forecasting

Growth vs Behavior in Power Demand Forecasting

cars road
cars road
cars road

In our previous article we talked about what forecasting is, why it matters, and that at Endre we tackle this from the perspective of maximizing the information about the future, not only predicting specific events.

Today we will dive deeper down the rabbit hole of forecasting and specifically talk about two very important aspects when it comes to forecasting for power grid planning: growth and behavior.

As mentioned in our first article, forecasting today tends to be about predicting events and trends. In the energy system this often results in predictions about how many electric vehicles there will be in the future, the number of solar PV systems that will be installed, or the size of the Battery Energy Storage (BES) market. 

This is obviously important information for a power grid operator to know today, but is it enough? Assume that would like to analyze the impact of Electric Vehicle (EV) charging in an area.

The data analysts in your company have developed a very good model to forecast the number of EVs. You understand the model’s main limitations and assumptions and the performance is spot-on. You are confident in its output. But you wanted an analysis of EV charging, not the number of EVs. So you are now left with calculating the EVs charging demand. Your data analyst has done a great job though, he found a good research article: that presents the charging demand for each EV. Perfect! You then know the number of EVs and how much each vehicle will charge. You have everything you need. Or?

 One problem is that the research article you read has already made assumptions about how these vehicles will charge. Assumptions that you are likely not aware of. A recent research publication found that small changes in assumptions could increase charging demand from EV with more than 100%. And this didn’t consider two additional major factors: variation in driving demand and the likelihood that high charging demand occurs during high peak hours.

In Sweden we have an annual winter break in February. During this period, many families leave for vacation homes in the northern part of the country to ski. EVs suddenly drive many hundreds of km resulting in large charging peaks, and famous queues at charging locations. Since this occurs during winter, the likelihood of high-power demand due to heating is high. Variation in driving fluctuates significantly between different locations, month-to-month variations of up to 50% are not uncommon.

What does this mean for your EV charging demand forecast?

Even if you made a perfect forecast model for the number of EVs, your charging demand forecast can be off with more than 200%! If that is, you haven’t considered local driving variations, likelihood of simultaneity with peak load, and assumptions about charging behavior.

 At Endre we focus on maximizing information on behavior for our customers using alternative data sources, which constitutes the core of our digital societal twin. If you want to learn more about how we work with behavior aspects of EV charging demand or power grid forecasting, don’t hesitate to reach out! 


In our previous article we talked about what forecasting is, why it matters, and that at Endre we tackle this from the perspective of maximizing the information about the future, not only predicting specific events.

Today we will dive deeper down the rabbit hole of forecasting and specifically talk about two very important aspects when it comes to forecasting for power grid planning: growth and behavior.

As mentioned in our first article, forecasting today tends to be about predicting events and trends. In the energy system this often results in predictions about how many electric vehicles there will be in the future, the number of solar PV systems that will be installed, or the size of the Battery Energy Storage (BES) market. 

This is obviously important information for a power grid operator to know today, but is it enough? Assume that would like to analyze the impact of Electric Vehicle (EV) charging in an area.

The data analysts in your company have developed a very good model to forecast the number of EVs. You understand the model’s main limitations and assumptions and the performance is spot-on. You are confident in its output. But you wanted an analysis of EV charging, not the number of EVs. So you are now left with calculating the EVs charging demand. Your data analyst has done a great job though, he found a good research article: that presents the charging demand for each EV. Perfect! You then know the number of EVs and how much each vehicle will charge. You have everything you need. Or?

 One problem is that the research article you read has already made assumptions about how these vehicles will charge. Assumptions that you are likely not aware of. A recent research publication found that small changes in assumptions could increase charging demand from EV with more than 100%. And this didn’t consider two additional major factors: variation in driving demand and the likelihood that high charging demand occurs during high peak hours.

In Sweden we have an annual winter break in February. During this period, many families leave for vacation homes in the northern part of the country to ski. EVs suddenly drive many hundreds of km resulting in large charging peaks, and famous queues at charging locations. Since this occurs during winter, the likelihood of high-power demand due to heating is high. Variation in driving fluctuates significantly between different locations, month-to-month variations of up to 50% are not uncommon.

What does this mean for your EV charging demand forecast?

Even if you made a perfect forecast model for the number of EVs, your charging demand forecast can be off with more than 200%! If that is, you haven’t considered local driving variations, likelihood of simultaneity with peak load, and assumptions about charging behavior.

 At Endre we focus on maximizing information on behavior for our customers using alternative data sources, which constitutes the core of our digital societal twin. If you want to learn more about how we work with behavior aspects of EV charging demand or power grid forecasting, don’t hesitate to reach out! 


In our previous article we talked about what forecasting is, why it matters, and that at Endre we tackle this from the perspective of maximizing the information about the future, not only predicting specific events.

Today we will dive deeper down the rabbit hole of forecasting and specifically talk about two very important aspects when it comes to forecasting for power grid planning: growth and behavior.

As mentioned in our first article, forecasting today tends to be about predicting events and trends. In the energy system this often results in predictions about how many electric vehicles there will be in the future, the number of solar PV systems that will be installed, or the size of the Battery Energy Storage (BES) market. 

