The possession or control of energy generation and supply is owned and operated either by private companies or the state and has hardly ever been characterized as a sustainable pattern for the electricity sector. Energy and financial crises, and catastrophic events affect the monopoly gravely and extensively, while lack of proper regulation has severely affected the price, reliability, and quality of electric power supplied to end customers, either residential, commercial, industrial, or in the public sector. System operators, regulators, and governing and policy bodies that introduce and oversee standardized codes, practices, and protocols regarding the operation and control of power systems have only been active since the 1980s. Additionally, since the first 1990s, several concerns have been raised about the harmful, global effects of fuel emissions on the environment and therefore the climate. A large portion of these emissions come from the conventional electricity generation sector, meaning that a shift to cleaner and more efficient sources and technologies are required.
The initiatives focused on the promotion of generation from renewable energy sources (RES). Several subsidizing mechanisms attracted participants as small as residential customers in the energy market. These mechanisms, although critical to the vision for the shift far away from fossil fuels, have also introduced considerable irregularities within the electricity sector. The aforementioned two factors have been the main drivers of what has been described as the deregulation of the energy markets. This article also follows the same mindset by organizing what is understood as modern control technologies from the perspective of the characteristics of each of these technologies. In other words, the aim of this article is to enable the reader to identify not application-specific power system challenges and how some control methods address them, but rather what should one expect and need as minimum or average features of any novel approach employed in power grid control methods.
2. Classic Control Applications in Power Systems
To have a far better understanding of what the challenges are of employing control in modern power systems, one must assess the present status of the classic control requirements and functions in today’s power systems and therefore the very recent past. During this sense, a quick overview of the foremost common control problems is provided within the next subsections.
2.1 System Area Frequency Control and Voltage Regulation
Regulation of frequency and voltage at any given point of an operating area and at any given moment are the 2 most ordinarily understood operations in power systems stability and control. The electrical frequency of a system expresses the angular speed of the rotating electric currents altogether electric machines and must remain on the brink of par value (50 or 60 Hz) for multiple reasons of infrastructure integrity and system protection. Voltage magnitude also must be kept as on the brink of its rated values as possible to ensure efficient behavior of end-users equipment and avoid collapse of any part of the system, which can cause interruptions and blackouts. Both of those control functionalities got to be managed efficiently consistent with standardized grid codes while taking under consideration their specific features. Frequency is common throughout an interconnected system and is an electromechanical phenomenon within the time-frame of tens of seconds, while voltage regulation can only be handled by local resources and measures, and, as a physical phenomenon, may require response times as fast as milliseconds.
Thanks to the aforementioned features, power system operators have, traditionally, controlled frequency and voltage using hierarchical schemes. In other words, the hierarchy of the control action may concern either the time-scale of operation (e.g., within milliseconds or up to minutes) or the world (e.g., starting closer to load centers instead of generating plants which will be involved in some following action) or another perceived perspective/structure of the matter. Commonly, hierarchical control architectures comprise three main levels: primary, secondary, and tertiary, The first control usually realizes control actions as a kind of automatic response to a change of a standing variable.
The aim of primary control is to limit excursions from nominal control points. Droop control is one among the foremost popular sorts of primary control that regulates the frequency by adjusting the output active power (commonly) of a controlled asset (usually a generator). Voltage control, although not explicitly, has traditionally been organized in primary and secondary levels, the first being fast control action offered by automatic voltage regulators (AVRs) of generators or any dispersed assets, like loads on demand response programs, properly equipped with controller topologies that monitor and drive accordingly the corresponding resources. The secondary control follows the first, resets the supply of primary reserves utilized in the previous stage, and aims to revive frequency and voltage to nominal values if that has not been achieved earlier. As primary control is, commonly, an area proportional sort of control, it is going to not be ready to achieve the reference values accurately and a few offsets should exist, thus requiring.
Pathways to a better power grid an integral sort of control. Since it follows primary control, secondary control is implemented for extended time horizons and maybe a regional sort of control handling a zone of an electrical system. In this sense, the control also becomes a market problem with optimal formulation characteristics. Hence, at this stage one may identify the secondary voltage control and load-frequency control (LFC) — otherwise referred to as automatic generation control (AGC). AGC has relevance to a different crucial matter of power grid control, the world control error, and therefore the interarea exchanges of active power and their effects on the steadiness of neighboring systems.
