A review of some methods of generator scheduling in the power system

Abstract:
Reinforcement Learning (RL), Generation Flexibility Assessment, Optimization Model for
Photovoltaic Generators, Coordinated Generation Scheduling, Mixed Integer Polynomial
Programming, Short-term Preventive Maintenance with DGs and Batteries, and Multi-Agent
Stochastic Energy Management Framework are just a few of the methods examined in this
review that are used to schedule generators. The relative advantages, disadvantages, and
prospective uses of each technique in contemporary power systems are explored. While
Generation Flexibility Assessment is essential for adjusting to large levels of renewable power
integration, RL shows promise in addressing real-time scheduling issues. A trustworthy
foundation for effective solar energy usage is provided by the Optimization Model for
Photovoltaic Generators. Through precise forecasts and intelligent power conversion,
coordinated generation scheduling successfully reduces the fluctuation of renewable energy
sources. The efficient and stable operation of power systems is ensured by mixed integer
polynomial programming, which enables thorough maintenance planning. By taking into
account the support capabilities of distributed generation and batteries, short-term preventive
maintenance with DGs and Batteries significantly lowers load losses. A powerful paradigm for
efficient energy management, particularly in scenarios with huge, nonlinear, and dynamic loads,
is introduced by the multi-agent stochastic energy management framework. This analysis gives
practitioners a comprehensive grasp of each method's unique characteristics, helping them to
make wise judgments in the dynamic world of power system operation and planning.
Keywords: generator, scheduling, reinforcement learning, generation flexibility assessment,
optimization model, coordinated generation scheduling
1.0 Introduction
In order to meet the rising demand for electricity while assuring dependability and cost
effectiveness, a power system must operate efficiently. In order to meet demand at the lowest
possible cost while maintaining system restrictions, generator scheduling is a basic component of
power system operation [1] [2].
In the past, real-time choices about the generation levels of power plants were made
using deterministic models like Economic Dispatch (ED) and Unit Commitment (UC) [3]. Under
a variety of restrictions, such as generator restrictions and demand requirements, ED focuses on
reducing the overall cost of generating. Contrarily, UC broadens this idea by taking launch and
shutdown expenses into consideration when committing or uncommitting units over a given
period of time.
The scheduling of generators has undergone a substantial change as a result of the
introduction of cutting-edge optimization techniques. Large-scale scheduling issues can now be
solved using linear programming (LP), mixed-integer linear programming (MILP), non-linear

1
programming (NLP), and quadratic programming (QP) [4]. Additionally, heuristic techniques
like dynamic programming and priority list methods have been extensively used to locate close
to optimal solutions in realistic, time-constrained circumstances [5][6][7].
Artificial intelligence (AI) and machine learning (ML), two emerging technologies, have
changed the way that generator scheduling is done. Artificial neural networks (ANNs) and fuzzy
logic (FL) have shown that they are capable of handling non-linearities and uncertainties related
to the integration of renewable energy [5]. Reinforcement Learning (RL) has also demonstrated
potential in terms of improving scheduling plans in dynamic and unpredictable contexts [8].
Generator scheduling is made more difficult by the increase in renewable energy sources,
which are notable for being sporadic and non-dispatchable. Scheduling hybrid systems and
stochastic models have become crucial methods for successfully integrating renewable energy
sources into the grid .
This review attempts to thoroughly assess the various approaches and strategies used in
generator scheduling in light of these advances, offering insights into their benefits, drawbacks,
and possible uses in contemporary power systems.
2.0 Problem Statement
In the context of power systems that increasingly rely on energy sources it has become
crucial to schedule generators efficiently. The main challenge lies in optimizing the allocation of
generation resources while considering factors, like the real time variability of outputs
integrating demand response and ensuring grid reliability through preventive maintenance. To
tackle these complexities, advanced techniques such as Reinforcement Learning (RL) and
Optimization Models for Photovoltaic Generators need to be employed. However striking a
balance between complexity, model accuracy and handling uncertainties is a challenge in this
ever changing landscape. Therefore it is essential to have an understanding of the strengths,
limitations and applications of these methodologies to make informed decisions, for efficient
power system operation and planning.

3.0 Methodology
This review compared and contrasted the methodologies used in generator scheduling
highliting their strengths and limitations. It further discussed future directions of the methods.

