Artificial Intelligence Techniques For Enhancing T

This article has been accepted for publication in IEEE Open Journal of Industry Applications.
This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/OJIA.2023.3338534
IEEE OPEN JOURNAL OF INDUSTRY APPLICATIONS 1
Artificial Intelligence Techniques for Enhancing the

Performance of Controllers in Power
Converter-based Systems – An Overview
Yuan Gao, Member, IEEE, Songda Wang, Member, IEEE, Tomislav Dragicevic, Senior Member, IEEE,
Patrick Wheeler, Fellow, IEEE, and Pericle Zanchetta, Fellow, IEEE
Abstract—The integration of artificial intelligence (AI) tech- system. However, these rule-based systems were limited in
niques in power converter-based systems has the potential to their ability to adapt to changing conditions and had difficulty
revolutionize the way these systems are optimized and controlled. handling more complex systems. Over the years, AI techniques
With the rapid advancements in AI and machine learning
technologies, this paper presents the analysis and evaluation of have evolved to include more sophisticated approaches such
these powerful tools as well as in computational capabilities of as expert systems, neural networks, and fuzzy logic systems,
microprocessors that control the converter. This paper provides etc. These techniques have been applied to a wide range of
an overview of AI-based controllers, with a focus on online/offline power electronic systems, including power converters, active
supervised, unsupervised, and reinforcement-trained controllers. rectifiers, inverters, and motor drives. However, despite the
These controllers can be used to create surrogates for inner
control loops, complete power converter controllers, and external significant progress made in this field, there are still many
supervisory or energy management control. The benefits of challenges that need to be addressed to fully realize the
using AI-based controllers are discussed. AI-based controllers potential of AI in power electronic systems.
reduce the need for complex mathematical modeling and enable Recently, there has been an interest in the use of AI
near-optimal real-time operation via computational efficiency. techniques in power electronic systems, driven by the rapid
This can lead to increased efficiency, reliability, and scalability
of power converter-based systems. By using physics-informed advancements in machine learning (ML). These advancements
methods, a deeper understanding of the underlying physical have been made both in the algorithm area and in the com-
processes in power converters can be achieved and the control putational capabilities of power converter controllers, which
performance can be made more robust. Finally, by using data- led to a significant improvement in the ability of AI systems
driven methods, the vast amounts of data generated by power to handle power electronic systems. The design, control, and
converter-based systems can be leveraged to analyze the behavior
of the surrounding system and thereby forming the basis for preventative maintenance are three key areas where AI is
adaptive control. This paper discusses several other potential currently being used in power converter-based systems [1].
disruptive impacts that AI could have on a wide variety of power Three parts are illustrated in Fig. 1.
converter-based systems. The design optimization of converter-based systems usually
Index Terms—Artificial Intelligence (AI), Machine Learning requires a huge computing power to simulate many feasible
(ML), Neural Network (NN), Power converter control, Energy design candidates and select the optimal one, especially if
management, Renewable energy integration. their fitness needs to be evaluated in various mission profiles.
The main reason for the high computing load is that converter
I. I NTRODUCTION needs to be simulated in heat defined design process in itself
is complex in nature where mutually coupled performance
T HE integration of artificial intelligence (AI) techniques
in power electronic systems has been an active area of
research for several decades. The earliest attempts to integrate
indices such as efficiency, power density, cost, and reliability,
should be considered [2]. In addition, intelligent control is
AI in power electronic systems focused on rule-based systems, essential for the reliable, robust, and stable operation of power
where a set of predefined rules were used to control the converter-based systems. For example, when AI algorithms
are added to the traditional PI/PR or advanced non-linear
Manuscript received May 14, 2023; revised September 21, 2023; accepted controllers, the dynamic response and tracking performance
November 27, 2023. This work was supported by the Marie Skłodowska-
Curie Actions (MSCA) under the European Union’s Horizon 2020 Research may be improved in several aspects, such as reliability to
and Innovation Staff Exchange (RISE) under Grant 872001. (Corresponding changing system conditions, reduction of computational load
author: Yuan Gao) or consequent performance improvement [3], [4], [5]. Fi-
Yuan Gao is with the School of Engineering, University of Leicester, LE1
7RH Leicester, U.K. (e-mail: [email protected]). nally, preventive maintenance of systems, including condition
Songda Wang is with Shell Global Solutions International B.V., 2596 HR monitoring, fault tolerance, and fault diagnosis, are effective
Den Haag, The Netherlands (e-mail: [email protected]). approaches to ensure a system’s healthy operation. With the
Tomislav Dragicevic is with the Department of Energy Technol-
ogy, Technical University of Denmark, 2800 Kongens, Denmark (e-mail: help of AI, the desired prediction and monitoring can be
[email protected]). achieved much faster and more precisely than with traditional
Patrick Wheeler and Pericle Zanchetta are with the Power Electronics methods [6], [7], [8].
Machines and Control (PEMC) Group, University of Nottingham, Not-
tingham NG7 2RD, U.K. (e-mail: [email protected]; Peri- In [1], a general review of AI algorithms and applications
[email protected]). for power electronics was presented from a life-cycle per-
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in IEEE Open Journal of Industry Applications. This is the author's version which has not been fully edited and
spective. The paper generally considered most AI techniques:

expert systems, fuzzy logic, heuristics, and ML, in three
distinctive life-cycle phases: design, control, and maintenance.
Therefore, it is not specific to the controller domain in power
electronics. Supervised and unsupervised learning techniques
for electrical power systems were reviewed in [9]. Though Linear Condition
popular ML algorithms were introduced, such as recurrent Monitoring
neural networks and random forest, the application areas in Fault

Non-
design, control, and maintenance phases were not clearly Linear
Tolerance &
categorized and reinforcement learning (RL) was not consid- Diagnosis
ered. For microgrid applications, the survey of AI techniques Drive Parameter