The third level of control, the tertiary control, is driven mostly by cost criteria of sorts, that is, optimal economic dispatching, minimization of losses, operation scheduling, and other optimization problems of comparable nature. It is a system-level, multiple time-steps ahead control technique, which takes under consideration the entire system (consisting of the many, loosely connected regions) over an extended time horizon (to the length of hours, days, etc.).
It is, essentially, the operators’ framework consistent with which all reserves for primary and secondary control are allocated and supported minimum cost considerations. Popular samples of tertiary control are economic dispatch (ED) and unit commitment. The distinguishing characteristics of primary, secondary, and tertiary control schemes are delineated in Table 10.1. In modern power systems with high penetrations of DG, it is critical to involve in frequency and voltage control as many of such units as possible, thanks to the displacement of conventional sources they cause, which severely affects the grid stability.
Despite the strong motivation, there are numerous challenges. for instance, when employing DG in frequency regulation, thanks to their unpredictability and operational constraints like those imposed by power extraction strategies, the performance of the respective control has got to be monitored continuously, while some sources will not be ready to contribute with reserves in the least, thanks to the said strategies, which do not add up otherwise. For the case of voltage regulation by DG units equipped with an inverter (a particularly common topology for many renewables), their output is going to be actively controlled in response to some voltage sensing, depending also, though, on the network characteristics; obviously, the inverters got to be ready to adapt and ensure proper control performance altogether cases.
2.2 Dispatching of System Resources
The steady-state dispatch problem of the power grid is typically formulated as an optimization problem to seek out the control set-point of all assets within the system with reference to an objective function (usually the system operational cost) subject to a group of operating constraints (power balance, voltage limit, line flow limit, etc.) and given parameters (load, equipment status, etc.). The control models of the power grid take different mathematical forms. The ED model neglects transmission constraints and only implements steady-state power balance constraints. On top of ED, the direct-current (DC) optimal power flow (OPF) model adds transmission constraints, but still ignores the reactive power and assumes a flat voltage magnitude of all buses to stay a linear model. Alternating-current (AC) OPF further calculates reactive power and its impact on bus voltage and considers voltage constraints. As AC OPF is nonlinear and nonconvex, its computational complexity is significantly above ED and DCOPF. there is a trade-off between convergence and accuracy. Since its first formulation, ACOPF and its approximate models have attracted intensive research. Within the conventional power grid operation, the independent system operator (ISO) solves the ED problem of the available generation units on an hourly basis by scheduling the on/off cycles of generation units by solving the day-ahead unit commitment (UC).
There is upscale literature on different solutions to both ED and UC problems, with certain objectives and constraints. The increasing complexity and enormous scale of power grids motivate the necessity for novel algorithms to enable efficient computations of power system operation and control. This is, however, a challenging task due to various sources of uncertainty, the physical constraints of the network, and therefore the novel technologies introduced by the smart grid (e.g., RES and demand response programs).
3. Characteristics to Best Serve Modern Power System Control Applications
The matter of addressing the control of large numbers of actors and for multiple purposes is evident and urgent following the deregulation of the energy markets and the wide promotion of the highest penetrations of DG units. Units of sizes as small as a few kilowatts are dispersed throughout the power system, and especially at the distribution level where no direct control of the assets has traditionally been employed or practically explored. Recently, decentralized/distributed and hierarchical control and optimization techniques have been widely deployed in the power system control to answer the challenges by the scalability requirements. They introduce potential advantages, as compared with centralized methods, including but not limited to the following :
● Each controlled asset needs to broadcast limited information with some neighbors.
● Limited information exchange requirement ensures more privacy among the assets, as well as reduced costs related to communication infrastructure.
● The algorithms are more robust in terms of communication link failure and certainly against single-point failures.
● Due to the distributed nature of computation, the run-time reduces considerably.
3.2 Adaptive Control and Topologies
Traditionally, a controller is adaptive if it is ready to update its parameters (for the sake of performance or stability) in response to changes occurring to the controlled asset/equipment. this is able to translate to regulate methods or systems, which either estimate the state of the asset parameters online and adapt accordingly or adapt once they identify changes to the asset consistent with a particular model representation of the said asset. Fig. 2 offers a schematic abstraction of adaptive control. Although adaptive control methods are not new within the field, the facility system and therefore the electricity markets became more susceptible to showing changing behaviors, thanks to the involvement of multiple market players, the operation of various DG units throughout the system, and therefore the unprecedented interactions among all of them . “Changing behavior,” within the context of adaptive methods, shouldn’t be confused with “variability,” since the latter implies that there are inherent characteristics that would be described through some probabilistic models.
