Debanjali Sarkar, Taimoor Khan, Fazal A. Talukdar, Sembiam R. Rengarajan
©SHUTTERSTOCK.COM/AGSANDREW
Several practical engineering optimization problems are computationally demanding, requiring a large amount of computer time, processing power, and memory. These challenges can be mitigated by human-engineered systems exhibiting intelligent behavior. With the evolution of high-speed digital computers, the use of computational intelligence (CI) techniques has increased rapidly. According to Bezdek [1], “A system is called computationally intelligent if it deals with low-level data such as numerical data, has a pattern-recognition component and does not use knowledge in the artificial intelligence (AI) sense, and additionally when it begins to exhibit computational adaptivity, fault tolerance, speed approaching human-like turnaround and error rates that approximate human performance.” Another definition, by Engelbrecht [2], states that “CI is the study of adaptive mechanisms that enable or facilitate intelligent behavior in complex and changing environments. These mechanisms include those Artificial Intelligence paradigms that exhibit an ability to learn or adapt to new situations, to generalize, abstract, discover and associate.” Thus, CI is the general term used to classify all such nature-inspired methodologies and their associated theories and applications. The five important paradigms of the CI technique are artificial neural networks (ANNs), swarm intelligence (SI), evolutionary computation (EC), and fuzzy systems (FSs). The origin of each technique can be connected to a natural system; for example, an ANN imitates the biological neural system. SI models the behavior of organisms living in swarms, whereas EC models the natural evolution system. Similarly, an FS originates from human thinking processes. Many of the problems involved in designing next-generation systems can be resolved using these CI techniques or their combinations.
Among the different classes of CI techniques, ANNs, EC, and SI have received the most attention from researchers across different domains for solving complex design problems. An ANN is a data processing system that can be trained to learn any nonlinear input–output relationship. Speech processing, image processing, data mining, classification, pattern recognition, and forecasting are some of the real-world applications of ANNs. ANNs are trained by mapping the input onto the goal output with numerous neurons in several layers. Neurons are processing elements that connect two layers, and the strength of these connections is determined by weights. The weights are adjusted during the learning process. Training an ANN model with datasets to generate an approximate objective function is part of the analysis or synthesis process. The trained network can reliably and quickly estimate output parameters for given input parameters, skipping the repetitious analysis of traditional techniques. EC is designed to replicate characteristics of natural selection and survival of the fittest. In each generation, an initial population is formed and updated iteratively by discarding unsuitable chromosomes based upon their fitness values. Genetic algorithms (GAs), genetic programming, evolutionary methods, evolutionary programming, differential evolution, and cultural evolution are all examples of evolutionary algorithms. Among these different techniques, GAs are the most widely explored algorithmic model that employs evolution-based computation techniques. GAs rely on the use of selection, crossover, and mutation operators to create successive generations of better adapted individuals. The search is guided only by the fitness value associated with every individual in the population. SI approaches are used to model the social behavior of species living in swarms or colonies. SI systems are made up of homogeneous individuals that interact with one another and the environment on a local level. Because there is no centralized structure to manage individual behavior and movement, their interactions lead to global behavior. Particle swarm optimization (PSO), ant colony optimization, the cuckoo search algorithm, and the bat algorithm are some frequently applied SI algorithms. As all the CI techniques have their limitations, it is common practice to combine two or more paradigms to form hybrids of paradigms. The basic idea is to exploit the relative merits of individual techniques while eliminating their weaknesses. Among the different evolutionary and swarm algorithms, GAs and PSO have attracted the most attention from the research community [3], [4], [5], [6].
Recent years have witnessed a paradigm shift in wireless technologies from a source of communication to delivering energy. This development helps low-power devices to sense, link, and obtain power anywhere at any time. Traditional methods of charging the device are replaced by wireless charging, bringing a great deal of convenience to the user. The idea of energy transmission into free space and the conversion of the wireless energy into dc power has resulted in the development of two power supply techniques, RF energy harvesting (RFEH) and wireless power transfer (WPT) [7], [8]. Both RFEH and WPT systems have been investigated and designed over the years for different applications based on requirements. However, while designing them, it is essential to take into consideration the influence on the system performance of every constituent element. The antenna, impedance matching network, rectifying circuit, and power management unit generally feature on the receiving side of both RFEH and WPT systems. An antenna in combination with a rectifying circuit and matching network is commonly defined as a rectenna. The efficiency of RFEH and WPT systems is highly determined by the performance efficiency of rectennas. However, rectennas exhibit nonlinear behavior during the RF-to-dc conversion phase, which degrades their performance in practical scenarios. Therefore, it is essential to model this nonlinearity during simulation to predict the performance of the rectenna systems accurately. In recent years, CI techniques have been widely used for modeling and optimization in order to mitigate these design challenges of RFEH and WPT systems [9].