2
4.0 Literature Review on Methods of Generator Scheduling Application
Real-time scheduling of renewable power systems using reinforcement learning based on
planning was studied in the work of [8]. In a genuine power grid setting, they used cutting-edge
reinforcement learning (RL) algorithms. Significant reductions in load shedding and renewable
curtailment are achieved by the well-trained RL-based scheduling agent. Compared to
conventional scheduling techniques, which mainly rely on potentially wrong day-ahead
projections, this is a significant improvement. The study also emphasizes the advantages of high-
frequency control decisions, exploiting flexibility in the current infrastructure to reduce hardware
changes and associated costs.

Figure 1: Real Time Scheduling [8]

Day-Ahead Scheduling (DAS), in accordance with the future 36-hour forecasts, which
take into account load and renewable generation, executes day-ahead unit commitment to
establish the startup/shutdown schedule of thermal generators. However, as the prediction time
lengthens, the predictability of renewable generation will decrease. This adds mistakes into the
unit commitment optimization process, producing an arbitrary output of a 0-1 matrix. The effects
of errors will be largely mitigated by the upcoming intraday economic dispatch, but when the
error scale is severe, renewable curtailment and load shedding will happen. RTS, on the other
hand, uses a finer time resolution to modify power generators. Instead of the day level of the
traditional day-ahead scheduling, the control interval is at the minute level. Accurate ultra-short-

3
term forecasts and data-driven, computationally efficient reinforcement learning algorithms are
used to achieve this. The scheduling issue returns to becoming a deterministic control issue
during the brief window of precise ultra-short-term forecasts. In order to jointly optimize unit
commitment and economic dispatch, the RL agent simultaneously regulates output power and
the startup and shutdown of generators. Additionally, our suggested GridZero has the ability to
schedule tasks in advance for upcoming choice scenarios.
The study by [1] was designed to plan for flexible generators and energy storages under
Taiwan's high renewable energy penetration rate. This study's main goal is to investigate how
generation flexibility affects unit scheduling costs. A multi-scenario study for the integration of
renewable power is taken into consideration, and a flexibility index is employed to assess the
generating flexibility in the Taiwan power system. The findings of this study can serve as a guide
for power systems as they develop flexible generating units and energy storage systems for the
integration of a significant proportion of future renewable energy generation.
Flexibility of the power system can be explicitly measured using a number of criteria. In
multi-conditional judgment problems, the fuzzy analytic hierarchy process (FAHP) and the
analytic hierarchy process (AHP) have been particularly useful. AHP is employed for making
decisions in unclear situations. AHP measures are frequently used in the power sector to evaluate
plant dependability or conduct repair operations. When expert judgments are translated into
values, [1] indicated that AHP might not be able to assess human subjectivity, while FAHP
might be able to do so with greater objectivity. In order to assess the flexibility of power
generation, this study uses FAHP [9].
The estimation of power generation flexibility often takes into account the unit's start-up
and shut-down periods, ramp-up rate, minimum generator output, and operating range. The
flexible index of each generator is determined by this study using the aforementioned variables
and the FAHP approach [8]. Since the units and ratios of the flexibility variables at each
generator differ, these variables are normalized using a technique that takes the total system
flexibility into account, which is stated as follows:
x −base min
I= , (1)
base max−basemin
x−base min
I =1− (2)
base max −base min