With the growing concerns about cyber-physical security issues, adaptive control is going to be put center stage because of the only viable countermeasure to attacks of that sort. Resiliency pertains to showing adequate control performance under failures of the control architecture or changes to the controlled asset that might otherwise affect the performance. Further to the present, if the control performance remains equivalent the adaptive control and its topology can also be described as reliable.
Flexibility is a control property that plays a crucial role for the transition of traditional power systems, many of them based on fossil fuels, toward power systems that can efficiently accommodate high shares of variable distributed energy generations. Flexibility pertains to control implementations that can handle a problem in several time scales, regions, or other structured organization blocks that have a minimum effect among them and may thus be handled separately. Flexibility is best expressed in power systems by employing hierarchical schemes of control actions. As hierarchical control schemes have multiple levels of timescales and control areas, they are proven very effective, because they offer a high degree of practical flexibility for power systems. This is especially so for the very timely case of widely deployed DG units based on RES that differ from conventional power plants — prominently due to their high intermittency, efficient management of their operation is essential.
Control hierarchy is not a new concept in power systems. However, the diversity of the characteristics of DG and RES implies that flexible paradigms and controls must be conceptualized for the case of better handling of such characteristics. A very common and obvious such concept is putting together hybrids of intermittent resources and storage systems , or more abstractly and at a greater extent combining stochastic and “deterministic” generators in microgrids or virtual power plants . Based on such set-ups, it introduces hierarchical control to integrate RES and energy-storage devices in the existing power system, while reviewing advanced control techniques for microgrids, including hierarchical control scheme as one of the proposed paradigms for integration of RES with the main grid.
An analogy of how flexibility should be understood for the case of modern power systems is described in Fig.3. Any events and phenomena occurring at the smallest scale (one unit or a set of units behind a point of common coupling/control) can be handled through hybrids of energy sources that ensure a consistent behavior with regard to power quality standards. At the next level,
microgrids are paradigms that can offer services to the grid and control the encapsulated resources in a fully vertical manner. At the highest level, virtual power plants can execute control actions and respond to operator requests by accounting for the diversity of the resources they comprise, while not relying on any given status of the power system, but rather ensuring that the control is efficient under a variety of circumstances.
3.4 Advanced Hardware Platforms and Relevant Enablers
With the increasing numbers and diversity of power system assets, the classic centralized control centers will be required either to handle unprecedented amounts of data and control actions or to follow the characteristics of the assets and adapt accordingly. The Internet of Things has been lately introduced in power systems with the deployment of smart meters , which have been gradually yet widely replacing the classic end-users’ energy meters throughout the grid. The IoT poses firstly the challenge of handling huge amounts of data that are required (online or in later times) for planning and controlling actions on behalf of system operators. Pathways to a Smarter Power System exchange of data with other hardware units hosting agents throughout the grid or any lesser part of it .
This latter requirement describes the challenge of standardizing the respective hardware platforms and their interactions with the software entities that they will be hosting. These platforms will be equipped with sensors gathering systems readings, controllers driving power system assets and resources, and the integrated software for the operation of this equipment. The interface of the integrated software with the agent may or may not follow standardized protocols , thus accounting for added complexity to the control problems at hand. . This implies that the interactions among all of these options will appear as an eminent challenge in the very near future. Lastly, the direct control of power system assets, especially those that mainly express the path toward the deregulation of the energy markets, that is, power electronics and automation of DG units, has been required to adapt to increased power system requirements .
Handling Uncertainty In power systems, “the only certainty is uncertainty,” and addressing this is necessary and crucial to the stability and economics of power system control. The power system uncertainties root from both supply and demand. On the supply side, the penetration of intermittent RES brings significant uncertainties to the system, since the wind speed and the solar irradiation are hard to estimate accurately even in the near future, for example, half a day from now. On the demand side, it is also challenging to make accurate predictions on the random energy consumption of each household.
The power system uncertainties are usually modeled with various stochastic methods, according to the control scenarios and the characteristics of the uncertainty. Robust optimization, which models the uncertainties with confidence intervals as an interval-based method, has been popular in recent decades . This methodology has the merits of lower requirements for forecasting data and computational burden compared to other stochastic approaches, while still yielding practical results for the power system The other challenge is how to determine the optimal control policy given appropriate uncertainty modeling.
