Multi-agent modeling for the simulation of a simple smart microgrid
Introduction
In the last years, energy systems are moving away from a centralized and hierarchical structure, under strict control of the electricity supply companies, toward a new system where distributed actors influence the energy supply. Production is no longer limited to large energy providers, as small decentralized producers in the form of distributed generation (DG) enter the network and are able to inject energy at much lower voltage levels than before.
This paradigm shift involves new challenges for the modeling and simulation of energy systems for which decentralized models are needed. Among the presented in a review in [1], there are, for example, several commercial tools used in power engineering, such as Eurostag or PSS/E. Further, a number of non-proprietary tools exist, such as several toolboxes based on Matlab/Simulink, for example Matpower or PSAT.
Since some years, the term smart grid has become widespread in the energy sector. The introduction of smart grids involves a change from manual operations toward an intelligent, ICT based and controlled network. These changes will especially affect the distribution grid [2], and in this way, microgrids.
A number of models have been developed to analyze and understand the behavior of microgrids. Some of them focused on decentralized control strategies, usually using Matlab/Simulink and similar classic tools, are of special interest [3], [4], [5]. Particularly relevant is the work on demand side management, using virtual powers plants [6] and on multi-agent platforms [7]. And also worth mentioned as significant the recently extended idea, discussed in some conferences [8] and accepted very well in some countries, of using microgrids as building blocks of the future smart grid [9].
Here, we recall that traditional methods used for analysis of electrical networks are based on static power flow calculations [10]. But these methods are not suitable for computing the system response to special events (such as power changes in generators fed by renewable energies, sudden connection and disconnection of loads and sources, or even when the network structure changes after a disaster occurs), because in these cases, the values of the variables have to be updated in a short time, as they are used for the network control or simulation, and therefore, a new “steady” state computation is required each time one of such special event occurs, which is not efficient at all.
Trying to give a solution to this problem the authors designed a decentralized power flow algorithm for this kind of models (see below). This method provides a more flexible and dynamic way than traditional methods and is able to cope with sudden changes and disasters. The method is described in [11] and was successfully validated against the Matlab PSAT Toolbox [12].
In [11], the complexity for modeling smart grids was identified. A first smart grid model was developed, which represents the system on the physical layer, by integrating a distributed load flow algorithm. The model was tested by running different simulations, letting interact a wind generation unit, a photovoltaic panel, a battery, two loads and a diesel generator. By modeling the individual elements as agents, a modular and flexible approach was used, where the agents can be programmed to have different behaviors such as the charging and discharging times for the battery system, or the integration of variable wind speeds by adding a wind speed simulator module which directly interacts with the turbines. On the logical layer, a first approach was made.
This approach is extended in the current work by adding real-time communications to the simulation, which represent one of the main features of the logical layer. Because of their scarce resources, microgrids need a flexible demand side [13], so introducing communication is essential as it allows performing monitoring, control [14] and demand side management, among others.
As some specific aspects of the author’s work in this area were presented in some conferences [11], [15], [16], [11], [17], in a rather informal manner, here we intend to offer a more understandable and reproducible description of them, by using some elements from the ODD protocol [18].
Section snippets
Overview
An agent-based approach was chosen for modeling a simple microgrid, trying to represent a minimalistic smart grid (or smart-microgrid). The implementation was done in the multi-paradigm modeling environment AnyLogic.
Design concepts
As mentioned before, two kind of interactions occur in the model. At the physical layer, these interactions are intended to perform the electrical behavior of the network, by implementing a power flow algorithm, and therefore, they do a completely deterministic, imposed, process. But at the logical layer, interactions are as result of agent communications and so agent behaviors can became rather complex, because they use specific rules which possibly depend of a number of environmental, most of
Elements of the microgrid
The microgrid model aims to include most of the aspects of future smart grids: distributed generation, renewable energy sources and communication flows are represented. The model consists of the following elements:
- •
1 Point of Common Coupling (PCC) – bus 1
- •
1 Diesel generator – bus 2
- •
2 Load (consumers) – buses 3 and 4
- •
1 Wind turbine – bus 5
- •
1 Photovoltaic panel, linked to
- •
1 Battery storage system – bus 6
The load model includes the necessary elements to allow its management through a smart meter device.
Load modeling
Simulation results
The simulation main window appears in Fig. 7. It shows the model structure, including the logical layer and the physical layer, with the bus index and type, the voltage magnitude and angle, and the active and reactive powers, at each bus. Note that the connections in both layers are the same as in Fig. 2. The figure also shows the icons of the main instances in the model: the environment, the agents and the Excel file where the data to build the example are written.
The PCC includes a control
Further considerations
There are two parallel networks, which are however closely related to each other. The logical layer represents the information transfer, based on ICT [23]. The electrical flows and phenomena can be influenced by decisions obtained with the information of the communication network. Therefore, communication and inter-layer devices are needed.
Because AnyLogic gives facilities to easily implement communications between agents, the user only have to instantiate and use them, by calling the
Conclusion and outlook
The given example model implements a very simple load management algorithm, to show feasibility of the simulation model. The communication mechanism between the PCC and the consumers allows monitoring powers and send signals to the consumers, which leads to a load management.
The combination of several approaches allows the creation of models that might abstract some details from the single unit models, but all in all create a much more realistic representation at the system level. The inclusion
References (23)
- et al.
Grid-connected renewable energy source systems: challenges and proposed management schemes
Energy Convers Manage
(2010) - et al.
A standard protocol for describing individual-based and agent-based models
Ecol Model
(2006) - et al.
Emergent synchronisation properties of a refrigerator demand side management system
Appl Energy
(2013) - Valov B, Heier S. Software for analysis of integration possibility of renewable energy units into electrical networks....
- et al.
The evolution of distribution
IEEE Power Energy Mag
(2009) - et al.
High-quality power generation through distributed control of a power park microgrid
IEEE Trans Ind Electron
(2006) - et al.
Centralized and decentralized control of microgrids
Int J Distr Energy Resour
(2005) - et al.
Distributed intelligent energy management system for a single-phase high-frequency ac microgrid
IEEE Trans Ind Electron
(2007) - Palensky P, Kupzog F, Zaidi AA, Kai Z. Modeling domestic housing loads for demand response. In: 34th Annual conference...
- et al.
Software agents in industry: a customized framework in theory and praxis
IEEE Trans Ind Inform
(2009)
Cited by (60)
Challenges, advances and future trends in AC microgrid protection: With a focus on intelligent learning methods
2023, Renewable and Sustainable Energy ReviewsA framework for the resilience analysis of complex natural gas pipeline networks from a cyber-physical system perspective
2021, Computers and Industrial EngineeringIntelligent local energy management through market mechanisms: Driving the German energy transition from the bottom-up
2020, Energy ReportsCitation Excerpt :The simulation model that was created had to be able to represent an energy system with the following characteristics: To tackle these challenges, the following approach is taken: to address the first two points, an agent-based modelling approach [3,4] is used. Agent-based models have shown to be one of the few approaches being able to capture complex system phenomena in energy systems.