4
x is one of the variables at a unit, I is the value after normalization, and base-max and base-min
are the maximum and minimum values at each variable, respectively. It is vital to give each
variable the proper weight, which was determined through an expert questionnaire, in order to
indicate the contribution of each variable to the system's overall flexibility. based on the
suggestions of ten power-related specialists [8] [10].
A reliable quadratic convex model for the best scheduling of solar generators in
unbalanced bipolar DC grids is presented in the work of [11]. The suggested model is founded
on Taylor's series expansion, which loosens the hyperbolic relationship between constant power
terminals and voltage profiles. To lessen the inaccuracy caused by assumed relaxations, the
suggested model is also solved in the recursive form. The solutions of (Formula presented) kW,
(Formula presented) kW, and (Formula presented) kW for 0%, 50%, and 100% of the capacity
PV rating, respectively, show the usefulness of the proposed recursively solved convex model in
achieving the global optimum for three levels of PV generator capacity.
Similar to this, [12] proposed the ideal coordinated generation schedule taking into
account the uncertainty of the day-ahead PV and wind power forecasts. Wind and photovoltaic
power generation in particular are unpredictable, however forecasting algorithms can be used to
predict them. A hyper-parameterized long short-term memory (LSTM) regression model is used
in the suggested framework to forecast the weather for the day ahead using past time-series
weather data. Finally, a day-ahead weather forecast is utilized to estimate the power conversion
for a particular PV module and wind turbine using an empirical formula. The case studies
demonstrate that coordinated generation scheduling can enable stable supply-demand operation,
ameliorate the power fluctuation problem associated with RESs, and reduce operating costs
associated with electricity generation.
Additionally, [5] proposed a method for generator scheduling maintenance in a power
system that incorporates demand response by transforming mixed integer polynomial
programming. The GMS approach is intended to handle different system needs and demand
response characteristics within a power system. Data from the Korean power system were used
in a case study to show how well the suggested approach for figuring out the ideal maintenance
plan works. The findings demonstrate that by taking into account the relevant system restrictions,
the suggested GMS approach may be employed to facilitate the efficient and dependable
operation of a power system.

5
 Additionally, [2] describe a novel approach to short-term preventive maintenance that
explicitly takes into account the support potential of batteries and DGs as well as the
unpredictability of DG power generation. There are two main concerns covered. The network is
divided into zones using a depth-first-search clustering algorithm to address the network's large-
scale complexity. Additionally, a technique is put forth for capturing the impact of maintenance
choices in the model of the served load from batteries and DGs through the creation of
topological constraints. The short-term maintenance schedule is then established using a mixed-
integer non-linear programming problem based on stochastic scenarios. Through a case study
based on a modified IEEE-34 bus distribution network, where the researchers also compare a
branch-and-bound and a particle swarm optimization solver, we demonstrate the effectiveness
and efficiency of the suggested approach. The findings also demonstrate that load losses can be
significantly reduced by preventative maintenance scheduling thanks to the supporting capacity
of DGs and batteries.
Additionally, [13]examined the optimization of generation scheduling in power networks
with the integration of renewables on several time scales, including medium- and long-term,
short-term, and real-time, respectively. The scheduling paradigm and method are outlined first.
The links and distinctions between the multi-source mathematical model with uncertainty are
also highlighted, as well as the market mechanisms for thermal power, hydroelectric power, wind
power, solar power, and energy storage. Second, the two dimensions of certainty and uncertainty
are used to separate the scheduling method and strategy. Between the traditional scheduling
problem and the scheduling problem with renewable energy, an innovative method and its
differences are described. In the meanwhile, each section highlights how the various temporal
scales interact and are coupled. Dispatchers are also given references for future directions and
information on the difficulties and weaknesses of present research.
Techniques and resolution approaches were looked at in a different study by [14]. It was
suggested to review the unit commitment problem. The main objective of this state-of-the-art is
to identify the research gaps that might be amenable to additional advancements as well as to
facilitate new researchers' knowledge of the various methodologies used in the literature.
Additionally, according to its dominant predominance when the unit commitment problem is
addressed, a summary of the evolution of the Mixed Integer Linear Programming formulation in
terms of the advancements of commercial solvers is shown. The case studies' computational

6
performance, power system representation, and modeling specifics have all been accurately
analyzed. Comparing this characterisation to traditional assessments, which only provide a
general view of the modeling, represents a considerable advancement.
Similar to this, [15] proposed an efficient stochastic operation paradigm for the shipboard
power systems, encompassing huge, nonlinear, and dynamic loads, to manage energy efficiently.
The suggested architecture separates the ship power system into a number of agents who
communicate with one another in accordance with their demands and supply. The reformulated
distributed energy management problem in the ship is solved using a multi-agent framework
based on the alternating direction method of multipliers (ADMM). The suggested system model
takes into account both single-shaft and twin-shaft turbo-generators to improve part-load
efficiency at specific moments when dealing with variable speed operation. A potent
optimization technique based on the modified-firefly algorithm (M-FOA) is suggested because
each agent's nonlinear cost function necessitates it. A phasor algorithm like this one aids in
avoiding premature convergence and becoming stuck in local optima. On an actual dataset of a
ship power system, the proposed stochastic model's suitability is evaluated. The simulation's
findings demonstrate the system's high robustness, guaranteed consensus, economical operation,
and workable solution in the event of a load-shedding-based power generating deficit.