can be found in [10], [11]. [10] discussed the AI applica- Sizing Identification
tions in different aspects of microgrids: energy management,
Design
load/generation power forecasting, power converter control,
fault detection, cyber attacks, and protection schemes. In Fig. 1: Diagram of AI applications in power electronics.
[11], authors reviewed the RL-enabled power management
in grid-tide microgrids. Also focusing on RL, [12] reviewed On/off-line
AI applications in power and energy systems, based on the Learning Mode
Learning Methods
Application
introduction of RL principles and categorization.
(a). Unsupervised Trained off-line,
Due to the specific control features and challenges of Regression
Learning Tested online
power converter-based systems, such as high-speed switching, (b). Supervised
complex modulation, and high-computational burden, AI im- Learning
(c). Reinforcement
plementation in the control of these systems will be different Classification Learning Applied online
from the other two life-cycle phases, namely design, and
maintenance. Therefore, there is an urgent need for an analysis
of AI applications in the control field for power converter- (a). (b). (c).
based systems. In recent years, dramatically increasing at-
x o
…
tention has been paid to AI applications in the control of
…
both individual converters and converter-based (micro)grid
systems. This paper presents the analysis and summary of Fig. 2: Diagram of ML methods, modes, and on/off-line
the state-of-the-art research on AI and its application status applications.
in enhancing controller performance. Regarding the system
topology, both inner-loop converter control and (micro)grid-
level energy management will be considered in this paper. Three main groups of ML algorithms are supervised learn-
The remainder of this paper is organized as follows. Section ing (SL), unsupervised learning (UL), and reinforcement
II will list the popular AI techniques (mainly in machine learning (RL) [1], [14], as shown in Fig. 2. Generally, both
learning) used in this target domain. After that, in Section SL and UL require collecting data before training; An SL
III & IV, this paper will focus on two aspects of AI-aided dataset should give outputs/labels corresponding to the inputs,
controllers: linear and no-linear. Both design and control while there is no label defined in UL. For RL, there is no
applications are discussed in these two sections. Furthermore, data collection or offline training, because it is determined
the specific AI application areas will be reviewed in Section by the trial-and-error exploration of agents in an unknown
V. Finally, this paper is concluded in Section VI. environment in order to maximize the cumulative rewards.
As shown in Fig. 2, two learning modes in ML are regres-
II. A RTIFICIAL I NTELLIGENCE T ECHNIQUES sion and classification. And in practice, different on/off-line
approaches should be investigated for different applications.
AI can be implemented using predefined rules, but in most In this paper, both online and offline ML model training
cases, ML algorithms are used to perform specific AI tasks are reviewed for the parameter design of controllers; but,
based on learning from the collected data. ML algorithms can for controller imitation learning, a common way is using the
effectively learn rules and relations from training data and collected data for offline training. Based on the input/output
improve the trained models automatically through experience. model design and the data-driven process, ML can be an
Therefore, the largest use of AI in power converter-based effective way to enhance the control performance of converter-
systems is with ML. Other common AI methods include based systems. Three ML groups will be introduced below,
expert systems, fuzzy logic, and meta-heuristic methods. To then ML-based emerging techniques will be discussed.
distinguish ML from the meta-heuristic search algorithms,
[13] proposed a simple algorithm categorization that comprises
Search Algorithm and Surrogate Algorithm (2SA). Both cat- A. Supervised Learning (SL)
egories were used for the same optimization problem for a If there are outputs (also called targets and answers) defined
converter-based actuation system. in the training data of ML, SL methods are usually applied.
These targets can either be continuous numbers or integers. D. ML-Based Emerging Techniques
If the targets are all integers/categories, the learning task is The above three subsections individually introduced three
known as classification, otherwise, it would be a regression groups in ML; However, the practical problems may need
task. Therefore, both regression and classification are training multiple groups of algorithms to address, and even an in-
the sampled input-output pairs but their output data features dependent algorithm could encompass more than one type
are different. Regression has no requirement for sampled of learning method. For example, deep deterministic policy
data but, in classification learning, the training outputs should gradient usually combined SL (deep neural network) with RL
include predefined categories/classes or even only 0 or 1 for [19], [20].
some specific problems. With the fast development of AI and ML, there are emerging
According to the statistics in [1], usage of SL is 91% of all ML techniques that came to worldwide attention in recent
ML applications in power electronics. In particular, Artificial years. Federated learning (FL), also known as collaborative
Neural network (ANN) is one of the most common algorithms learning, is one of the emerging techniques. It is first intro-
in SL where neurons are the fundamental information process- duced by Google in 2016 [21]. FL typically applies when
ing units and the building blocks. Other SL algorithms include individual actors need to train models on larger datasets
Linear Regression [15], Support Vector Machine (SVM) [13], than their own, but cannot afford to share the data in itself
[16], Random Forest [17], etc. with others (e.g., for legal, strategic, or economic reasons).
The technology yet requires good connections between local
B. Unsupervised Learning (UL) servers and minimum computational power for each node [22],
[23]. FL trains an algorithm across multiple decentralized edge
For UL, there is no predefined output in the training data. devices or servers holding local data samples, without ex-
UL learns how to discover patterns and information from a changing them. This approach stands in contrast to traditional
dataset without preset outputs/labels/features. Such methods centralized ML techniques where all the local datasets are
can also be used to reduce data dimensionality without losing uploaded to one server, as well as to more classical decentral-
important information. Therefore, the training data for UL can ized approaches which often assume that local data samples
only have certain elements. For example, in a flower category are identically distributed. There are three parts to federated
case, there are 4 elements: sepal length, sepal width, petal learning: centralized, decentralized, and heterogeneous. FL has
length, and petal width. The UL can discover the information been effectively applied to self-driving [24], wireless power
that was previously undetected. An example is to categorize control [25], and large-scale energy systems [23]. With the
different kinds of flowers by using some sampled data of 4 continuing development of distributed large-scale converted-
elements. based systems, FL could become a promising AI technique
UL problems can be further grouped into clustering and for wireless computing and power control.
association problems. Clustering mainly deals with finding a Another popular technique in ML is Transfer Learning (TL),
structure or pattern in unlabelled data. The flower category it has found several applications in electrical power systems.
case is indeed a clustering problem. Association rules can TL leverages knowledge gained from one task and applies
discover associations/relationships between elements in large it to a related but different task to pursue high performance
databases. For example, people that buy a new house are most and efficiency. TL is desirable for the practical problems that
likely to buy new furniture. the training data is expensive or impossible to recollect [26].
Some application areas of TL in electrical domain are energy
C. Reinforcement Learning (RL) management [27], battery health [28], impedance estimation
[29], [30], and thermal performance of motor [31].
The third group of ML is RL, which is learning how
intelligent agents can perform a task by interacting with their
environment. Unlike classical and heuristic optimization meth- III. AI A PPLICATIONS IN L INEAR C ONTROLLERS OF
ods, RL does not require an accurate model of the system or P OWER C ONVERTERS
environment to generate an optimal solution [11], [18]. In the Linear controllers, such as cascaded voltage and current
learning process, these agents ought to take action (updating loops, are widely used in power converter systems to regulate
the policy at the same time) in a specific environment. They the power flow and maintain stability. However, these con-
would get rewards from each step/trying and the final target trollers are designed to operate in linear systems and can have
is usually to maximize the cumulative reward. difficulty performing optimally in nonlinear systems. There-
Therefore, in the beginning, how to operate in the environ- fore, for optimal performance, the parameters of linear control
ment is unknown, and no trustworthy data can be collected. loops should be adapted to the changing operating conditions
The agents should try different actions in the environment to in nonlinear surrounding environments. In this context, AI-
receive the corresponding rewards. Based on the cumulative based techniques can be used to adapt the parameters of linear
rewards, the agents can learn a good policy (which means how controllers in power converters.
to act in a certain position) in the environment. Obviously, RL Due to the possession of non-linearity, and the ability of
is an online learning process. It differs from SL or UL because online learning (transferring the real-time experience to AI
no input/output data pairs are collected before the learning models), the AI-aided PI/PR online control methods have
process. been widely used in different areas, such as dc-dc converter
Output LC Filter 2) Offline design: Here, a surrogate model of the optimal