Usually, power system control aims to optimize the objective function. This manner enables the utilization of the most updated forecasting information, and its performance can be guaranteed under certain conditions. Uncertainty has been and will continue to be one of the major concerns in power system control, as we move forward to a low-carbon energy sector. Without meeting this requirement, the stability, economics, and resilience will all be compromised.
4. The Path Forward in Power System Control
Many cases of novel power system control approaches are discussed in the following subsections. Research by academia and the industry in the field has been multilevel and has spanned many applications. Nevertheless, all of the proposed ideas bear the five main characteristics analyzed in the previous section, in order to address the control challenges in the new framework, and also what it is expected of the power systems in the coming decades.
4.1 Control of Distributed Resources
4.1.1 Decentralized and Distributed Optimization Methods
The different characteristics and features of employing distributed and decentralized methods on the voltage regulation problem are described as follows. Let it be noted here that for this problem, a decentralized approach is, practically, equivalent to employing local controllers with very infrequent (if none at all) updates of their parameters with regard to some central entity recalculating some relevant objectives. There are two major categories of distributed algorithms for voltage regulation in power distribution systems. The first category, relevant to the majority of the aforementioned works, relies on using linearized power flow equations around the operation point . The other category is based mainly on convexification of the problem using second-order cone programming or semidefinite programming relaxation . Distributed algorithms provide feasible solutions to the reactive power compensation problem , as well as simultaneous voltage regulation and loss minimization . In , an average consensus on reactive power injection ratios is used for optimizing a reactive power sharing objective. There are also several methods based on decomposing the original voltage control problem into subproblems, including online distributed optimization , SDP, ADMM, and leader-follower methods. First a linearized network model is introduced to enable the design of distributed voltage control strategy. Then an analytical scheme is proposed to verify the convergence of the distributed method to local minimum in the nonconvex branch-flow model. The error during iterations is also characterized. Minimizing deviations of voltage from rated values and the reactive power injections (as a form of inverter operation cost) comprise a weighted objective function for the problem. In order to enable the distributed computation, two communication schemes are considered using a tree communication network and complete communication network, with the latter case showing improved performance and efficiency in achieving the voltage regulations goals.
4.1.2 Novel Model Predictive Control
Understanding MPC has been intensively studied for power system applications in recent years. It became popular because of its effectiveness in addressing uncertainty issues that are temporally coupled. The researchers proposed an MPC-based model that controls the aggregation of heterogeneous thermostatically controlled loads (TCL) to provide frequency regulation service. The model aims to decide which TCL in control to switch on/off to track the frequency regulation signal while keeping the temperatures within a certain range. Similarly, proposes an MPC model that controls residential air conditioners to track solar power fluctuations. The TCL dynamics and regulation signal/solar power are both temporally coupled in this application. Energy storage is a key solution that provides flexibility to address the intermittency of RES production and fast ramping capability to enhance power system stability. The control of energy storage is subject to the physical power constraint and energy constraints, while the latter is temporally coupled; the accumulated energy has to be kept within a range. Therefore, many studies have applied MPC in storage control . The degradation issue of electrochemical energy storage (EES) can also be addressed with MPC incorporated in an intertemporal framework .Pathways to a Smarter Power System Intrinsically, electric vehicle (EV) charging is EES control. In this case, the arrival time of future EV is usually random and needs prediction, and the consumer usually requires its EV to be fully charged within a certain amount of time, which brings temporally coupled constraints. MPC was applied to determine the near optimal real-time EV charging profiles in a charging station. analyzed the performance of MPC in EV charging scheduling and concluded that MPC performs very closely to the finite-horizon continuous DP and could reduce computational complexity.
From the whole system perspective, MPC can be applied to control each agent in the system including demand response, energy storage, FACTS, fast-responding generation, etc., and consequently determine the power flow. proposed an MPC-based DC-power-flow model that alleviates line temperature overloads to prevent cascading failures. further expanded the MPC-based dispatch model by accounting for voltage magnitudes and reactive power. MPC can be incorporated in a distributed/decentralized manner, which can enhance the robustness of the system against communication failure. In the distributed control scheme, there is typically no centralized coordinator, and each agent in the system communicates with its neighbors to determine the optimal actions. proposed a Consensus + Innovations scheme for microgrid energy management, using MPC to develop a multi-time-step control model. proposed a noniterative distributed MPC model for a class of spatially interconnected systems with communication constraints, in which a cost function switching strategy is developed to guarantee the stability of the overall system.