7
5.0 Comparative Analysis of the Strengths, Limitations, and future work on Generator Scheduling Methods

Table 1: Strengths, limitations and future work on generator scheduling method

Method Application Strengths Limitations Future work

Reinforcement Real-time scheduling of - A significant decrease in Doesn't describe any Further study could
Learning (RL) renewable power load shedding and renewable potential difficulties or concentrate on developing
systems through energy curtailment. limitations encountered RL algorithms for more
planning-based - Makes use of a cutting-edge when using RL to intricate grid
reinforcement RL algorithm, demonstrating power scheduling. environments and
learning [16] the potential for investigating uses outside
advancements in real-time of real-time scheduling.
scheduling.
Generation Planning of Flexible -Provides helpful Does not specifically The application in
Flexibility Generators and Energy suggestions for setting up mention the difficulties practice and cost-benefit
Assessment Storages under High flexible generating units and or computational evaluation of suggested
Penetration of energy storage systems. - complexity involved in flexible generation
Renewable Power in Addresses the important need calculating the systems could be the
Taiwan Power [1] for flexible generation in the flexibility index. subject of further study.
face of rising renewable
integration.
Optimization Robust quadratic Achieves global optimum for Does not discuss any Testing the concept in
Model for convex model for various PV generating computational issues or various grid setups and
Photovoltaic optimal scheduling of capacities, highlighting the potential investigating scalability
Generators photovoltaic resilience of the suggested implementation for larger systems could
generators [11] approach. difficulties for the be the main topics of
suggested model. future research.

Coordinated Coordinated generation - Uses coordinated Lacks a description of The forecasting models
Generation scheduling considering generation scheduling to any potential could be improved, and
Scheduling day-ahead PV and wind address the variability uncertainties or the methodology could be
power forecast problem with renewable problems with the tested in various grid
uncertainty [12] energy sources. forecasting model. scenarios, in future
8
 -Shows the potential for cost studies.
reduction and safe supply-
demand operation.

Mixed Integer Generator Maintenance Outlines a strategy that may Does not address The practical usefulness
Polynomial Scheduling Using be adjusted to deal with potential difficulties or and scalability of the
Programming Mixed Integer different system needs and restrictions encountered suggested technique in
Polynomial demand response when applying the various power systems
Programming [5] characteristics. suggested strategy. might be investigated in
further detail.

Short-term Short-term preventive Focuses on network Does not specify any The robustness of the
Preventive maintenance method complexity concerns on a potential computing technique in various
Maintenance with considering support broad scale and efficiently difficulties or network topologies and
DGs and Batteries potential of DGs and integrates DGs and batteries difficulties in applying grid configurations could
batteries [2] into preventive maintenance the suggested be investigated in more
planning. technique. detail.
Multi-Agent Stochastic operation Uses a multi-agent Does not address any Future studies should
Stochastic Energy framework for optimal architecture to address the difficulties that might concentrate on examining
Management energy management of complexity of shipboard arise in putting the the suggested
Framework shipboard power power systems, assuring suggested stochastic framework's applicability
systems [17] reliable and efficient operation paradigm into to larger ships and
operation. practice. validating it using actual
onboard power systems.

9
Figure 1 presents a number of methods used for generator scheduling with the applications,
strengths, limitations and future work. A closer look at this methods present an interest insight
into their usefulness in the power system. A comparative analysis of these methods is presented
below:

Reinforcement Learning (RL) presents leading-edge method to generator scheduling in

real-time situations. It utilizes AI-driven algorithms to make decisions based on feedback from
the environment. RL shows high compliance to change in conditions and is capable of handling
uncertainties connected to renewable energy generation. Although, it might need extensive
computational resources and training data that can be disadvantage in some contexts. In addition,
the performance of RL depends on the quality of the underlying model and the representation the
state space. Its has the potential to decrease renewable curtailment and load shedding makes it a
promission methodology, particularly in the terms of integrating unpredictable renewable
sources into modern power systems.