VSI if L io
parameter adaptation method is created in an offline setting,
AC using data from the model of the environment and the con-
Vdc C
Load
verter system. This model can be used to determine the optimal
parameters for different operating conditions, which can then
S
be implemented in the controller. This approach does not
PI/PR+NN Measurements require real-time adaptation, but it does require a large amount
Online design of simulation data to be collected and processed in order
Offline design Output LC Filter to determine the optimal parameters. Moreover, if simulation
VSI if L io data is not sufficiently well representing the real system, this
approach may yield sub-optimal or even unstable results.
AC
Vdc C The online/offline ANN-based design diagrams in a linear
Load
controller are shown in Fig. 3. This figure uses a dc-ac voltage
source inverter (VSI) feeding an ac load via an output LC filter
Data Collection &
NN Offline Training Measurements but, the general application of ANN should be similar in other
S topologies, such as ac (micro)grid-connected using an LCL
filter. The biggest difference between online and offline design
PI/PR Controller with NN
Online Predictions
is that there is no intensive online training in offline design;
As a trade-off, independent offline learning usually requires a
Fig. 3: ANN applications in linear controllers. high number of offline simulations to collect the training data.
For dc-dc converters, [32] proposed using ANN to adjust both
PID coefficients and the controller structure. [3] proposed an
[32], [3], [33], [34], spacecraft [35], exoskeleton systems [36], ANN-based predictor to modify the output voltage reference
and electric vehicle charging system [37]. However, there are values in a PID controller, which can improve the transient
generally two groups of approaches utilized to do that, design response. In these studies, ANN was trained offline to pursue
and control, as discussed below. high prediction accuracy for the output voltage of the dc-dc
converter.
A. Parameter Adaptation
1) Online design: In this approach, an AI-based signal B. Control Surrogate Modelling
processing method is used to take measurements from the en- Different from the above parameter adaptation studies, in
vironment and learn the optimal parameters in real-time. This [33], an ANN was trained as an adaptive controller which
can be achieved by exploring an online AI-based surrogate directly outputs the duty cycle and the frequency of gate-
model of the linear system, which is trained using data from driving signals. This belongs to the other group of AI ap-
the system and the environment. AI techniques such as ANNs plications in the linear controllers, the controller surrogate
and heuristic algorithms can be used to build this model. By model. In this approach, a surrogate AI-based model of the
continuously searching through the model, the controller can system is designed that blends linear control and nonlinear
adapt the parameters for different operating conditions in real system dynamics, e.g. by matching the linear impedance model
time. However, it can be computationally intensive and require of the converter with the nonlinear impedance model of the
a high-frequency sampling rate. grid. This model can then be used in an online or offline
In [38], two types of ANNs were proposed to design a search for parameters of the linear controller, as indicated in
Hammerstein model identifier and a PID controller for the previous subsections. The AI-based model can be created with
adaptive control of a converter-based Hybrid Energy Storage techniques such as physics-informed neural networks, which
System (HESS); The online learning of Hammerstein ANN can incorporate the physical laws of the system into the model.
can supply an appropriate power flow reference for the HESS
to improve unsatisfactory frequency deviations and tie power
IV. AI A PPLICATIONS IN N ON - LINEAR C ONTROLLERS OF
oscillation. In [39], the student psychology optimization al-
P OWER C ONVERTERS
gorithm is used for the online tuning of control parameters
within a PI-incorporated reinforced learning ANN controller, If only simple linear controllers are employed, system
for power system applications. dynamics and external disturbances may not be well addressed
Apart from ANN applications, in [40], local model networks though they are computationally light without much complex-
(also recognized as Neuro-fuzzy systems) were used to iden- ity. In contrast, some advanced control methods can integrate
tify the model of the dc-dc buck converter, which can model nonlinearities and consider system constraints into the model,
the nonlinear dynamics for the implementation of a local linear for example, model predictive control (MPC). One of the key
controller regulating the converter output voltage. To improve characteristics of MPC is that the control objectives and system
the transient response performance of dc-dc converters, [41] limitations can be simply and intuitively included in the cost
combined two fuzzy-logic controllers with a PID compensator function (CF), directing the easy generalization for different
for the coefficients and error modifications. converter topologies.
Two main categories in MPC include the continuous control Output LC Filter
set MPC (CCS-MPC) and the finite control set MPC (FCS- VSI if L io
MPC). CCS-MPC uses the control vector as a continuous
control signal thus the output of optimization can be any vector AC
Vdc
Vdc C
within the control region defined by available voltage vectors Load
of the converter [42], [43]. Differently, FCS-MPC considers a
set with a limited number of input candidates thus the output
of optimization can only be one of the considered vectors in S
MPC Measurements
the set [44]. AI techniques have been used for both the design
and imitation control of MPC. The following two sections will
look at these two aspects in detail. Data Collection &
NN Offline Training
A. Parameter Design for Non-linear Controllers (a) Data collection and ANN training offline
Weighting factor design Output
continues to be a hot topic in MPC
LC Filter Output LC Filter
VSIusually VSI
because there are if multiple L control terms
io in the used if L io
CF. [45] proposed an ANN approach to automatically select
the weighting factors in the CF ofC FCS-MPC. The AC trained AC
Vdc Vdc C
Load Load
ANN serves as a surrogate model of the converter that can
provide fast and accurate estimates of the performance metrics
for any weighting factor combination. [46] presented an ANN-
S Measurements
based real-time tuning methodMPCof a weighting factor to achieve
Measurements
the desired average switching frequency and track the current S
reference.
Reference [47] proposed Dataan inverse& application method of
Collection x o
…
…
NN Offline
AI that can effectively provide Training and coefficients for
references
…
the control of a power converter-based system. Two different
cases were used for the method validation. One is the current Trained NN Online Test
sharing for a converter-based microgrid; The other is the (b) Trained ANN online test
extension of the MMC operation region under unbalanced grid
faults. An ANN-based droop coefficient design method was Fig. 4: Diagram of MPC imitation learning
proposed in [48] for improved load sharing.
Some published works studying ANN-based predictive con-