This is obviously important information for a power grid operator to know today, but is it enough? Assume that would like to analyze the impact of Electric Vehicle (EV) charging in an area.

The data analysts in your company have developed a very good model to forecast the number of EVs. You understand the model’s main limitations and assumptions and the performance is spot-on. You are confident in its output. But you wanted an analysis of EV charging, not the number of EVs. So you are now left with calculating the EVs charging demand. Your data analyst has done a great job though, he found a good research article: that presents the charging demand for each EV. Perfect! You then know the number of EVs and how much each vehicle will charge. You have everything you need. Or?

 One problem is that the research article you read has already made assumptions about how these vehicles will charge. Assumptions that you are likely not aware of. A recent research publication found that small changes in assumptions could increase charging demand from EV with more than 100%. And this didn’t consider two additional major factors: variation in driving demand and the likelihood that high charging demand occurs during high peak hours.

In Sweden we have an annual winter break in February. During this period, many families leave for vacation homes in the northern part of the country to ski. EVs suddenly drive many hundreds of km resulting in large charging peaks, and famous queues at charging locations. Since this occurs during winter, the likelihood of high-power demand due to heating is high. Variation in driving fluctuates significantly between different locations, month-to-month variations of up to 50% are not uncommon.

What does this mean for your EV charging demand forecast?

Even if you made a perfect forecast model for the number of EVs, your charging demand forecast can be off with more than 200%! If that is, you haven’t considered local driving variations, likelihood of simultaneity with peak load, and assumptions about charging behavior.

 At Endre we focus on maximizing information on behavior for our customers using alternative data sources, which constitutes the core of our digital societal twin. If you want to learn more about how we work with behavior aspects of EV charging demand or power grid forecasting, don’t hesitate to reach out! 


In our previous article we talked about what forecasting is, why it matters, and that at Endre we tackle this from the perspective of maximizing the information about the future, not only predicting specific events.

Today we will dive deeper down the rabbit hole of forecasting and specifically talk about two very important aspects when it comes to forecasting for power grid planning: growth and behavior.

As mentioned in our first article, forecasting today tends to be about predicting events and trends. In the energy system this often results in predictions about how many electric vehicles there will be in the future, the number of solar PV systems that will be installed, or the size of the Battery Energy Storage (BES) market. 

This is obviously important information for a power grid operator to know today, but is it enough? Assume that would like to analyze the impact of Electric Vehicle (EV) charging in an area.

The data analysts in your company have developed a very good model to forecast the number of EVs. You understand the model’s main limitations and assumptions and the performance is spot-on. You are confident in its output. But you wanted an analysis of EV charging, not the number of EVs. So you are now left with calculating the EVs charging demand. Your data analyst has done a great job though, he found a good research article: that presents the charging demand for each EV. Perfect! You then know the number of EVs and how much each vehicle will charge. You have everything you need. Or?

 One problem is that the research article you read has already made assumptions about how these vehicles will charge. Assumptions that you are likely not aware of. A recent research publication found that small changes in assumptions could increase charging demand from EV with more than 100%. And this didn’t consider two additional major factors: variation in driving demand and the likelihood that high charging demand occurs during high peak hours.

In Sweden we have an annual winter break in February. During this period, many families leave for vacation homes in the northern part of the country to ski. EVs suddenly drive many hundreds of km resulting in large charging peaks, and famous queues at charging locations. Since this occurs during winter, the likelihood of high-power demand due to heating is high. Variation in driving fluctuates significantly between different locations, month-to-month variations of up to 50% are not uncommon.

What does this mean for your EV charging demand forecast?

Even if you made a perfect forecast model for the number of EVs, your charging demand forecast can be off with more than 200%! If that is, you haven’t considered local driving variations, likelihood of simultaneity with peak load, and assumptions about charging behavior.

 At Endre we focus on maximizing information on behavior for our customers using alternative data sources, which constitutes the core of our digital societal twin. If you want to learn more about how we work with behavior aspects of EV charging demand or power grid forecasting, don’t hesitate to reach out! 


Smarter grid planning using societal insights


Email hello@endre.tech

Visit us at Läraregatan 3, 411 33 Gothenburg, Sweden

Post address Stena Center, 412 92 Gothenburg, Sweden

Copyright 2025 | Endre Technologies

Smarter grid planning using societal insights


Email hello@endre.tech

Visit us at Läraregatan 3, 411 33 Gothenburg, Sweden

Post address Stena Center, 412 92 Gothenburg, Sweden

Copyright 2025 | Endre Technologies

Smarter grid planning using societal insights


Email hello@endre.tech

Visit us at Läraregatan 3, 411 33 Gothenburg, Sweden

Post address Stena Center, 412 92 Gothenburg, Sweden

Copyright 2025 | Endre Technologies

Smarter grid planning using societal insights


Email hello@endre.tech

Visit us at Läraregatan 3, 411 33 Gothenburg, Sweden

Post address Stena Center, 412 92 Gothenburg, Sweden

Copyright 2025 | Endre Technologies

Smarter grid planning using societal insights


Email hello@endre.tech

Visit us at Läraregatan 3, 411 33 Gothenburg, Sweden

Post address Stena Center, 412 92 Gothenburg, Sweden

Copyright 2025 | Endre Technologies