A Decentralized Energy Management System Based on Multi-objective Optimization
Bahman Ahmadi is a PhD student in the department Mathematics of Operations Research. (Co)Promotors are prof.dr. J.L. Hurink, dr.ir. G. Hoogsteen and dr.ir. M.E.T. Gerards from the faculty of Electrical Engineering, Mathematics and Computer Science, University of Twente.
The shift to cleaner energy is changing the way we generate and use energy worldwide. This transition is driven by the need to combat climate change and decrease reliance on fossil fuels and it is facilitated by the switch from fossil fuels to renewable energy sources. By that it also necessitates the electrification of various appliances like Electric Vehicles (EVs) and Heat Pumps (HPs). As a consequence, energy systems are being redesigned to remain reliable while integrating more renewable sources, supported by policies like the Paris Agreement and national clean energy programs.
The energy transition now leads to electrification in many sectors, leading to challenges in the electricity grid where the mismatch between the supply from renewables and demand from electrical appliances results in increased network stress, asset degradation, and potential blackouts. This situation highlights the need for more adaptable decentralized models to effectively manage distributed energy resources (DERs) and handle fluctuations in renewable energy generation. Solutions like demand-side management, dynamic pricing, and smart device integration are being introduced to tackle these issues. For this, traditional power grids, which were designed as centralized systems, are being transformed into more flexible and decentralized networks, supporting bidirectional energy flows and real-time adjustments to changes in supply and demand.
A significant step in shifting to clean energy is the emergence of Energy Communities (ECs) and Energy Management Systems (EMSs). EMSs play a key role in managing the integration of resources like solar panels, batteries, and flexible energy loads. They manage energy distribution using real-time data, forecasts, and intelligent algorithms to improve efficiency and system stability. Furthermore, the proposed decentralized EMS enhances local control and flexibility, bridging gaps in current optimization efforts. Within the concept of ECs, homes, businesses, and public institutions connect to a local electricity grid and collaborate to form a local energy system. Hereby, an EMS aims to support the members of an EC with the integration of local energy production, amongst others, by supporting the sharing of renewable energy within the community. The goal is to make the community more resilient and self-sustainable, and to offer social and economic benefits by giving people more control over their energy use. Furthermore, EMS supports ECs to balance energy production and consumption, making the grid more resilient and efficient while supporting a decentralized energy future.
This thesis focuses on developing a decentralized energy management framework using multi-objective optimization. The goal is to efficiently manage DERs and flexible assets in future ECs while ensuring stability and meeting the various needs of the end-users.
The proposed approach models both the physical and cyber layers of the future energy systems and applies advanced optimization methods to balance technical, economical, environmental, and social goals. By emphasizing decentralized decision-making within the proposed EMS frameworks, this research introduces controller(s) that make local, optimized choice(s) to contribute to the overall performance of the EC and the underlying energy system. Hereby, an in-depth analyses of the cyber and physical layers, the implementation of power-flow analysis, and collaboration strategies that help mitigate operational challenges are presented.
An integrated model is proposed that unifies the physical grid model with a cyber communication network to facilitate data exchange and decentralized control. For this, a model to represent physical electricity grids and the interconnection of
electrical devices is developed, whereby in this model, the cyber aspects of the energy grids are integrated. The modeling of the cyber systems involves creating a framework that integrates models of the physical grid, communication strategies between the controllers, and data exchange between the physical grid edges, controllers, and users. This framework ensures that the management system can dynamically respond to changes in the grid conditions and user demands. As it covers both the physical and cyber aspects, the proposed approach allows for a comprehensive understanding and management of the EC grid, enhancing the reliability and efficiency of energy distribution.
The first contribution of this thesis is the Multi-Objective Energy Management System (MOEMS), which employs advanced optimization techniques to balance objectives such as cost efficiency, emission reductions, and grid stability in ECs. The cyber layer of the proposed EMS uses a centralized controller to manage the assets in the EC based on user-defined preferences. MOEMS uses a decentralized decision-making concept where all of the residents in the EC can manage the assets, incorporating their desired goals and constraints. The provided flexibility by users within MOEMS is used by the central controller to eliminate grid-level issues while determining acceptable solutions for the users.
Building upon MOEMS, a Decentralized Multi-Objective Energy Management System (DMOEMS) is introduced, which distributes the decision-making processes among local controllers. This decentralization enhances scalability and resilience, enabling each controller within an EC to autonomously optimize its operations while still contributing to system-wide performance metrics.
DMOEMS integrates a multi-objective optimization framework that aggregates conflicting goals, e.g., minimizing electricity cost and CO2, reducing photovoltaic (PV) curtailment, maximizing self-consumption, minimizing power congestion, and eliminating voltage violations by converting these objectives into a single objective using user-defined weight factors. Each local controller optimizes the operation of its attached physical grid assets (e.g., PV, Energy storage system (ESS), EV, and HP) based on local (grid and assets) constraints and user preferences. In contrast, the EC controller coordinates the aggregated power profiles through an iterative feedback mechanism. This coordination dynamically adjusts weight factors and curtailment strategies to resolve grid congestion without compromising individual privacy.
The centralized MOEMS approach is tested through simulations using a model of a real EC (the Aardehuizen community) and a real-world micro-grid set up in the KEZO research lab, thereby demonstrating its adaptability to various configurations.
To assess its effectiveness, the performance of MOEMS is compared with traditional EMS approaches, such as profile steering and optimization methods that focus on economic and environmental aspects. The results highlight the ability of MOEMS to address grid challenges such as power congestion and voltage violations, while meeting diverse stakeholder objectives. Notably, in the simulations and real-world implementation of MOEMS, the result shows up to 22% higher annual electricity cost savings and a 37% reduction in CO2 emissions compared to traditional EMSs.
In the decentralized setting of DMOEMS, corresponding user-centric and flexible coordination mechanisms are developed to enhance local engagement and acceptance, as this promotes active participation of users in energy management. Simulation studies on the Aardehuizen EC demonstrate that DMOEMS effectively mitigates overloading scenarios across diverse operating conditions (high EV charging, regular demand, and excess PV generation), resulting in user satisfaction, reduced operational costs, and lower CO2 emissions. The proposed framework highlights the potential of a democratic and decentralized approach to energy management in ECs. The numerical results for the management of physical grid assets using DMOEMS show improvements of 20% in CO2 emissions, 4% in electricity cost savings, and a 30% reduction in PV curtailment relative to baseline scenarios. Furthermore, the proposed mechanism in DMOEMS (compared to the MOEMS) shows an improvement in computational cost by converging faster to resolve grid congestion compared to conventional approaches.
Finally, this thesis explores the integration of communal assets to mitigate grid congestion and stabilize local voltage profiles. The simulation results for the proposed frameworks demonstrate practical applications in real-world settings like the Aardehuizen community and the KEZO research facility. Hereby, this thesis confirms the adaptability of the presented EMS frameworks in integrating evolving grid services and technologies, ensuring future readiness in various use cases.