B. Imitation Controller for Nonlinear Controllers trollers are discussed below. A deep neural network-based
The high computational burden is one of the main disad- predictive control strategy is generally presented in [49] for
vantages of MPC, especially for the implementation of multi- the application of high-frequency multi-level converters. How-
step predictions or/and multilevel converters. To address that, ever, none of the technical details of machine learning work
ANNs were used to learn the MPC model via offline training, was provided, for example, data collection, ANN training
which can keep an approximate control performance while & validation, etc. In contrast, [50] clearly presents an ANN
at the same time reducing the computational burden. The learning approach for the control of a two-level voltage source
theoretical basis of this imitation approach is that the predictive inverter feeding linear and non-linear loads. System descrip-
control process is completely deterministic, i.e. for the same tions, FCS-MPC principles, proposed ANN architecture, and
set of input variables (i.e. circuit measurements) and a given training/test/validation are all presented in detail. Moreover,
cost function, the outputs (inserted vectors/submodules) will the authors collected sample data from 60 circuit conditions
always be the same. In this context, while the conventional (covering the linear/non-linear load) and tested the trained
MPC uses exhaustive rolling optimization every time instant ANN for all the considered circuit conditions. Even though the
to identify the optimal actuation, this is not necessary. It proposed method in [50] was not validated in the experiment,
should be possible to represent the deterministic input-output it is presented with all necessary technical details thus easy to
relationship with a more computationally efficient structure. implement and generalize. Following the same approach, [51]
Therefore, the same control effect as the MPC can be achieved trained different ANN controllers for up to N=3 prediction
by an ANN, but with a lower online computational burden. horizons and tested them in the experiment rig.
The general way of MPC imitation is depicted in Fig. 4 Regarding the converter topology, [50], [51] used this su-
for an inverter system. There are three steps in this process: pervised learning method for two-level converters, and [52]
(1). Data collection from the original MPC model; (2). ANN further implement it in the three-level converter. [49], [53]
offline training; (3). Online test of the trained ANN. The first trained and validated the ANN imitation controller to a five-
two steps are shown in Fig. 4(a) while the last step is shown level flying capacitor converter. And [54], [55] proposed two
in Fig. 4(b). different ANN imitation controllers for the modular multi-level
converter (MMC). risk of running out of the training range. Thus, the response
of AI controller is not fully predictable. In [55], a case is
C. Online ANN for Nonlinear Controllers given with some results when the controller runs out of the
training range, the MMC is still able to control the system;
Apart from the above design and control studies using But, obviously, it is not for all the related control problems.
ANN offline training for MPC, there was plenty of research 2) Controller Design using AI: As mentioned in Section
on neural network-based nonlinear online-learning controllers, IV, apart from replacing the PI/PR controllers with AI, AI
which do not comprise data collection for the learning but technologies can be used to help better design the traditional
use an optimization method, such as Levenberg-Marquard converters, at the same time, keeping the existing controllers in
(LM) [56] and Genetic Algorithm (GA) [57]. For example, the converters. For example, in [61], the artificial bee colony
Reference [58] proposed an adaptive self-organizing fuzzy algorithm is used to design the PID parameters of the PV
wavelet neural network for an AC motor servo system, which converters. In [62] presents a recursive least square and genetic
uses adaptive GA to optimize ANN parameters. In [5], [59], algorithm-based tool for PID controller tuning, with this
recurrent fuzzy ANN-aided (for feature selection) sliding- method, the controller design process can be accelerated. In
mode control was used in the power converter-based systems [63], a particle swarm optimization (PSO) controller parameter
for intelligent control with an active power filter, where the design method is proposed for the DC-DC buck converter to
online learning laws can effectively capture unknown functions reduce the aforementioned transient and steady-state errors.
with a low computational burden. For ML applications, [45] proposed an ANN-based method
to design the weighting factors of the MPC-based grid-
V. A PPLICATION A REAS AND F UTURE P ERSPECTIVES connected controller. In [47], a special application of ANN is
After discussing the AI techniques for linear and nonlinear introduced to select the droop coefficients of microgrids; That
controllers, in this section, different AI applications are brieftly is, instead of inputting system parameters for system response,
introduced. Fig. 5 shows the general structure of this section. the desired system response is set as the input of ANN, then
Solar, wind, HVDC, energy storage, powertrain, and promising the trained ANN can tell the user what is the preferable input
Power to X (P2X) applications are presented to describe the parameters for the target system. The gain of ML approach is
differences between each other. that, it can get rid of optimization algorithms, no local optimal
Following the above sections, the common technologies will problem. This is because, after training, computation of ANN
be first introduced in this section, those methods are general is extremely light thus exhaustive algorithm can be used in
methods and thus available for most applications/situations. the design space. The optimal design can be quickly found
Furthermore, application-specific AI technologies are given through tightly sweeping the design space.
one by one, and those technologies are designed for specific 3) Converter Modelling using AI: Regarding the modeling
applications to achieve different goals. of the power electronics converters, traditional analytical meth-
ods can be used to model the converter, however, it heavily
relies on the accuracy of the system parameters. By using AI
A. General AI Technologies technologies, the modeling of the converter can be obtained
Four aspects will be summarized below for the general AI by measuring the converters. In [64], Bayesian Regularization
technologies; The first two (AI-based controller and controller (BR) along with ANN and random forest (RF) based machine
design using AI) have been elaborated in Section IV; The third learning methods are used to model the power electronics
(Converter modeling using AI) and fourth (Stability Analysis converter with collected experimental data.
by AI) are also important and practical topics regarding AI- In this case, most of the industrial applications will still
enabled controller. The pros and cons of each aspect will be treat AI converter model as a trained or simulated ’black
presented to help the potential scholars and engineers select box’. That means, this method can only be used to verify
suitable methods. the system modeling after the physical system has been built
1) AI Based Controller: The grid-connected converters play up. Therefore, even though this method is useful for modeling
important roles to regulate the power to the grid. Section IV verification, it may not be applicable in the early design phase
introduced the ANN application in replacing the traditional of the system.
PI-based controllers to reduce the computation burden. For 4) AI Based Stability Analysis: In the modern power
the photovoltaics (PV) application, [60] gave an example electronics-based power system, the stability of the system
that used an ANN-based controller to control a single-phase is an important factor due to the high share of inverter-
PV inverter. The results show that the output power can be based resources. For example, the phase lock loop of the
regulated by the proposed ANN-based controller. Similar AI- grid following the converter will cause system instability
based imitation methods could be generalized to different under weak grid conditions. AI technologies can contribute
systems and applications. to stabilizing power electronics-based power systems. In [65],
The advantage of the AI-based imitation controller is that, a random forests-based power oscillation damper (iPOD) for
the transient response of the AI controller is faster than PI grid forming converter is proposed, the proposed damper can
controller because there is no integral part in the controller automatically adjust the gain of the active damping controller
