of their contract.
Machine learning can do more than just giving you a more accurate CLV prediction. By inputting data like customer information, consumption habits, location, purchase history, and payment behavior, we can use machine learning models like deep neural networks to predict the overall value of an individual customer.
With machine learning, we can even take this one step further by suggesting ways to increase customer value. This could mean making highly-targeted offers to similar customers, or leveraging natural language processing (NLP) to help improve service for frustrated customers who might be ready to churn.
Predict the Probability of Winning a Customer
For utility providers in open markets, having a complete picture of potential customers is critical to staying ahead of the competition.
But machine learning can do more than offer you this complete picture. It can also provide the information you need to make data-driven marketing decisions.. This means that when a person lands on your website, you’ll be able to figure out whether that person will become a customer.
We can use machine learning to gather the information that person brings with them — things like where they live, what kind of computer they’re using, their browsing history, search history, and how many times they’ve been to your website — and come up with an accurate picture of that person as a consumer.
From there, not only can machine learning determine the likelihood of that person becoming a customer (this is known as scoring), it can also determine the best strategy for turning them into a customer. This involves using highly targeted advertising and delivering a very personalized experience — for example, using a picture of a family of four on the homepage when someone from a family of four visits the site.
Make Highly Targeted Offers to Customers
In open energy markets, consumers have a choice of utility providers. Personalized offers are essential to enticing new customers and keeping existing ones, especially since brand loyalty isn’t as strong as it used to be.
We can use machine learning to help you gain the insight you need to make irresistible offers that speak directly to a specific customer’s needs.
By analyzing spending habits and customer data, machine learning can help you determine the best kind of offer to make to a certain customer at any given time.
For example, if the data indicates that a customer is getting ready to move, you could send out an offer waiving the connection fee at their new location. This kind of personalized offer puts you ahead of the competition and makes the customer less likely to churn.
Optimize Energy Consumption
People have long been conscious of energy consumption, both at home and at work. But without doing a lot of manual calculations, we’ve only ever been able to get an overall sense of energy use, without knowing which appliances or devices use the most.
Smart meters and the rise of Internet of Things devices have changed all that. Non-intrusive appliance load monitoring (NIALM), also known as disaggregation, is an algorithm that uses machine learning to analyze energy consumption at the device-specific level.
We can help you figure out which appliances cost the most to operate. This will help both home and business customers fine-tune their consumption habits to save money and reduce energy use. They can either use high-cost appliances less often or replace them with more energy-efficient models.
Predict Merit Order of Energy Prices
Utility providers have a lot of options when it comes to sourcing where they get their energy — everything from renewables like wind and solar to fossil fuel and nuclear. When it comes time to sell power, these different sources are organized into a merit order based on price. This determines the order in which power from these various sources is sold.
Because we have access to data from a wide variety of sources, we can use machine learning to analyze both real-time data and historical data. Machine learning algorithms are also better at taking into account all the different factors that influence the price –things like weather, demand, how much energy is available from the various sources, historical usage, etc.– to predict an optimized merit order.
This helps you make more informed decisions about where you’re getting your power. This is especially helpful in markets where there is a lot of renewable energy, such as wind, because it’s hard to guarantee energy availability from these sources.