Relatedly, In order to adapt power systems to the growing integration of renewable

energy sources, Generation Flexibility Assessment is crucial. This process assesses a generator's
capacity to modify its output in reaction to shifting circumstances. It offers vital information on
how well the system can adapt to variations in renewable energy production. This strategy is
particularly useful for guaranteeing the stability and dependability of the grid. But it needs
precise information about the qualities and capacities of every generator, which isn't always easy
to come by. Furthermore, even though it tackles the flexibility issue, it does not recommend
certain scheduling choices, which makes it a supplementary approach rather than a stand-alone
one.

An efficient method for scheduling solar energy generation is the Optimization Model for
Photovoltaic Generators. It uses Taylor's series expansion and quadratic convex models to relax
the hyperbolic relations between voltage profiles and constant power terminals. This makes it
possible to predict the behavior of solar generators more precisely. The approach works
especially well in bipolar DC grids that are out of balance. However, compared to more universal
generator scheduling techniques, its applicability might be restricted to particular photovoltaic
generation conditions.

10
 A strategic method for addressing the unpredictability of renewable energy sources is
coordinated generation scheduling, which incorporates forecast uncertainty into the scheduling
procedure. It makes day-ahead weather predictions using sophisticated forecasting models, like
hyper-parameter-optimized LSTM regression. The estimation of power generation from
renewable sources is then done using this data. The approach provides enhanced supply-demand
operations security and reduces power variability by coordinating generation schedules based on
these projections. This method works very well to improve the way renewable energy is
integrated into power systems. Its effectiveness is dependent, nevertheless, on how accurate the
forecasting models used are.

An extensive optimization framework for solving the unit commitment problem in power
systems is provided by Mixed Integer Polynomial Programming. It makes it possible to take into
account several energy sources at once, such as solar, wind, hydroelectric, and thermal energy.
By expressing the scheduling choices as a mathematical optimization problem, this method
offers a solid solution. However, the complexity of large-scale power systems, where processing
time and computational resources become crucial, may restrict its application. Furthermore, it
makes the assumption that power generation and variables have a linear connection, which may
not always adequately reflect the nonlinear behavior of some generators.

This methodology presents a preventive maintenance strategy that specifically takes

energy storage systems and distributed generators (DGs) into account. It seeks to minimize load
losses by strategically scheduling maintenance tasks while taking DG output uncertainties into
account. It is efficient in handling large-scale distribution networks because it makes use of
stochastic scenario-based optimization and depth-first-search clustering. Its use, though, might
be restricted to situations in which DGs and batteries are important for grid support and power
generation. Furthermore, proper modeling of the behavior of distributed generation and
maintenance needs is essential to its success.

A practical method for achieving the best possible energy management in shipboard
power systems with big, nonlinear, and dynamic loads is presented by the Multi-Agent
Stochastic Energy Management Framework. It optimizes energy distribution by breaking the
system up into many agents that coordinate according to their supplies and requests. Robust
performance is ensured by the employment of a modified θ-firefly algorithm and the alternating

11
direction method of multipliers (ADMM) for optimization. Its use, meanwhile, might be limited
to shipboard systems and might not be immediately applicable in more general power system
settings.

Every one of the aforementioned approaches has unique advantages and specific uses in the field
of generator scheduling. The type of renewable integration, the specific system requirements,
and the available computational resources should all be taken into consideration when selecting a
methodology. Furthermore, taking into account the possible synergies between these techniques
in an integrated scheduling approach might produce better outcomes in contemporary power
systems.

6.0 Future Directions in Generator Scheduling Methods

The studied generator scheduling techniques have been used to address a variety of power
system issues. However, there is still plenty that can be done to guarantee stability and an
optimal power supply. Below is an outline of the future direction of research:

1. Planning-based reinforcement learning for real-time scheduling [16]: Additional study

may examine the incorporation of cutting-edge AI methods outside of reinforcement
learning, such as deep reinforcement learning or meta-learning, to improve real-time
scheduling responsiveness and accuracy. The scalability and adaptation of the RL-based
agent to larger and more complicated power systems may also be an important subject to
explore in the future.
2. Energy storage and flexible generator planning [1] : In order to accommodate even larger
levels of renewable energy penetration, future studies may concentrate on the creation of
sophisticated optimization algorithms for flexible generator and energy storage planning.
To further increase the flexibility index and the precision of generation flexibility
evaluations, research may further focus on the creation of predictive models for the
generation of renewable energy.
3. Solar Generator Scheduling Using Metaheuristic Algorithms or Machine Learning-Based
Approaches [11]: Future study may examine the use of advanced optimization techniques
to schedule photovoltaic generators more accurately and efficiently. A promising field of
research may also focus on how to expand this model to support multi-objective
optimization while taking into account aspects like cost and environmental effect.