to slow down the dynamics behavior. However, because the based on the different operating points. In [66], the proposed
controller is trained by collected data only, there may be a AI-based control design uses a deep neural network (DNN) to
Fig. 5: AI Applications
learn the nonlinear mapping between the virtual synchronous a wind farm is proposed to better control the frequency of
generator (VSG) input and output signals, enabling it to adapt the wind farm network. The network frequency variation is
to different operating conditions and disturbances. The paper minimized by the proposed controller. In [73], an IGBT Fault
presents simulation results to demonstrate the effectiveness Tolerant Control (FTC) strategy for wind turbine converter
of the proposed control design in improving the stability is proposed, and fuzzy logic is used to fast detect the wind
and robustness of VSGs under various operating scenarios. turbine converter IGBT fault, then the protection algorithm
[67] gives a comprehensive review of the AI applications to will activate the redundant converter leg to bypass the faulty
improve the stability of future power systems. leg.
For stability analysis, the key interest from the industry is 3) High Voltage Direct Current Application: The power
the physical principles and insights of stability issue in the needs to be transferred to the power grid using HVAC (High
system. With traditional transfer function or state-space-based Voltage Alternate Current)/HVDC (High Voltage Direct Cur-
stability analysis, there is a solid theory behind it to explain rent) technologies to reduce transmission losses from the
the stability. However, due to lack of interpretation, industries far away generation units. For HVAC technology, traditional
are still hesitant to embrace the AI-based stability analysis, power transformers are used which will not require many
though it is easy to use with promising results. power electronics controllers in the plant, then AI application
in the HVDC system is not within the scope of this paper.
For HVDC technology, 2 main power electronics converters
B. Technologies for Specific Application
are used: line-commutated HVDC converters, and voltage-
AI techniques can be applied in many specific areas, but the source HVDC converters. AI technologies can be used to
function and purpose may vary according to different require- better control the HVDC converters to achieve internal and
ments. For example, maximum power point tracking (MPPT) external performances. In [74], an ANN-based operational
should be considered for solar applications, however, AI-based region extension method for Modular Multilevel Converters
MPPT is not very common in powertrain applications. Some (MMCs) under unbalanced grid faults is proposed, the ANN
specific AI application areas are listed as follows: can automatically decide the injected circulating current to re-
1) Solar: For solar applications, the special application of duce the submodule voltage ripple under grid faults, hence, the
the AI technology is for the MPPT purpose, MPPT is widely submodule over voltage trip is avoided and then the operating
used in DC-DC converters in solar applications to track the region is extended. In [54] and [55], machine learning-based
maximum power from the sun. Traditionally, 2 main non-AI controllers are proposed for MMCs to achieve faster dynamic
MPPT algorithms are commonly used: Perturb and Observe response and lower the computation burden of the controller.
(P&O) algorithm [68] and the Incremental Conductance (INC) 4) Energy Storage System: Energy storage is an important
algorithm [69]. However, the traditional MPPT algorithms role in the future renewable-energy-based power system to
cannot properly track the maximum power point under partial shift the mismatch between generation and consumption. The
shading conditions. The AI-added MPPT methods can track battery is one of the most important ways to do energy storage.
the global maximum power point with higher computational Besides energy storage, the battery-connected converter can
speed and faster dynamic speed [70]. [71] compared different also provide the grid ancillary services, for instance, reduc-
AI-aided MPPT algorithms both under partial shading condi- ing the system harmonics, regulating system frequency, and
tions, the conclusion from the paper is that RBFC optimized supporting system voltage.
Fuzzy controller has the highest efficiency under partial shad- [75] proposed the design of a Multi-Objective-Based Artifi-
ing conditions. cial intelligence controller for a wind/battery-connected shunt
2) Wind: In wind energy systems, one challenge is the active power filter (SAPF). The controller aims to improve
changing of the wind speed over time might cause power the performance of the SAPF by simultaneously optimizing
fluctuation in the wind energy system, then the frequency of multiple objectives such as Total Harmonic Distortion (THD),
the wind network will also be fluctuating. In [72], an adaptive voltage regulation, and energy efficiency. [76] gave a com-
ANN controller for the energy capacitor system (ECS) in prehensive review of the AI applications of energy storage
systems. R EFERENCES
5) Powertrain for Transportation: Electric Drive Control: [1] S. Zhao, F. Blaabjerg, and H. Wang, “An overview of artificial intelli-
Several studies have been conducted for electric drive ap- gence applications for power electronics,” IEEE Transactions on Power
plications by using ML-aided controllers. For example, [77] Electronics, vol. 36, no. 4, pp. 4633–4658, 2020.
[2] R. Rajamony, S. Wang, G. Calderon-Lopez, I. Ludtke, and W. Ming,
used a multilayer perception feedforward ANN to learn the “Artificial neural networks-based multi-objective design methodology
long-horizon FCS-MPC, which solves the mixed-integer op- for wide-bandgap power electronics converters,” IEEE Open Journal
timal control problem offline to generate the training data of Power Electronics, vol. 3, pp. 599–610, 2022.
[3] H. Maruta, H. Taniguchi, and F. Kurokawa, “A study on effects of
of ANN. After training, a 2-level inverter-based motor drive different control period of neural network based reference modified
was controlled by the trained ANNs in real-time systems, and pid control for dc-dc converters,” in 2016 15th IEEE International
different horizon numbers were compared. Deep RL was used Conference on Machine Learning and Applications (ICMLA), pp. 460–
465, IEEE, 2016.
in [78] for a DC-DC Buck converter-fed DC motor to reduce [4] P. Pandey, S. Bodda, and P. Agnihotri, “Comparative analysis of the
the torque/current ripples. And [79] proposed ANN model control of llc resonant converter using proportional-integral and machine
predictive and variable structure controllers for the single- learning based techniques,” in 2020 IEEE 17th India Council Interna-
tional Conference (INDICON), pp. 1–6, IEEE, 2020.
degree-of-freedom rotary manipulator. [5] S. Hou, Y. Chu, and J. Fei, “Intelligent global sliding mode control using
6) Future Perspectives: Power To Hydrogen: With the recurrent feature selection neural network for active power filter,” IEEE
Transactions on Industrial Electronics, vol. 68, no. 8, pp. 7320–7329,
large-scale development of renewable energies, the constraint 2020.
of the power grid limits the maximum power that can be [6] I. Martin-Diaz, D. Morinigo-Sotelo, O. Duque-Perez, and R. J. Romero-
transferred to the grid, and also the intermittent nature of Troncoso, “An experimental comparative evaluation of machine learning
techniques for motor fault diagnosis under various operating conditions,”
renewable energies requires a long-term energy storage so- IEEE Transactions on Industry Applications, vol. 54, no. 3, pp. 2215–
lution for renewable energies. Producing hydrogen by using 2224, 2018.
energy can be an ideal solution for long-term energy storage. [7] Y. León-Ruiz, M. González-Garcı́a, R. Alvarez-Salas, J. Cuevas-Tello,
and V. Cárdenas, “Fault diagnosis based on machine learning for the high
AI has good potential to better control and design future green frequency link of a grid-tied photovoltaic converter for a wide range of
hydrogen systems. The power converters in green hydrogen irradiance conditions,” IEEE Access, vol. 9, pp. 151209–151220, 2021.
systems play important roles in safe, high-efficient, and low- [8] Z. Li, Y. Gao, X. Zhang, B. Wang, and H. Ma, “A model-data-hybrid-
driven diagnosis method for open-switch faults in power converters,”
cost green hydrogen production. This area is still relatively IEEE Transactions on Power Electronics, vol. 36, no. 5, pp. 4965–4970,
new and many questions need to be solved in the near future. 2020.
Nowadays, the power industry is still mainly using con- [9] S. Chen, “Review on supervised and unsupervised learning techniques
for electrical power systems: Algorithms and applications,” IEEJ trans-
ventional analytical methods to design, model, and control actions on electrical and electronic engineering, vol. 16, no. 11,