Although several review exercises have been reported in the literature that discuss the topologies and design challenges of RFEH and WPT systems, there are no such substantial contributions of review articles in the literature that investigate the available CI methods for modeling RFEH and WPT systems. This article presents a comprehensive review of the existing literature on CI methods for modeling RFEH and WPT systems. The review article will be beneficial for researchers working in the field as it bridges the gap between research and practice. The organization of the article is as follows: an overview of RFEH and WPT systems is discussed in the sections “Overview of RFEH Systems” and “Overview of WPT Systems,” respectively. The section “CI Modeling of RFEH Systems” covers CI modeling in RFEH systems, followed by a discussion of modeling in WPT systems in the section “CI Modeling of WPT Systems.” The section “Review Summary” provides a summary of the review.
The movement of technology toward low power has made it possible for some devices to run on energy harvested from the ambient environment. These energies are collected to convert into electricity and can be used on demand or can be stored for later use. Energy from the environment is available in the form of radiation, thermal, mechanical, nuclear, and magnetic sources. Energy harvesting is an alternative method for providing energy at low cost. Nowadays, many devices work wirelessly, but they must be charged repeatedly for proper functioning. To simplify this process, the same RF that is used to transfer data is used to charge the device. RFEH is a type of energy harvesting that can convert RF waves from the environment into electrical energy. It can provide energy to low-power-consuming devices. The major sources of RF are listed in Table 1 [10].
Table 1. Some RF sources.
The main components of a typical RFEH system, shown in Figure 1, are an antenna, a matching network, and a rectifier. An antenna helps in capturing the RF signal from the ambient environment and converts it into electricity. The frequency range that is chosen must find maximum usage at a particular location and time of use. When the antenna receives the signal, it creates a potential difference across the antenna. This causes a movement in the charge carriers, which then flows to the matching network. In a matching network, the impedance of an antenna adapts to the next stage so that the transmitted power can be maximized. The rectifier is used to convert the RF signal into dc and also to boost the rectified signal [11], [12].
Figure 1. A concept diagram of a typical RFEH system.
WPT is the transmission of electrical energy without the need for any wire or waveguide medium as a link. It is reliable, fast, and efficient, and it can be used for short-distance or long-distance transmission with low maintenance requirements. The two main categories of WPT techniques are near-field and far-field operation. In a near-field system, power transmission occurs over a short distance through the use of inductive coupling (IC) in a magnetic field or capacitive coupling (CC) in an electric field. In IC, a transformer is formed between the transmitter and the receiver coils, illustrated in Figure 2(a). An oscillating magnetic field is generated when an alternating current passes through the transmitter coil. When this magnetic field passes through the receiving coil, an alternating EMF is induced, creating alternating current at the receiver. This current may either directly drive the load or be rectified to dc at the receiver that drives the load. In CC, energy is transmitted between two electrodes. The transmitter and receiver electrodes form a capacitor. An alternating voltage generated by the transmitter is applied to the transmitting plate, which in turn induces alternating potential at the receiver plate, thereby producing an alternating current to drive the load. Another near-field WPT technology is the magnetic resonance coupling (MRC) method that operates on the principle of resonant coupling shown in Figure 2(b). According to the theory of resonant coupling, two resonant devices have maximum coupling between them if their operating frequency is the same as their resonant frequency [13].
Figure 2. A concept diagram of (a) a near-field IC WPT system and (b) an MRC WPT system.
In a far-field (radiative) WPT system, the RF signal is generated by a dedicated RF source. The RF signal is then radiated into free space and is captured by the receiving antennas. The received RF signal is then converted into dc using the rectifying circuit before delivering it to the load. A typical far-field WPT system is illustrated in Figure 3. Due to this aspect, the RFEH mechanism is referred to as a form of far-field WPT technique in some of the literature. However, there exist some distinct differences between the two technologies in terms of their design requirements. In RFEH systems, the energy is received from ambient sources that are unknown in nature. However, in the case of radiative WPT systems, a dedicated transmitting unit is used to generate the RF signal. Further, the rectennas designed for ambient RFEH should have omnidirectional radiation characteristics since the orientation of the incoming EM wave is unknown. Unlike RFEH, high-gain directional rectennas are preferred for far-field WPT systems to overcome the challenges of long-distance communication. In addition, RFEH can be a viable solution in smart cities and urban environments, whereas WPT can be helpful for powering devices in rural and semiurban areas [14], [15].