12
 4. Taking Forecast Uncertainty into Account When Scheduling Coordinated [12]: Future
research could concentrate on improving day-ahead forecast accuracy for renewable
energy sources using more complex forecasting models, possibly combining ensemble
methods or deep learning techniques. Future study may also benefit from looking into
how advanced control techniques like model predictive control can be used for
coordinated generation scheduling.
5. Generator Scheduling Maintenance with Demand Response [5]: Future research may
focus on creating more complex optimization models that take into consideration
changing grid dynamics, such as enhanced demand response and distributed energy
resource integration. Investigating the use of machine learning algorithms for predictive
maintenance scheduling may also present chances for process improvement and
automation.
6. Short-term Preventive Maintenance Using Batteries and Distributed Generation [2]:
Future research could examine how to handle even larger-scale distribution networks
with a greater number of zones and components by applying advanced clustering
techniques and optimization algorithms. Research may also concentrate on incorporating
more precise and thorough models for the behavior of distributed generating and battery
systems, taking aging and degradation into account.
7. Power system generation scheduling optimization with integration of renewable energy
sources [13]: Future studies might focus on the creation of more sophisticated
optimization algorithms that can manage the rising complexity and unpredictability
related to the integration of renewable energy sources. Additionally, investigating the
integration of machine learning methods for renewable energy projection and demand
forecasting could improve the precision of scheduling decisions over a range of time
scales.

7.0 Conclusion

This review has investigated a wide range of generator scheduling methodologies, from
cutting-edge approaches like Reinforcement Learning (RL) and Optimization Models for
Photovoltaic Generators to tried-and-true methods like Generation Flexibility Assessment and
Coordinated Generation Scheduling. Each approach has particular advantages and addresses

13
 particular problems in contemporary power systems. While Generation Flexibility Assessment is
essential for adjusting to the integration of renewable power sources, RL has considerable
potential to improve real-time scheduling. A reliable foundation for effective solar energy
consumption is provided by the Optimization Model for Photovoltaic Generators. The stability
and effectiveness of power systems are further ensured by techniques like Coordinated
Generation Scheduling, Mixed Integer Polynomial Programming, Short-term Preventive
Maintenance using DGs and Batteries, and the Multi-Agent Stochastic Energy Management
Framework. This review equips practitioners with the knowledge to make wise judgments in the
constantly changing world of power system operation and planning by recognizing the particular
characteristics and uses of these methodologies. A reliable foundation for effective solar energy
consumption is provided by the Optimization Model for Photovoltaic Generators. The stability
and effectiveness of power systems are further ensured by techniques like Coordinated
Generation Scheduling, Mixed Integer Polynomial Programming, Short-term Preventive
Maintenance using DGs and Batteries, and the Multi-Agent Stochastic Energy Management
Framework. This review equips practitioners with the knowledge to make wise judgments in the
constantly changing world of power system operation and planning by recognizing the particular
characteristics and uses of these methodologies.

REFERENCES

[1] Y. K. Wu, W. S. Tan, Y. S. Chiang, and C. L. Huang, “Planning of Flexible Generators and Energy Storages
under High Penetration of Renewable Power in Taiwan Power System,” Energies (Basel), vol. 15, no. 14,
2022, doi: 10.3390/en15145224.

[2] J. Fu, A. Nunez, and B. De Schutter, “A Short-Term Preventive Maintenance Scheduling

Method for Distribution Networks with Distributed Generators and Batteries,” IEEE
Transactions on Power Systems, vol. 36, no. 3, pp. 2516–2531, 2021, doi:
10.1109/TPWRS.2020.3037558.
[3] M. A. Tijani, G. A. Adepoju, and M. O. Okelola, “Optimization Approaches to Generation
Dispatch Problems: Review of Nigerian Power System,” Nigerian Journal of Technological
Development, vol. 19, no. 2, pp. 156–163, 2022, doi: 10.4314/njtd.v19i2.8.
[4] F. Glover, M. Lewis, and G. Kochenberger, “Logical and inequality implications for reducing
the size and difficulty of quadratic unconstrained binary optimization problems,” Eur J Oper
Res, vol. 265, no. 3, pp. 829–842, 2018, doi: 10.1016/j.ejor.2017.08.025.