the power electronics system, because of the well-developed pp. 1487–1499, 2021.
theory and interpretation of the conventional methods. An [10] E. Mohammadi, M. Alizadeh, M. Asgarimoghaddam, X. Wang, and
M. G. Simões, “A review on application of artificial intelligence tech-
example is the impedance-based method to judge the grid- niques in microgrids,” IEEE Journal of Emerging and Selected Topics
connected converter stability. In the near future, the power in Industrial Electronics, 2022.
industry will not fully trust AI techniques, nor quickly turn [11] E. O. Arwa and K. A. Folly, “Reinforcement learning techniques for
optimal power control in grid-connected microgrids: A comprehensive
to AI-based methods for power converter control. However, review,” IEEE Access, vol. 8, pp. 208992–209007, 2020.
AI-based methods still have great potential for industrial [12] D. Cao, W. Hu, J. Zhao, G. Zhang, B. Zhang, Z. Liu, Z. Chen, and
applications, especially when the system is too complicated to F. Blaabjerg, “Reinforcement learning and its applications in modern
power and energy systems: A review,” Journal of modern power systems
be analytically explained. The data-driven models could help and clean energy, vol. 8, no. 6, pp. 1029–1042, 2020.
electrical engineers to gain the system insights without doing [13] Y. Gao, T. Yang, S. Bozhko, P. Wheeler, and T. Dragičević, “Filter design
tedious analysis or modeling, this is indeed the key advantage and optimization of electromechanical actuation systems using search
and surrogate algorithms for more-electric aircraft applications,” IEEE
of AI techniques. Transactions on Transportation Electrification, vol. 6, no. 4, pp. 1434–
1447, 2020.
[14] T. Guillod, P. Papamanolis, and J. W. Kolar, “Artificial neural network
VI. C ONCLUSION (ann) based fast and accurate inductor modeling and design,” IEEE Open
Journal of Power Electronics, vol. 1, pp. 284–299, 2020.
[15] S. Weisberg, Applied linear regression, vol. 528. John Wiley & Sons,
This paper gave an overview of AI-enabled controllers 2005.
for power converter-based systems. To begin with, the fast- [16] J. A. Suykens and J. Vandewalle, “Least squares support vector machine
developing ML algorithms were introduced with three basic classifiers,” Neural processing letters, vol. 9, no. 3, pp. 293–300, 1999.
[17] M. Hannan, J. Abd Ali, A. Mohamed, and M. N. Uddin, “A random
groups: supervised, unsupervised, and reinforcement learning. forest regression based space vector pwm inverter controller for the
Then, the AI applications in linear and nonlinear controllers induction motor drive,” IEEE Transactions on Industrial Electronics,
were discussed separately from two perspectives: design and vol. 64, no. 4, pp. 2689–2699, 2016.
[18] L. P. Kaelbling, M. L. Littman, and A. W. Moore, “Reinforcement
control; In this part, online/offline application modes and their learning: A survey,” Journal of artificial intelligence research, vol. 4,
technical details were analyzed and discussed. In addition, the pp. 237–285, 1996.
common AI technologies were summarized with four aspects [19] M. Gheisarnejad, H. Farsizadeh, and M. H. Khooban, “A novel nonlinear
deep reinforcement learning controller for dc–dc power buck converters,”
that can be generally used for various applications; And some IEEE Transactions on Industrial Electronics, vol. 68, no. 8, pp. 6849–
specific AI application areas were further given which may 6858, 2020.
focus on typical requirements or targets. Finally, a future [20] A. J. Abianeh, Y. Wan, F. Ferdowsi, N. Mijatovic, and T. Dragičević,
“Vulnerability identification and remediation of fdi attacks in islanded dc
perspective on Power-To-Hydrogen was added as a promising microgrids using multiagent reinforcement learning,” IEEE Transactions
way for future green energy applications. on Power Electronics, vol. 37, no. 6, pp. 6359–6370, 2021.
[21] J. Konečnỳ, H. B. McMahan, F. X. Yu, P. Richtárik, A. T. Suresh, and dc switching converters,” Journal of Electrical Engineering & Technol-
D. Bacon, “Federated learning: Strategies for improving communication ogy, pp. 1–12, 2023.
efficiency,” arXiv preprint arXiv:1610.05492, 2016. [42] S. Vazquez, C. Montero, C. Bordons, and L. G. Franquelo, “Model
[22] P. Kairouz, H. B. McMahan, B. Avent, A. Bellet, M. Bennis, A. N. predictive control of a vsi with long prediction horizon,” in 2011
Bhagoji, K. Bonawitz, Z. Charles, G. Cormode, R. Cummings, et al., IEEE International Symposium on Industrial Electronics, pp. 1805–
“Advances and open problems in federated learning,” Foundations and 1810, IEEE, 2011.
Trends® in Machine Learning, vol. 14, no. 1–2, pp. 1–210, 2021. [43] T. Dragičević, S. Vazquez, and P. Wheeler, “Advanced control methods
[23] X. Cheng, C. Li, and X. Liu, “A review of federated learning in energy for power converters in dg systems and microgrids,” IEEE Transactions
systems,” 2022 IEEE/IAS Industrial and Commercial Power System Asia on Industrial Electronics, vol. 68, no. 7, pp. 5847–5862, 2020.
(I&CPS Asia), pp. 2089–2095, 2022. [44] J. Rodriguez, M. P. Kazmierkowski, J. R. Espinoza, P. Zanchetta,
[24] A. M. Elbir, B. Soner, S. Çöleri, D. Gündüz, and M. Bennis, “Federated H. Abu-Rub, H. A. Young, and C. A. Rojas, “State of the art of
learning in vehicular networks,” in 2022 IEEE International Mediter- finite control set model predictive control in power electronics,” IEEE
ranean Conference on Communications and Networking (MeditCom), Transactions on Industrial Informatics, vol. 9, no. 2, pp. 1003–1016,
pp. 72–77, IEEE, 2022. 2012.
[25] D. Liu and O. Simeone, “Privacy for free: Wireless federated learning [45] T. Dragičević and M. Novak, “Weighting factor design in model predic-
via uncoded transmission with adaptive power control,” IEEE Journal tive control of power electronic converters: An artificial neural network
on Selected Areas in Communications, vol. 39, no. 1, pp. 170–185, 2020. approach,” IEEE Transactions on Industrial Electronics, vol. 66, no. 11,
[26] S. J. Pan and Q. Yang, “A survey on transfer learning,” IEEE Transac- pp. 8870–8880, 2018.
tions on Knowledge and Data Engineering, vol. 22, no. 10, pp. 1345– [46] S. Vazquez, D. Marino, E. Zafra, M. D. V. Peña, J. J. Rodrı́guez-Andina,
1359, 2010. L. G. Franquelo, and M. Manic, “An artificial intelligence approach for
[27] P. Qin and L. Zhao, “A novel transfer learning-based cell soc online real-time tuning of weighting factors in fcs-mpc for power converters,”
estimation method for a battery pack in complex application conditions,” IEEE Transactions on Industrial Electronics, vol. 69, no. 12, pp. 11987–
IEEE Transactions on Industrial Electronics, 2023. 11998, 2021.
[28] S. Zhang, H. Zhu, J. Wu, and Z. Chen, “Voltage relaxation-based state- [47] Y. Gao, S. Wang, H. Hussaini, T. Yang, T. Dragičević, S. Bozhko,
of-health estimation of lithium-ion batteries using convolutional neural P. Wheeler, and S. Vazquez, “Inverse application of artificial intelligence
networks and transfer learning,” Journal of Energy Storage, vol. 73, for the control of power converters,” IEEE Transactions on Power
p. 108579, 2023. Electronics, vol. 38, no. 2, pp. 1535–1548, 2022.
[29] M. Zhang, X. Wang, D. Yang, Z. Cui, and M. G. Christensen, “Transfer [48] H. Hussaini, T. Yang, Y. Gao, C. Wang, T. Dragicevic, and S. Bozhko,
learning for identifying impedance estimation in voltage source invert- “Droop coefficient design in droop control of power converters for
ers,” in 2020 IEEE Energy Conversion Congress and Exposition (ECCE), improved load sharing: An artificial neural network approach,” in 2021
pp. 6170–6174, IEEE, 2020. IEEE 30th International Symposium on Industrial Electronics (ISIE),
[30] M. Zhang, Y. Zhang, and Q. Xu, “Transfer learning based online pp. 1–6, IEEE, 2021.
impedance identification for modular multilevel converters,” IEEE [49] D. Wang, X. Yin, S. Tang, C. Zhang, Z. J. Shen, J. Wang, and
Transactions on Power Electronics, vol. 38, no. 10, pp. 12207–12218, Z. Shuai, “A deep neural network based predictive control strategy for
2023. high frequency multilevel converters,” in 2018 IEEE Energy Conversion
[31] Y. Wang and T. Wu, “Temperature rising model of tubular permanent Congress and Exposition (ECCE), pp. 2988–2992, IEEE, 2018.
magnet linear synchronous motor based on transfer learning-deep neural [50] I. S. Mohamed, S. Rovetta, T. D. Do, T. Dragicević, and A. A. Z. Diab,
network,” in 2023 42nd Chinese Control Conference (CCC), pp. 6707– “A neural-network-based model predictive control of three-phase inverter
6712, 2023. with an output lc filter,” IEEE Access, vol. 7, pp. 124737–124749, 2019.
[32] J. Liu, T. Wei, N. Chen, W. Liu, J. Wu, and P. Xiao, “A neural network [51] M. Novak and T. Dragicevic, “Supervised imitation learning of finite-
pid controller with dynamic structure adjustment for dc-dc switching set model predictive control systems for power electronics,” IEEE