Figure 3. A concept diagram of a far-field WPT system.
The essential requirement of an RFEH system is high power conversion efficiency due to the limited amount of available harvested power and energy storage capacity. A high-gain antenna is a good candidate to deliver maximum RF energy to the rectifying circuit. The availability of several frequency bands for RFEH systems makes a multiband or broadband antenna suitable. However, the demand for compact devices requires the use of antennas that are low profile and low cost. Also, the rectifier should be designed to operate at a low input power level due to the low power density of ambient EM fields. Some other design requirements of an RFEH system are circular polarization for minimal mismatch loss, efficient nonlinearity modeling of the RFEH systems, miniaturization without degradation in efficiency, and high tolerance to environmental effects [12]. CI techniques have been widely used to overcome these design challenges in modeling and optimizing RFEH systems and their components.
The S-parameters of a zero-bias Schottky diode were modeled with an ANN in [16]. The measured S-parameter data samples of the diode were collected from a vector network analyzer and used to train a single network. The frequency and input power were the input data for the ANN model, and the outputs were all the S-parameters. The model utilized two input neurons, eight output neurons, and two hidden layers, each with five neurons. The authors of [17] present a technique based on ANNs for modeling the nonlinear harvesting behavior of RF energy harvesters. A harvester prototype operating at 915 MHz was created to generate the datasets required for training the ANN model. The ANN model based on a multilayer perceptron (MLP) neural network (NN) utilized two input neurons, the harvesting time and the distance between the transmitter and receiver; one output neuron, the output voltage of the harvester; and one hidden layer with sixty hidden neurons.
For multiuser transmission, it is difficult for wireless transmitters to acquire the RF energy harvester channel. To solve this problem, a deep NN (DNN) implemented at each wireless transmitter was trained using the harvested energy to estimate the transmit covariance matrices. After the DNNs were trained, they were applied online to the harvester with unknown channels. The harvested energy was supplied back using the same pilot transmit covariance matrices. The DNN developed the transmit covariance matrices while approaching the maximum total received power at the harvester [18]. A similar problem was addressed in [19]. Here the authors present a DNN model to find the optimal power allocation using the channel vectors and the information rate requirement as the inputs of the model. The channels were divided into several classes using a k-means clustering algorithm, and a DNN was trained for each class. The transmitter determined the class of the required channel and employed a DNN to identify the best transmit covariance matrix. The precision of the DNN model with respect to the power loss threshold is shown in Figure 4. In Figure 4, the precision is calculated as the ratio of Ns to NT, where Ns represents the transmissions having a power ratio greater than a preset threshold out of NT tests. In [20], a DNN model is proposed to optimize a dc-boosted gate bias rectifier to achieve the maximum power conversion efficiency. The proposed model consists of five hidden layers and utilized an exponential linear unit as the activation function. A CMOS rectifier for RFEH was optimized using the DNN to identify the maximum power conversion efficiency for various input RF power sources from the antenna for the load conditions in [21]. In [22], a Wi-Fi channel state information (CSI)-based presence detection system consisting of preprocessing for data representation, a convolutional NN (CNN) for motion detection, and postprocessing for the eventual presence detection is explored. The rationale for choosing a CNN is its ability to exploit CSI variation in multiple dimensions due to the multiple input, multiple output–orthogonal frequency-division multiplexing waveforms at the physical layer.
Figure 4. The precision versus the power loss ratio threshold as addressed in (a) [18] and (b) [19].
The relationship between the pixel connection of the dual-port antenna and the received power was analyzed using Z-parameters in [23]. The GA was then used to find the optimal pixel configuration that maximizes the received power in an ambient RF field. The GA was used to optimize a hybrid fractal antenna shown in Figure 5 at 2.45 GHz in [24]. The objective function used for optimization denotes the relation between the resonant frequency and the dimensions of the antenna and was generated using a curve-fitting method. In [25], a patch antenna with H- and E-shaped slots is presented for RFEH, and a GA was used to optimize the antenna for enhancing the gain. A dual-band antenna was designed in [26] to operate at 3.5 GHz and 5.8 GHz. The gain of the antenna was maximized by etching a rectangular unit cell optimized using the GA. In [27], a tetraband antenna to harvest RF energy from the GSM 900, the GSM 1800, the UMTS, and Wi-Fi sources was designed using a GA. A real-time matching circuit consisting of two shunt capacitors and one series capacitor connected in parallel with an inductor was proposed in [28]. The optimal combinations of the capacitor values were selected using a GA.
Figure 5. The application of a GA for the optimization of the radiator shape to maximize received power [24].