14
[5] H. C. Jo, R. Ko, and S. K. Joo, “Generator Maintenance Scheduling Method Using
Transformation of Mixed Integer Polynomial Programming in a Power System Incorporating
Demand Response,” Energies (Basel), vol. 12, no. 9, 2019, doi: 10.3390/en12091646.
[6] Y. Zhou, W. J. Lee, R. Diao, and D. Shi, “Deep Reinforcement Learning Based Real-time AC
Optimal Power Flow Considering Uncertainties,” Journal of Modern Power Systems and Clean
Energy, vol. 10, no. 5, pp. 1098–1109, 2022, doi: 10.35833/MPCE.2020.000885.
[7] J. H. Woo, L. Wu, J. B. Park, and J. H. Roh, “Real-Time Optimal Power Flow Using Twin
Delayed Deep Deterministic Policy Gradient Algorithm,” IEEE Access, vol. 8, pp. 213611–
213618, 2020, doi: 10.1109/ACCESS.2020.3041007.
[8] C. O’Malley, P. de Mars, L. Badesa, and G. Strbac, “Reinforcement learning and mixed-integer
programming for power plant scheduling in low carbon systems: Comparison and hybridisation,”
Appl Energy, vol. 349, Nov. 2023, doi: 10.1016/j.apenergy.2023.121659.
[9] M. H. Andishgar, A. Fereidunian, and H. Lesani, “Healer reinforcement for smart grid using
discrete event models of FLISR in distribution automation,” Journal of Intelligent and Fuzzy
Systems, vol. 30, no. 5, pp. 2939–2951, 2016, doi: 10.3233/IFS-151814.
[10] P. de Mars and A. O’Sullivan, “Applying reinforcement learning and tree search to the unit
commitment problem,” Appl Energy, vol. 302, p. 117519, Nov. 2021, doi:
10.1016/J.APENERGY.2021.117519.
[11] O. D. Montoya, W. Gil-González, and J. C. Hernández, “Optimal Scheduling of Photovoltaic
Generators in Asymmetric Bipolar DC Grids Using a Robust Recursive Quadratic Convex
Approximation,” Machines, vol. 11, no. 2, 2023, doi: 10.3390/machines11020177.
[12] S. Admasie, J. S. Song, and C. H. Kim, “Optimal coordinated generation scheduling considering
day-ahead PV and wind power forecast uncertainty,” IET Generation, Transmission and
Distribution, vol. 17, no. 11, pp. 2545–2562, 2023, doi: 10.1049/gtd2.12868.
[13] K. Chen et al., “Optimization of multi-temporal generation scheduling in power system under
elevated renewable penetrations: A review,” Front Energy Res, vol. 10, no. January, pp. 1–11,
2023, doi: 10.3389/fenrg.2022.1054597.
[14] L. Montero, A. Bello, and J. Reneses, “Techniques , and Resolution Methods,” MDPI Energies,
vol. 15, no. 4, 2022.
[15] M. A. Mohamed, H. Chabok, E. M. Awwad, A. M. El-Sherbeeny, M. A. Elmeligy, and Z. M.
Ali, “Stochastic and distributed scheduling of shipboard power systems using MθFOA-ADMM,”
Energy, vol. 206, p. 118041, 2020, doi: 10.1016/j.energy.2020.118041.
[16] S. Liu et al., “Real-time scheduling of renewable power systems through planning-based
reinforcement learning,” 2023, [Online]. Available: http://arxiv.org/abs/2303.05205
[17] S. Goleijani and M. T. Ameli, “A multi-agent based approach to power system dynamic state
estimation by considering algebraic and dynamic state variables It is valid until 50 days after
July 26 th 2018 . Share Link :,” no. July, 2018, doi: 10.1016/j.epsr.2018.07.019.

15
16