converters,” in 2022 7th International Conference on Integrated Circuits Transactions on Industrial Electronics, vol. 68, no. 2, pp. 1717–1723,
and Microsystems (ICICM), pp. 356–360, IEEE, 2022. 2020.
[33] M. Jafari and Z. Malekjamshidi, “Design, simulation and implementa- [52] M. Novak and F. Blaabjerg, “Supervised imitation learning of fs-mpc
tion of an adaptive controller on base of artificial neural networks for a algorithm for multilevel converters,” in 2021 23rd European Conference
resonant dc-dc converter,” in 2011 IEEE Ninth International Conference on Power Electronics and Applications (EPE’21 ECCE Europe), pp. P–
on Power Electronics and Drive Systems, pp. 1043–1046, IEEE, 2011. 1, IEEE, 2021.
[34] H. Maruta and D. Hoshino, “Transient response improvement of [53] D. Wang, Z. J. Shen, X. Yin, S. Tang, X. Liu, C. Zhang, J. Wang,
repetitive-trained neural network controlled dc-dc converter with over- J. Rodriguez, and M. Norambuena, “Model predictive control using
compensation suppression,” in IECON 2019 - 45th Annual Conference artificial neural network for power converters,” IEEE Transactions on
of the IEEE Industrial Electronics Society, vol. 1, pp. 2088–2093, 2019. Industrial Electronics, vol. 69, no. 4, pp. 3689–3699, 2021.
[35] R. Wang, Z. Zhou, and G. Qu, “Fuzzy neural network pid control based [54] S. Wang, T. Dragicevic, Y. Gao, and R. Teodorescu, “Neural network
on rbf neural network for variable configuration spacecraft,” in 2018 based model predictive controllers for modular multilevel converters,”
IEEE 3rd Advanced Information Technology, Electronic and Automation IEEE Transactions on Energy Conversion, vol. 36, no. 2, pp. 1562–1571,
Control Conference (IAEAC), pp. 1203–1207, IEEE, 2018. 2020.
[36] M. P. Belov, D. D. Truong, and P. Van Tuan, “Self-tuning pid controller [55] S. Wang, T. Dragicevic, G. F. Gontijo, S. K. Chaudhary, and R. Teodor-
using a neural network for nonlinear exoskeleton system,” in 2021 II escu, “Machine learning emulation of model predictive control for mod-
International Conference on Neural Networks and Neurotechnologies ular multilevel converters,” IEEE Transactions on Industrial Electronics,
(NeuroNT), pp. 6–9, IEEE, 2021. vol. 68, no. 11, pp. 11628–11634, 2020.
[37] Z. Li, J. Pang, C. Xu, and C. Zhang, “Electric vehicle charging system [56] W. Zhao, K. Li, and G. W. Irwin, “A new gradient descent approach
pressure control based on fuzzy neural network pid control,” in 2021 for local learning of fuzzy neural models,” IEEE Transactions on Fuzzy
IEEE Conference on Telecommunications, Optics and Computer Science Systems, vol. 21, no. 1, pp. 30–44, 2012.
(TOCS), pp. 873–877, IEEE, 2021. [57] M. R. Mashinchi and A. Selamat, “An improvement on genetic-based
[38] D. Xu, J. Liu, X.-G. Yan, and W. Yan, “A novel adaptive neural network learning method for fuzzy artificial neural networks,” Applied Soft
constrained control for a multi-area interconnected power system with Computing, vol. 9, no. 4, pp. 1208–1216, 2009.
hybrid energy storage,” IEEE Transactions on Industrial Electronics, [58] R. Hou, D. Shi, Q. Gao, and Y. Hou, “Implementation of adaptive self-
vol. 65, no. 8, pp. 6625–6634, 2017. organizing fuzzy wavelet neural network: Modeling ac servo system,” in
[39] M. K. Das, P. Bera, P. Sarkar, and K. Chakrabarty, “Pi-rlnn controller 2021 IEEE 10th Data Driven Control and Learning Systems Conference
for lfc of hybrid deregulated power system based on spoa,” in 2021 (DDCLS), pp. 1128–1134, IEEE, 2021.
IEEE 18th India Council International Conference (INDICON), pp. 1– [59] S. Hou, Y. Chu, and J. Fei, “Robust intelligent control for a class
5, IEEE, 2021. of power-electronic converters using neuro-fuzzy learning mechanism,”
[40] K. Rouzbehi, A. Miranian, E. Rakhshani, A. Luna, and P. Rodriguez, IEEE Transactions on Power Electronics, vol. 36, no. 8, pp. 9441–9452,
“Identification and local linear control of a dc-dc buck converter using 2021.
local model networks,” in 2013 International Conference on Renewable [60] M. Demirtaş, I. Cetinbas, S. Serefoğlu, and O. Kaplan, “Ann con-
Energy Research and Applications (ICRERA), pp. 745–750, 2013. trolled single phase inverter for solar energy systems,” in 2014 16th
[41] J. Liu, T. Wei, N. Chen, J. Wu, and P. Xiao, “Fuzzy logic pid controller International Power Electronics and Motion Control Conference and
with both coefficient and error modifications for digitally-controlled dc– Exposition, pp. 768–772, IEEE, 2014.
[61] A. Oshaba, E. S. Ali, and S. M. Abd Elazim, “Pi controller design using
artificial bee colony algorithm for mppt of photovoltaic system supplied
dc motor-pump load,” Complexity, vol. 21, no. 6, pp. 99–111, 2016.
[62] L. Galotto, J. Pinto, J. Bottura Filho, and G. Lambert-Torres, “Recursive
least square and genetic algorithm based tool for pid controllers tuning,”
in 2007 International Conference on Intelligent Systems Applications to
Power Systems, pp. 1–6, IEEE, 2007.
[63] A. Debnath, T. O. Olowu, S. Roy, I. Parvez, and A. Sarwat, “Particle
swarm optimization-based pid controller design for dc-dc buck con-
verter,” in 2021 North American Power Symposium (NAPS), pp. 1–6,
IEEE, 2021.
[64] H. S. Krishnamoorthy and T. N. Aayer, “Machine learning based mod-
eling of power electronic converters,” in 2019 IEEE Energy Conversion
Congress and Exposition (ECCE), pp. 666–672, IEEE, 2019.
[65] G. N. Baltas, N. B. Lai, L. Marin, A. Tarraso, and P. Rodriguez, “Grid-
forming power converters tuned through artificial intelligence to damp
subsynchronous interactions in electrical grids,” IEEE access, vol. 8,
pp. 93369–93379, 2020.
[66] Q. Xu, T. Dragicevic, L. Xie, and F. Blaabjerg, “Artificial intelligence-
based control design for reliable virtual synchronous generators,” IEEE
Transactions on Power Electronics, vol. 36, no. 8, pp. 9453–9464, 2021.
[67] W. Liu, T. Kerekes, T. Dragicevic, and R. Teodorescu, “Review of
grid stability assessment based on ai and a new concept of converter-
dominated power system state of stability assessment,” IEEE Journal of
Emerging and Selected Topics in Industrial Electronics, 2023.
[68] J. Ahmed and Z. Salam, “An improved perturb and observe (p&o)
maximum power point tracking (mppt) algorithm for higher efficiency,”
Applied Energy, vol. 150, pp. 97–108, 2015.
[69] D. Sera, L. Mathe, T. Kerekes, S. V. Spataru, and R. Teodorescu, “On the
perturb-and-observe and incremental conductance mppt methods for pv
systems,” IEEE journal of photovoltaics, vol. 3, no. 3, pp. 1070–1078,
2013.
[70] K. Y. Yap, C. R. Sarimuthu, and J. M.-Y. Lim, “Artificial intelligence
based mppt techniques for solar power system: A review,” Journal of
Modern Power Systems and Clean Energy, vol. 8, no. 6, pp. 1043–1059,
2020.
[71] S. R. Kiran, C. H. Basha, V. P. Singh, C. Dhanamjayulu, B. R. Prusty,
and B. Khan, “Reduced simulative performance analysis of variable step
size ann based mppt techniques for partially shaded solar pv systems,”
IEEE Access, vol. 10, pp. 48875–48889, 2022.
[72] S. Muyeen, H. M. Hasanien, and J. Tamura, “Reduction of frequency
fluctuation for wind farm connected power systems by an adaptive artifi-
cial neural network controlled energy capacitor system,” IET Renewable
Power Generation, vol. 6, no. 4, pp. 226–235, 2012.
[73] A. Bouzekri, T. Allaoui, M. Denai, and Y. Mihoub, “Artificial
intelligence-based fault tolerant control strategy in wind turbine sys-
tems,” International Journal of Renewable Energy Research, 2017.
[74] S. Wang, T. Dragicevic, Y. Gao, S. K. Chaudhary, and R. Teodorescu,
“Machine learning based operating region extension of modular mul-
tilevel converters under unbalanced grid faults,” IEEE Transactions on
Industrial Electronics, vol. 68, no. 5, pp. 4554–4560, 2020.
[75] S. Koganti, K. J. Koganti, and S. R. Salkuti, “Design of multi-objective-
based artificial intelligence controller for wind/battery-connected shunt
active power filter,” Algorithms, vol. 15, no. 8, p. 256, 2022.
[76] A. N. Abdalla, M. S. Nazir, H. Tao, S. Cao, R. Ji, M. Jiang, and L. Yao,
“Integration of energy storage system and renewable energy sources
based on artificial intelligence: An overview,” Journal of Energy Storage,
vol. 40, p. 102811, 2021.
[77] I. Hammoud, S. Hentzelt, T. Oehlschlaegel, and R. Kennel, “Long-
horizon direct model predictive control based on neural networks for
electrical drives,” in IECON 2020 The 46th Annual Conference of the
IEEE Industrial Electronics Society, pp. 3057–3064, IEEE, 2020.
[78] R. Anugula and S. P. K. Karri, “Deep reinforcement learning based adap-
tive controller of dc electric drive for reduced torque and current ripples,”
in 2021 IEEE International Conference on Technology, Research, and
Innovation for Betterment of Society (TRIBES), pp. 1–6, IEEE, 2021.
[79] N. Nikdel, P. Nikdel, M. A. Badamchizadeh, and I. Hassanzadeh, “Using
neural network model predictive control for controlling shape memory
alloy-based manipulator,” IEEE Transactions on Industrial Electronics,
vol. 61, no. 3, pp. 1394–1401, 2013.

