Close

Belgian DSO

THE CHALLENGE

Enable accurate energy forecasting to guarantee efficient, safe and reliable power distribution

Our client is a large Belgian electricity and gas Distribution System Operator (DSO). Its primary function is to transport electricity and gas to users via medium and low voltage (electricity), and medium and low pressure (gas) networks. They are responsible for safe and continuous supplies to all users.

 

Why forecasting matters

Accurate consumption and system load forecasting is essential for DSOs, so they can:

  • properly size the distribution grid
  • estimate the long term purchase of energy
  • determine the distribution tariffs
  • spot anomalies in consumption/load
  • respond to public regulator requirements
  • foresee necessary future investments.

What makes forecasting difficult

Energy consumption is always a moving target, but in recent years has become increasingly difficult for DSOs to predict. The energy sector is in transition as the world moves from fossil fuels to more sustainable energy choices. Consumers have become prosumers, actively involved in changing their consumption behavior in their homes and businesses – from installing solar panels or alternative heating systems to buying electric vehicles or developing circular energy systems.

 

THE SOLUTION

Developing granular energy consumption forecasting system, providing medium to long term visibility and understanding of energy consumption drivers.

Identifying the business needs

The Client asked Jetpack.AI to start with business analysis to help them get a better and deeper understanding of their users’ energy consumption behavior and their underlying components. Together we identified the requirements:

  • track consumption behavior at the lowest granularity possible (consumption points / EAN)
  • analyse consumption in relation to key sociodemographics (income levels, household or business size, etc.)
  • more accurate network flow predictions for long-term forecasting (1 – 10 years)
  • better identification and definition of the variables needed for reliable forecasting and how to use them (e.g. historical data, weather statistics, calculating the number of prosumers, number of clients, etc.)
  • better cost and revenue management by setting the correct distribution tariffs and accurately planning future investments.

The Jetpack.AI solution

The Jetpack.AI team developed a complete forecasting solution: a tailored data product, with:

  • a serverless cloud data infrastructure, seamlessly integrated with client systems
  • a data processing engine to collect, process data and apply forecasting models
  • a service layer to serve the user interfaces and internal production systems
  • a web based user interface developed specifically to share complex insights through a user-friendly tool that is fully integrated into their regular business practices.

This data product allows them to:

  • explore energy consumption by consumption point, district, city or region
  • get many consumption details for each of the selected areas (energy breakdown by type, consumer category or index frequency, equipment data, quantity evolution timeline, etc.)
  • analyse sociodemographic variables impacting consumption variation
  • predict consumption baselines for the next 4 years using time series modelling
  • reconsolidated historical consumption data, starting from each individual consumption
  • create scenario simulations for the next 10 years including assumptions on transitions to electric vehicles, heat pumps, etc.
  • break down models by geographical area, but also by consumer type (residential, non-residential, protected residential).

Technical details

  • front-end web application in JavaScript/React/
  • data engineering with Scala/Spark
  • prediction models (time series and machine learning (ML) gradient boosting)
  • serverless infrastructure on Amazon Web Services (AWS)

THE RESULTS

CAPEX planning optimisation through improved forecasting and long-term planning

The impact

We are proud to have contributed to  our Client’s business by delivering them a tool improving their strategic planning. They are now equipped with an application and methodology providing:

  • better understanding of the factors impacting users’ energy consumption
  • highly accurate consumption forecasts (error margin less than 1%)
  • ability to generate scenarios with consumption hypotheses for 10 years
  • bottom-up visualization, from the consumption point to a global overview.

What’s next

Jetpack.AI is continuing its collaboration with this Client to support their forecasting. Our next steps include the development and integration of a scenario simulator into the application. It allows users to create long term simulations based on adjustable evolution scenarios, for which the users can modify assumptions and parameter values. The application will also further integrate additional network assets and connections, relaxing its focus on consumption points to also cover the central elements of the distribution network more extensively.

Client:
Resa
Category:
Date:

23 April 2021