Because MGs come in many different shapes and sizes, their controllers must be strong, flexible, and capable of essential computations at high speeds. Using a hierarchical control scheme with multiple time levels is especially interesting when you think about how fast things are for managing outcomes and how slow things are for economic scheduling. An MG controller design with data transmission, a global controller, and localized control systems can be centralized or decentralized. An MG with centralized control uses a transmission network and a central hub. An MG with decentralized control uses local data and cooperation between micro-sources, capacity controllers, and MGCCs to control tasks. While a centralized server maintains global control performance by being aware of every network location, the vulnerability of massive-scale MGs to single-point breakdowns and under attacks renders the adoption of centralized processors impractical93,94. On the same point, an entirely distributed strategy cannot be implemented since special hardware is too expensive. As a result, a modern mix of the two strategies is adopted. This strategy is known as multilevel MG administration, consisting of three operational layers, as illustrated in Fig. 4: (1) primary (field type), (2) intermediate (MG type), and (3) auxiliary (grid type).
Hierarchical MG control systems have been chosen as the primary control method. Still, the changing behavior of alternative energy DPRs and demands makes it hard to set limits on hierarchical control. So, decentralized and distributed strategies are being used to make it easier to use MG control systems95,96.
Recent research suggests how the traditional three levels of the hierarchical management scheme are combined into smaller systems to realize plug-and-play functionality97. Optimal control designs, also known as laminar monitoring designs, were stated in98 for the development of distributed equipment and networks by incorporating limitations on data flow among layers (formation of control schemes and segmenting targets into sub-tasks). Such integrated control systems reduce control messages and adequately express control algorithms between systems while still optimizing. However, integrating levels to improve MGs’ speed, effectiveness, and acceptability requires further investigation. Even so, hierarchical MGs control is still the best99, as flexible control structures and precise control mechanisms are still being made. Hierarchical control attains a good balance between fully centralized and distributed control mechanisms and manages many customizable parts while meeting strict quality goals. It has a dependable and robust networking system and compatible routing protocols. Thus, the subsequent sections elaborate on these three control layers (primary, intermediate, and auxiliary).
Primary-level control systems consist of localized controllers. These controllers govern energy distribution across micro-sources, regulate the current and voltage characteristics of digital power inverters, and maintain the frequency of the entire system. It also serves as the initial line of defense against electrical voltage and frequency fluctuations in an MG, and it works on the slowest time scales95. By using neighborhood data, variations can be reduced to a minimum; therefore, a communication network may or may not be needed (see Table 2).
Current source converters (CSCs) and voltage source converters (VSCs) are two kinds of power semiconductor inverters that may be used to transform the alternating or direct current from clean energy sources47. A grid-forming VSI regulates potential differences and system frequency, making it well-suited for MGs operating in an isolated configuration. In contrast, a grid-following VSI regulates both reactive and active energy.
However, control of inverters is required for controlling production characteristics of both voltage and current, in addition to the proper proportions of outer feedback loops, i.e., droop or non-droop-based regulation and maximum power point tracking100. The subsequent section analyses in further depth the inner control dynamic characteristics of such inverter classes and the outside energy-sharing control approaches.
A DPR integrated with a grid-following voltage source converter coupled to the power grid depicted in Fig. 5 demonstrates the corresponding 3-phase dynamical model represented by (9)109:
For dq benchmark, the corresponding model for currents and potential difference is represented as given below:
Here (10) and (11) are the reference values for the d-axis and q-axis components of the output voltage respectively. The model (10) and (11) are linked with and final potential is shown in model (12) and (13) as provided below:
Here (12) and (13) are dq-frame voltages. The coefficients of the filter resistance and inductance are denoted by and , respectively, while co indicates the system frequency.
For integrating energy from green sources (such as photovoltaic and wind power), the most common type of inverter used is a VSC that follows the grid or feeds electricity back into it. Its primary function is interacting with an alternating current system and trading reactive and active electricity. A VSC is often built to supply current. This paper also takes into account this design pattern. A power station or grid-forming inverter is necessary for island mode functioning of an MG because the MG cannot produce the required voltage as well as frequency without them110.
Reference currents and are often supplied by an energy regulator in grid-following inverters that regulate the energy sent to the system and consumed by the localized demand. The following expressions represent the power elements.
In Eq. (14), P denotes active power, where Q represents the reactive power. As shown in Fig. 5, may also be employed to change the potential111.
Figure 6 displays the control design of a grid-forming converter109, which is comparable to that of a grid-following converter except for the two circular cascaded cycles. An outside cycle monitors the benchmark voltages to compute current for an inside current control cycle. The dynamic model for grid-forming is given as below:
In model (15), and represents the demand current and voltage respectively.
