iBKM – Intelligent Management of Balancing Groups
The increasing use of renewable energies to generate electricity, the growing availability of decentralized and controllable consumption devices as well as the electrification of the entire energy system contribute to the increasing complexity to regulate electricity networks. This results in rising balancing energy costs for the energy industry.
Today, the market for balancing-energy costs in Germany already amounts to approximately 1.5 billion euro per year – and is still increasing. It can be assumed that the complexity will continue to increase as a result of the forthcoming energy revolution and new statutory regulations such as "Redispatch 2.0".
In the energy marked, the balancing group stands for a virtual energy quantity account which is used to regulate the electricity and gas markets.
The aim is to balance the energy quantities that are consumed and those that are injected into the grid in order to prevent over- or underproduction within the balancing group.
In Germany, there are multiple balancing groups that are managed by so-called balancing group managers (BGM). They prepare daily forecasts, also called schedules, for the energy quantities to be processed on the following day.
A wide variety of external factors can lead to deviations from the forecast schedule. To resolve the resulting discrepancies and ensure grid stability, missing electricity quantities must be physically provided in the form of control energy. This regulates the flow of energy.
However, since energy is a product that is traded on markets, the corresponding cash flow must also be adjusted. For this purpose, so-called balancing energy is used to charge the responsible parties for balance deviations within the balance group. This legal construct is an accounting tool meant to maintain the necessary balance between the energy quantity accounts. The price for balancing energy is usually higher than the regular electricity price to ensure that all balancing group managers equalize the feed-in and off-take energy.
New technologies, in particular from the research fields of data science and artificial intelligence (AI), are developed to solve these challenges. In particular, neural networks and deep learning methods are already in use due to their flexibility and wide range of application possibilities to optimize processes across all industries. These modern algorithms and software applications can derive novel insights from data and generate predictions from real-time energy related data streams.
Neural networks and deep learning methods are already in use due to their flexibility and wide range of application possibilities to optimize processes across all industries.
Also EXXETA has asked the question of how AI can be optimally applied to the energy sector and has initiated a corresponding research project for the use of AI in balancing group management.
In order to optimize balancing groups as precise as possible, accurate forecasts is the essential information. Currently common forecasts used in balancing group optimization include (a) load forecasts, (b) electricity price forecasts, (c) feed-in forecasts of volatile electricity generation from wind and solar, and (d) forecasts of variable electricity storage options.
Current tools for energy management represent various types of balance curves, their forecasts, and only a minimal set of instructions for further actions. The information is then manually evaluated by humas. However, forecast uncertainties and causes of forecast offsets are insufficiently represented. Human actors also often form prejudices about the reliability of these forecasts, which can negatively influence their actions.
The goal of our holistic innovation approach is to provide users with precise and explainable forecasts as well as reliable uncertainty estimates, based on the underlying stochastic processes.
The goal of our holistic innovation approach for future AI-enhanced balancing group management (iBKM) is therefore to provide the users with precise and explainable forecasts as well as reliable uncertainty estimates, based on the underlying stochastic processes. This increased amount of information should be presented as efficiently and comprehensibly as possible through a user-oriented interface.
In addition to the development of AI-based software, we lay the focus also on the presentation of data and forecasts, as well as a user-oriented integration of the software into the decision-making process in companies.
A frequently underestimated obstacle in the use of AI is the integration of calculated mathematical and statistical multivariate parameters into the companies' course of action. In this context, seamless integration into decision-making processes and usage-optimized presentation of the data is indispensable for stakeholder acceptance and ongoing use of the application.
Therefore, the goal is not only to present the calculated information as intuitively and efficiently as possible but coordinate also the interaction among the human and AI stakeholders. An optimal solution should increase trust in the machine actor for decision making in critical situations and save important time without reducing mathematical precision.
Today's common methods of automated balancing group optimization are primarily based on linear optimization. Here, various energy network parameters, such as averaged electricity consumption curves, are made available to an optimization unit. This allows forecasting future grid balance deviations which can, subsequently, be controlled in advance. A common result of this optimization methodology is an estimation of the expected required control power of a balancing group.
The solution to be explored in the iBKM project pursues ambitious goals, such as a 3-fold improvement in forecast quality, 15-fold improvement in time resolution, and near real-time optimization (>15-fold improvement). Thus, in addition to a strong reduction in forecast error, the iBKM solution generates recommended actions for current measures to obtain the most efficient balancing group optimization. In sum, these advances will enable significant cost savings for grid operators as balancing energy costs can be avoided.
By combining interdisciplinary expertise within the consortium partners KOS Energie GmbH (KOS), OmegaLambdaTec GmbH (OLT) and Octothorpe GmbH, we at EXXETA aim to develop a novel product for the digital future of the Bavarian energy industry by the end of 2022 (see figure).
The software is targeted at all market participants in German wholesale energy trading, who can achieve price savings by exploiting forecasts of the relevant control parameters in the balancing group.
For the first time in this sector, large volumes of data are thus to be collected and processed in near real-time. The use of AI should also enable the calculation of meaningful forecasts and the development of recommendations for action. The goal is to seamlessly integrate the software into existing decision-making processes in balancing group management and to provide the newly available information to human stakeholders in an efficient and comprehensible manner in order to achieve the highest possible degree of optimization and benefit. In turn, the enhanced information flow should achieve increased efficiency in Bavarian balancing group management, which will lead to an optimized allocation of resources.
KOS Energie GmbH (KOS), founded in February 1999, is a cooperative association of medium-sized, 100 percent municipally owned municipal utilities from the southern Bavarian region. As a horizontal cooperation platform, KOS offers municipal utilities decisive advantages through the creation of important synergy effects, which strengthen the economic efficiency and thus the independence of municipal utilities as regional infrastructure providers.
OmegaLambdaTec GmbH is a leading Data Science & AI startup based at the Garching Technology and Startup Center Gate. Since 2015, OLT has taken a pioneering role in the development of customized smart data and physical analytics solutions. The focus is on data-driven forecasting, anomaly detection, digital twin simulations and simulation-based optimization with broad applications in the future fields of Smart City, Smart Energy, Smart Mobility and Industry 4.0.
Octothorpe GmbH is a consulting company focusing on information security, data protection, and regulation management in the energy industry, for example in the context of smart meter rollout. The introduction of a holistic management approach for the various legal requirements is becoming increasingly important. Due to many years of experience in the information sector, e.g., in the implementation of SAP systems, the necessary practical relevance is given.