4.2 Frequency Control and Active Power Reserves
4.2.1 Extending Frequency Control Implementation and Application
A state-of-the-art review of LFC schemes was offered in . The work proposed a robust LFC using genetic algorithms and linear matrix inequalities. Similarly, a fuzzy logic controller was proposed to control the load frequency in two areas of power systems. Frequency of power systems can be controlled using varying controllable loads, energy storage devices, and generation units. In this sense, presented a frequency control scheme by controlling Hybrid Electric Vehicles, controllable loads, and a generation unit whereas real-time pricing control technique was utilized to regulate frequency of powers in . At a higher level of realizing this type of regulation and due to the recent expansion and growing complexity of power systems, the requirement to involve DG resources in tasks as the LFC requires the accurate and timely transmission of enormous amounts of data. To deal with this challenge, a singular value decomposition (SVD) based real-time LFC can be realized . According to the proposed strategy, the measured data in each control area is decomposed using SVD before being transmitted through the communication network, and only the most valuable information is transmitted to the control center.
4.2.2 Active Power Reserves
Beyond the Load-Frequency Control Problem Catering for closer control of active power became crucial as RES-based DG grew rapidly and vastly, due to its high efficiency compared to conventional generation and its limited environmental impact. RES, however, depends heavily on weather and the surrounding terrains, thus accounting for the intermittent and stochastic nature of the relevant DG units. The employed control approaches are essentially driving the operating setpoints of dispatchable assets; hence, they are less of control in the classic sense and more of planning tools.
The size of the reserves scheduled were determined according to reliability indexes in light of the increased penetration of wind power. A somehow symmetrical approach was employed in , where the reliability of the intermittent RES (a photovoltaic system) was assessed, so as to adjust reserves accordingly. As is evident, the fact that both these methods leverage forecasting tools characterizes them as robust or adaptive, depending on whether the forecasts are incorporated as probability functions or otherwise lead to frequent updates of the set-points of the controlled assets. Stochastic methods have been another approach to handle RES variance by accounting for their output through probability density functions incorporated as cost factors and inequality constraints in security-aware dispatching set-ups.
Machine learning was used (binary decision trees, BDTs) to prepare dispatching set-points for handling multiple levels of excess or deficit of DG power. The BTDs were trained with actual cost and AC power flow data for either case of DG variance. This control method is a definitive example of a robust approach, while its application in an actual test-bed, as discussed in , highlights the requirements for versatile and low-cost hardware platforms with dispersed processing power. Pathways to a Smarter Power System
4.3 Voltage Regulation
Voltage regulation has become a major control concern in the past few years, especially at the level of distribution and for a twofold reason. The reason to consider the opportunity cost of reactive power generation is that it affects the frequency control capability of the generator to some degree. This paper proposed a distributed nonlinear control-based algorithm to achieve the optimal reactive power generation for multiple generators in a power grid. The reactive power control setting update for each generator only requires local measurement and information exchange with its neighboring buses. It was demonstrated that the proposed algorithm could reduce the nonconvex objective function monotonically until convergence and achieve comparable solutions to the centralized technique: particle swarm optimization with faster convergence speed.
The ever-increasing complexity of power systems, as large-scale dynamical systems, originates from integration of novel smart grid technologies (e.g., RES, demand response, and distributed energy storage), constraints Pathways to a Smarter Power System imposed by limitation of physical infrastructure, and growth of electric consumption.
These resources and technologies, on the one hand, can potentially improve the reliability of power systems by reducing dependency on bulk generation units, using more distributed and environmentally friendly alternatives, and enabling intelligent demand-side management. On the other hand, they introduce more uncertainty that makes the control and operation of power systems a challenging task. These uncertainties may also lead to unexpected operation difficulties, such as voltage/frequency deviations caused by mismatch between demand and supply. Further, deregulation of power systems has increased the number of stakeholders and entities, which results in a major concern regarding the scalability of the existing power systems control approaches.
From the above article, it is natural to focus future efforts in modern power systems control on enhancing these five characteristics by means of integrated design of efficient algorithms and software, such as distributed algorithms and parallel computing platforms, as well as deploying modern hardware-based technologies. As understood by the examples presented, most of the ongoing research in the field attempts to address as many of these characteristics as possible to offer readily applicable solutions to the ongoing challenges of the deregulated energy markets and the highly interacting control areas and regions throughout the grids worldwide.
Authors: Aman Deshmukh,Kunal Gaikwad,Priyanka Ghosh,Sudarshan Deshmukh.