Machine learning for real-time Advanced Multi-Energy Trading

Context

Europe is pushing for a more sustainable and efficient energy grid based on the introduction of more renewable energy in a prosumer market. By 2030, the 27% [1] of energy consumption should derive from renewables. In addition, a 40% reduction of GHG should be reached, compared to the levels obtained at 1990. A possible solution and opportunity is to change the traditional energy system into Local Energy systems (LES) managed by a Local Energy Community (LEC). The challenge is to correctly energetically balance and financially optimize such a complex system with multiple connected, decentral devices that need to be controlled to guarantee the overall quality and safety. A total system optimization is achievable only if various other types of energy vectors (electric, thermal, HVAC, mobility, data, …) are managed together from a macro perspective.

The aim of MAMûET is to turn these specific multi-energy challenges via research into tangible industrial knowledge. In the longer term the goal is to bring this knowledge via a living lab and demonstrator to the market. Accordingly, the context in which MAMuET will be elaborated is key as it is the fundamental part of its valorization. The ‘Green Energy Park’ (GEP) is focused on the development of a living lab and a demonstrator. The consortium endeavors to deeply research the management, control and exploitation of the CO₂-neutral, self-sufficient multi-energy microgrid that will be built within the context of the Green Energy Park (www.greenenergypark.be). As a future perspective (outside underlying project proposal), the Green Energy Park will accommodate an experimental living lab to develop, test, and validate market ready products and services for microgrids in real-life conditions. To accomplish the larger valorization impact of the Green Energy Park the knowledge build-up of MAMuET is fundamental. The valorization potential of GEP is large as it interconnect various prosumers: a large datacenter, an incubator for start-ups, a large parking lot (150-400 vehicles) with electric charging infrastructure and 70 companies from different sectors. In addition, the CO₂-neutral microgrid will integrate renewable energy production systems (3- 4MW solar, 3-4MW Wind Energy), cogeneration (or Combined Heat Power CHP) and energy storage capacity. GEP will be conceived for accommodating a wide range of technologies, offering an experimental and versatile platform for the Flemish industry, and will be used to implement and assess innovative products and services that meet the challenges of the grid-of-the-future.

The Green Energy Campus project in the Research Park Zellik and the different living lab focus points

Scope and objectives of MAMûET

The main essence of the innovation of the MAMûET project is on research enabling the co-design optimization and the smart management, control and exploitation of a multi-energy microgrid. Accordingly, the facilitation and development of a Cooperative Ecosystem is key inside this project. Inside this collaborative scenario, three main goals have been described for MAMûET:

1. Cost effective exploitation of a microgrid by optimal co-design and real time control of the assets.
2. Optimization of multi-energy management for increased Total System Efficiency.
3. Standardize, Scalable and Replicable Proof of Concept of the Cooperative Eco Platform.

To realize these goals, following main knowledge acquisition is essential:

• Self-learning algorithms for State estimation of assets and predictive analytics for the integration of a wide variety of connected devices and advanced renewable energy production and storage.
• Optimal management and control strategies for reliable operation, optimal balancing and stabilization of voltage and frequency, with maximal use of renewables, supporting fast and reliable off-on-grid transition.
• Business and exploitation models based on predictions of multi-energy virtual merit orders, enabling beneficial trading of energy within the microgrid and between the microgrid and the grid.

The self-learning, predictive algorithms to be researched in this project will optimize the operational multi-energy management of the microgrid in such a way that prolongs the asset lifespan and exploit the assets’ generation and storage capacities. The algorithms will be based on big-data analytics and self-learning approaches. A multi-vector multi-agent system is a complex system that combines different types of devices (such as wind turbines, PV-panels, batteries, electric vehicles, …) from different energy-vectors (such as electricity, heating, cooling, ventilation, water, …). The potential of optimizing such a grid is large considering the array of possible source dispatching and storage capacities in response to different demand types. The self-learning algorithms that will be developed will predict the demand/supply and optimize the co-design (CAPEX reduction estimated up to 30%) as well as the operation and maintenance (OPEX reduction estimated up to 15%). The optimization includes a complex set of external factors such as grid prices, levelized cost of local energy, weather, CO₂ taxation and environmental.

The basic is covered by VUB being the academic partner with a proven track record in State and Behavior Predictive Models and design tools for energy systems. The industrial partners; ABB, Power Pulse, Priva and SDM, will ensure innovation technical and financial feasibility. The industrial partners have complementary specializations that ensure project success, motivated by improving their operations and business processes.

[1]EC (2030 climate & energy framework https://ec.europa.eu/clima/policies/strategies/2030_en