The bulk of energy exchange strategies for grid-forming converters depend on droop management as an external feedback controller to maintain the standard MG frequency as well as voltage during PCC112 (see also Table 2). To accomplish that, we reduce the stable operation of synchronous generators by using linear exchange equations involving voltage (V), frequency (f), reactive power (Q), and real power (P). The visual correlation across the droop coefficients is shown in Fig. 7.
Energy distribution varies owing to demand and grid redesign113, however, it continues to be a complicated process to calculate the magnitude of this impedance. Thus, developing more complex droop mechanisms for precise reactive energy pooling is a current field of study. While these controls’ concepts, parameters, and used cases vary, a few examples are shown below.
An increase in information exchange control techniques for inverter control118,119 can be attributed to address the shortcomings of droop regulation approaches, like sluggish dynamic response and laborious tweaking of control settings. Droop-less network-based control approaches for dynamic demands and unpredictable situations enhance frequency and voltage control. They depend on a data network to regulate the power produced by sustainable DPRs and may be deployed in a centralized or distributed design. One option is a bounded control setting approach that has reduced objective function computations and computation latency. Other applications involve advanced control approaches such as inductive reasoning, neural network computing, and evolutionary computing120,121.
On the primary layer of control, much research has been carried out, although they are still in the early stages. The primary purpose of this work is to update existing methods even though droop reduction solutions for MGs with many random components have already been the topic of substantial research. As a result, cutting-edge techniques such as the meta-heuristic algorithm described in122 are utilized to regulate the frequency and the voltage in an offshore MG in relation to the limits imposed by its operational capabilities.
In the MG system, any power and frequency irregularities caused by primary control are brought back under secondary or intermediate control for correction. It functions on a shorter time scale to send task signals to the primary layer, which the primary element uses to manage the economy and synchronize MG networks with the utilities. Secondary administration can be implemented in either a hierarchical or distributed design. This helps to decrease distortions and energy imbalances by controlling DPRs in accordance with the primary regulators of said relevant devices (refer Table 3). In a configuration where the MG is controlled centrally, the MG administrator plays a pivotal role in synchronizing DPRs and achieving optimal MG performance73. A few examples are shown below:
Even though they work effectively in various functioning situations, these approaches are vulnerable to breakdown at a specific point since they rely heavily on centralized transmission. On the opposing side, the decentralized design divides the task of regulation optimizing between the allocation network provider, the MGs control system, and the local processors. Distributed secondary supervision approaches employ independent and specialized regional controllers who use neighbour data to lower the price of operations and the characteristics of the system that backs them up. For example, in128, a secondary control scheme that relies on two types of processors for every DG unit is presented. The mechanism can accurately share reactive energy, and return frequency as well as voltage to their reference values. Secondary control systems are positioned across the network connectivity, and the primary control system generates a command signal for the first layer. Further examples of this can be seen in Table 3.
As the uppermost level of governance in grid-connected MGs, tertiary or auxiliary control governs how the MGs communicate with each other and the upstream system. Table 4 lists the research on tertiary control techniques. For efficient scheduling, optimal performance planning, and controlling power flow in both directions between the power network and MG30,137, tertiary management functions are performed at the minimum time scale. Centralized tertiary regulation systems were more generally utilized to control MGs; however, in138, a highly decentralized tertiary controller has been established for every DG unit. Decentralized primary control systems (DPCS), decentralized secondary control systems (DSCS), and decentralized tertiary control systems (DTCS) are the three components that make up the unified supervision approach, which offers enhanced adaptability and dependability. A conventional droop mechanism is the foundation of DPCS, whereas a non-convex droop adjustment serves as the basis for DTCS. The control scheme is modelled by model (16), which has the following form:
The gain parameters for DTCS and DSCS are denoted by and , while the yield frequency and observed frequency are indicated by and . Several decentralized tertiary control systems are now working to handle a variety of MG difficulties, including black start functioning, reserve power planning, and general harmonic aberration adjustment46,143. Tertiary control also helps balance demand and makes switching between grid-connected and standalone operations easy. For example, the authors of123 describe a master-slave peer-to-peer fusion control method that allows smooth changes among the two operational modes of MGs. The MG controls frequencies and amplitudes under grid-connected functioning while keeping necessary energy production constant. MGs are more resilient because they use a peer-control strategy that takes precedence when the information network fails. Furthermore, in a three-phase delivery network, voltage fluctuations can be introduced by single-phase demands or asymmetrical electrical transmission. This could cause damage to voltage-sensitive devices. Tertiary supervision is a cost-effective method for balancing the restoration operations of numerous DGs. Further compensation devices, such as series or shunt-activated power diffusers in144,145, were introduced to minimize imbalance. The technique considers various lines and DGs’ compensating constraints and voltage stability requirements.