Artificial Intelligence Techniques For Enhancing T

Uploaded by

Copyright:

Available Formats

Artificial Intelligence Techniques For Enhancing T

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Artificial Intelligence Techniques For Enhancing T

Uploaded by

Copyright:

Available Formats

This article has been accepted for publication in IEEE Open Journal of Industry Applications.

IEEE OPEN JOURNAL OF INDUSTRY APPLICATIONS 1

Artificial Intelligence Techniques for Enhancing the

IEEE OPEN JOURNAL OF INDUSTRY APPLICATIONS 2

spective. The paper generally considered most AI techniques:

neural networks and random forest, the application areas in Fault

ered. For microgrid applications, the survey of AI techniques Drive Parameter

IEEE OPEN JOURNAL OF INDUSTRY APPLICATIONS 3

IEEE OPEN JOURNAL OF INDUSTRY APPLICATIONS 4

Output LC Filter 2) Offline design: Here, a surrogate model of the optimal

IEEE OPEN JOURNAL OF INDUSTRY APPLICATIONS 5

Some published works studying ANN-based predictive con-

IEEE OPEN JOURNAL OF INDUSTRY APPLICATIONS 6

IEEE OPEN JOURNAL OF INDUSTRY APPLICATIONS 7

IEEE OPEN JOURNAL OF INDUSTRY APPLICATIONS 8

IEEE OPEN JOURNAL OF INDUSTRY APPLICATIONS 9

IEEE OPEN JOURNAL OF INDUSTRY APPLICATIONS 10

You might also like