Case Study: IO.Energy CLEF – Clean Energy Forecaster

Written by Louise Marlein

6 October 2020

As a GigaWatt partner of the IO.Energy ecosystem, we have contributed to the Clean Energy Forecaster use case. We worked together with Fluvius, BAAM-Consulting and Digibri to complete the sandboxing phase.

In this article we share the goal of the use case and the results of the sandboxing phase.

Summary

An energycommunity can only be as effective as the algorithms it is using to predict the consumption and production within the community.

  • During the ideation phase, CLEF had tested communities’ desire to forecast their consumption and production in order to optimize their consumption of clean and locally produced energy
  • During the sandboxing phase, CLEF had developed the forecasting algorithm and tested it in a commercial building
  • Moving forward, CLEF is ready to focus on the consumer interface of their solution

Challenge & Opportunity

Digitization, decentralization and the pressing need for decarbonization results in an energy transition where consumers become prosumers with an active role in the production, storage and optimal usage of energy. This energy transition creates opportunities for citizens and communities to produce clean and affordable energy towards a low carbon economy. However, energy communities are not that easy to set-up and operate. Some key services still need to be developed in order to highlight the value for the consumers to engage within a community and to facilitate their interactions.

With that in mind, Baam-Consulting, Digibri, Fluvius and Zelospark have narrowed down the challenge that they wanted to work on in the framework of IO.Energy Ecosystem to the following question:

How can we provide the necessary energy-relevant information within a community to highlight the potential for some prosumers to leverage their clean energy produced locally?

The Vision

To leverage the local production of energy within a community, it is essential to match the demand and the supply for energy at a local level. More specifically, it is important to anticipate when a prosumer will produce more than he needs and who will need energy at that time in order to resell his extra energy within his community.

The Clean Energy Forecaster (CLEF) use-case contributes to the shift to a carbon-neutral energy system by enhancing the flexibility of (pro-) consumers at the residential areas. To do so, they make it possible to share within the community key information to take consumption decision by providing accurate and reliable production and consumption forecast.

The Product

To enable households to become an active participant in new energy services provision like demand response, offered by service providers like retailers or aggregators, the Clean Energy Forecaster use-case aims at designing and implementing an energy management platform that truly engages its end-users to manage and control their energy use inside their houses and buildings. The platform will enable households to optimize their energy use within their community by:

  • exploring the potential flexibility in the electricity consumption
  • facilitating the energy management of their houses and buildings
  • managing and controlling existing loads, generation and storage resources according to their flexibility, and whenever possible, make use of local energy generation and storage.

In order to achieve that, a variety of functionalities regarding control and management of in-building energy resources have to be defined along with an interoperable strategy between components, systems and actors. The 3 key features of the platform will be:

  1. a software tool and algorithm to forecast energy production and consumption,
  2. an interactive communication and information platform,
  3. recommendations for optimal planning in order to manage self-consumption of PV energy at community level.

For CLEF, an energy community can only be as effective as the algorithms it is using to predict the consumption and production within the community.
The envisaged solution is expected to be highly configurable and adaptable to economical contexts, scenarios (e.g. types of asset, available loads and generation) and types of end-user, allowing its use to be as broad as possible.

As a product, the Clean Energy Forecaster platform is targeting special groups of users (environmentalist and socio-ecologist) to address (cover) their need. The project is focused on the existing technologies in a way that allows the successful implementation of an energy management solution. Given the high TRL expected of the final solution of Clean Energy Forecaster technology should be close to the market implementation.

The Data Approach

CLEF-app connects to the weather data-provider database, downloads the weather forecast data, historical solar production data and predicts PV energy generation.

CLEF’s AI based software detects patterns and models the aggregated smart meter data from pro- & consumers (via DSO and IOE platform).

The solution shares, a day ahead, energy generation, consumption and balance among community members with a simple and user-friendly software.

The scope of the sandboxing phase

During the sandbox, the goal of CLEF was to develop some building blocks of the Energy Management Toolbox that they have the ambition to create.

To test the customer-desirability, the technical feasibility and the financial viability of their idea, CLEF has joint the sandboxing phase of the IO.Energy program. In this 9-months period, the 4 companies aligned to work on the following objectives:

1. Energy consumption modeling and patterning
2. PV production forecasting
  • Using smart meter data of consumers throughout Fluvius and/or IO.Energy platform.
  • Deployment of big data analytics and ML tools for the energy consumption data-modelling.
  • Building predictive algorithms for (1) energy generation with applying solar irradiation and weather data (2) predicting when we have an excess of energy on the local grid for the next 24h in order to encourage local members to use this excess of energy.
3. Information / communication platform and reanalysis
4. Benchmarking for market validation analysis and hypothesis
  • Providing an application which gives prediction on PV production and energy consumption day ahead.
  • Reanalysis data and evaluating performance of (pro-) consumers regarding energy usage as well as the quality of the prediction.
  • Providing individual information and comparative report at a flexible time frame.
  • What key factors are related to the market and actor involvement influencing the effectiveness and success of a smart community? Which type of information would trigger the practical actions towards adapting the behaviour from the consumers’ side? How to keep consumers engage over a long period?
  • How accurate does the forecasted generation/consumption data need to be? What time frame and resolution are required to transfer and process data in order to get a reliable and accurate prediction.
  • What would be the impact of consumer’s behaviour adaption on the grid stress (in terms of peaks and voltage)?

The sandbox results

 
1. Where the hypotheses validated?

Technical assumption #1: All necessary data already exists

The assumption has been validated as the weather forecast and irradiation measures are provided by ECWMF services collected by Copernicus satellites. With a charge service, it is also possible to use irradiation forecast from EnergyWeather or Solargis for example. Then we finally need data consumption and production of users, those can be collected with dedicated devices or directly from Fluvius for example.

Technical assumption #2: An accurate algorithm can be developed

Yes, we did.

Technical assumption #3: The process can be integrated and automated to be able to apply to different communities and scale

Yes, it can.

 
2. What are the lessons learned?
  • Learning about our potential customers’ segmentation and their needs
  • With IOE ecosystem we gained valuable experience in partner search and match and the formation of the consortium
  • Our forecasting models and applied algorithms can predict reasonably accurate all three input datasets, i.e., electricity offtake, consumption and PV production.
  • The electricity offtake is predicted within supervised machine learning with an accuracy level of around 85%. However, for PV production the level of accuracy is a bit lower which is due to the sensitivity of PV production on cloud condition on the sky.
 
3. What are the next steps?

As CLEF we believe that local energy community members will change their behaviour if we provide them with an algorithm which predicts when it is the optimal time to consume energy. With that in mind, as a next step, we want to:

  • Provide users with an interface displaying the forecast. As we assume that most of the users want something simple to use, it will simply be a green or red indicator for next 24h to show on which time slot energy will be in excess or in shortage. A more detailed user interface for the few users who want to see more details.
  • Measure users’ engagement based on the energy forecast
  •  

Other articles