According to existing scholarly publications, this part presents several contemporary control strategies. In contrast to traditional linear control systems, these approaches offer higher dynamic characteristics under all working situations despite disadvantages, including the stuttering problem and extensive mathematical formulation (for details, please see Table 5).
To achieve the MG monitoring objectives, a multi-agent system (MAS) involving a group of autonomous agents can be employed147,148. Agents are distinguished in terms of MG security, control efficiency, and economic functioning by their independence, societal interactivity, responsiveness, and identity149,150. A decentralized MAS method may more easily share data and coordinate operations by dividing a vast power grid into smaller, more manageable components. A single controller governs similar units in a cooperative MAS system to draw inferences about their control systems. To sum up, MAS employs hierarchy to improve MG management in a dynamic context by classifying agents based on their respective power status. Entities in a MAS’s system primary, intermediate, and auxiliary levels undertake diverse control roles and rely on distinct information pathways to carry out their activities autonomously151. The accessibility of elements, the control functions performed by those agents, and the type of data provided between those agents affect the aggregate efficiency of the managed system. These three parts are controlled and built the same way as most MAS structures.
MGs increasingly use MPC in their planning to overcome nonlinear economic optimization barriers152. Both the grid-level and converter-level variants of the MPC algorithm include a prediction framework, a solution mechanism, and an objective function. These three essential components are included in the grid-level version as well. Unlike its predecessor, this technique maximizes the MGs’ efficiency while considering several competing objectives and constraints153. For example, the work of154 presents an MPC-based regulator for home power systems with a distributed generation and storing system. The controllers perform storing device optimization, power consumption forecasting, and power trading.
Also, in155, researchers looked at how MPC can be applied to control backup devices in MGs. A solar energy system and rechargeable batteries in an ESS MG are managed using a combined MPC156. In157, an MPC-based technique is used to control a faster magnetic energy storage device so that changes in the unit’s electrical current and power don’t cause chromatic oscillations. When applied to voltage control with a DC/DC compressor in an SMES network in MGs, the proposed control approach can reduce eddy current losses for the DC power of a superconducting circuit, resulting in favorable effects.
When problems with reliability and unpredictability are worst, more than logical analysis and computer simulations of MG networks are needed for reliable control systems. Intelligent controllers powered by artificial intelligence can adjust to complexities and do not necessitate any existing understanding of the system’s functioning. Thus, their potential use in MGs, including electricity supply and demand restoration, capacity proportioning, security, energy delivery optimization, and demand management, has garnered a lot of attention158.
The photovoltaic forecast program in159, an artificially intelligent neural network, allows MGs to be regulated in real-time. MG controllers use the predictions application’s outputs as one factor in making predictions about photovoltaic units’ electricity production. As a result, MG recovery mechanisms like fuel inverters only activate when required. In addition, the following are examples of other machine learning-based controllers used for MG electricity supply and consumption regulation:
Integrated MGs have yet to be completely investigated despite the increasing significance of ANN-enabled solutions in the MG area. Hence, approaches for autonomous market pricing, energy pooling, and optimal energy supply control in networks of MGs must be investigated in upcoming studies.
Traditional controllers are no longer useful because of changes to the network, such as the redesign and integration of more dispersed generation and demand. MGs are particularly vulnerable to the effects of these occurrences since they happen so regularly. Improved agent behavior while optimizing the objective functions is the goal of sophisticated control mechanisms, including RL-oriented management approaches. Arithmetically, RL consists of a Markov chain of events with three independent variables: the actor, the incentive indication, and the surrounding context. An agent’s actions in the world have consequences, both deterministically and stochastically163. Some examples of RL-oriented approaches that have been used in the past are as follows:
The MAS idea was considered in MGs, and the consensus model has indeed been employed as the fundamental concept for cooperative control of bots167. Every agent uses the data available locally and the consensus algorithm to cooperatively collaborate with the neighbors in its neighborhood to arrive at a consensus168,169.
Second-level multilevel MG management has implemented a consensus technique to restore frequency and voltage characteristics in MGs. This is especially true when synchronizing grid building and grid tracking inverters, for which consensus-based collaborative control mechanisms have been developed. The authors of138 provide a joint control analysis of grid-forming and grid-following converters, which guarantees MG plug-and-play functionality and maximizes energy produced from sustainable sources. To bring the voltages of multiple DGs into agreement, we use a paradigm federalization technique to combine control conditions and design a global second-order feedback control technique.
Power administration in MG networks is another area where consensus supervision has been used. For instance, in170, the consensus concept has been used to address the energy dispatch issue for an MG with five lines using a quadratic objective model. To strike a good compromise between production and consumption, we determine the incremental cost that each device should update. Using a consensus mechanism to maximize the cost-effectiveness of MG electricity networks has much promise. How to use this concept to provide synchronization among many variable components, such as storage devices and generation machines, in an MG network is an intriguing area for further study.