To collect the maximum power from RF energy sources and enhance the efficiency of a dc-to-dc converter, a power management circuit was designed and optimized using PSO in [29]. In the PSO, the converter efficiency was used as the fitness function, and inductor and on-time were chosen as the optimized parameters. An RFEH circuit consisting of a patch antenna, multistage matching network, and a rectifying circuit was proposed in [30]. The impedances and the transmission line length of the matching network were determined using PSO. In [31], a broadband rectenna covering six frequency bands was proposed to harvest energy from multiple ambient RF sources at the same time. To allocate multiple users to harvest energy in a cloud-based cellular network, an optimization approach was presented. A quantum-behaved PSO (QPSO) algorithm was utilized to solve the optimization problem, ensuring maximum energy efficiency while satisfying the energy consumption constraint and the total data rate requirement. The convergence curve of the QPSO for 10 users with an increasing number of particles is depicted in Figure 6. In [32], an energy-harvesting algorithm based on PSO is proposed for optimal resource allocation and for power allocation to find a tradeoff between the two-way relaying device-to-device link rate and the energy efficiency. The proposed algorithm was able to find the power-splitting factor and the optimal relay selection. Recently, the PSO technique was implemented in [33] to optimize the height and position of an unmanned aerial vehicle relay and its time-switching ratio.
Figure 6. The energy efficiency versus the number of iterations [31].
An Internet of Things (IoT)-based network for monitoring and controlling a museum’s interior environment was developed where always-on and power-hungry sensor devices were energized through RFEH [34]. Rectenna arrays were used to collect the RF energy and convert it to electric power to prolong the lifetime of the sensor nodes. To find the daily trends in the collected environmental data, a deep learning framework based on a gated recurrent unit was utilized. In [35], a fractal antenna for RFEH systems was optimized using genetic swarm optimization (GSO), which is a combination of a GA and PSO. GSO was also utilized in [36] to design a loop wire antenna with high efficiency. In [37], a salp swarm algorithm (SSA) was used to realize a modified E-shaped antenna with dual-band characteristics for RFEH systems. The operating frequencies of the proposed antenna designed using SSA cover the allotted bands of LTE-2600 and 5G NR mobile communication networks. SSA is inspired by the swarming mechanism of salps in oceans. The SSA’s searching abilities are strengthened by the salps’ dynamic movements as they help them to avoid local optima.
WPT systems have been developed for a variety of applications ranging from charging low-power devices to electric vehicles to biomedical devices. Different applications have different requirements and result in different parameters for their respective designs. Calculating these unknown parameters through trial-and-error methods is time-consuming. Further, the outcome of such an approach may not always agree with the theoretical simulation and calculation. CI techniques can simplify the optimization process and enhance the performance accuracy of WPT systems. The following subsections discuss some of the significant contributions of CI modeling in WPT system design.
A feedforward ANN-based automatic impedance matching method is proposed to improve the power transfer efficiency (PTE) of an WPT system in [38]. Initially, a mapping connection was established between the equivalent load impedance seen by the transmitter and the optimal matched capacitor. Then the impedance was matched through the ANN, and consequently, the PTE improved up to 78.33%. This improved impedance matching was retained within a distance of 10 to 30 cm, as shown in Figure 7(a). ANN models for real-time range-adaptive automatic impedance matching WPT systems, as shown in Figure 8, were implemented in [39]. The parameters of the matching network and the range-adaptive transmitter coils were optimized to realize an efficient automatic impedance matching network.
Figure 7. The PTE as a function of distance according to (a) [38] and (b) [71].
Figure 8. A block diagram of the WPT system in [39].
A modeless prediction scheme was introduced in [40] using an ANN approach for WPT systems that can efficiently predict the variable mutual inductance among the transceiver coils. In [41], a DNN-based estimation method was implemented for estimating the frequency spectra and design parameters of a metamaterial unit cell for WPT applications. In [42], a deep feedforward NN (DFNN) was proposed to optimize the secure transmission parameters of a WPT system that maximizes effective throughput. In [43], a WPT system based on magnetic resonance was proposed where a multicoil transmitter was conceptualized for transferring energy simultaneously to multiple single-coil receivers. A strategy integrating offline search and online learning was presented in the corresponding work to enhance the system PTE with numerous receivers with variable placements. To maximize the system PTE of the multiple receivers with unknown positions, a method combining offline search and online learning was proposed. Two algorithms based on random forest and DNN were implemented for online prediction of the receiver positions. A novel channel-tracking scheme was developed using a deep recurrent NN architecture by uniting long short-term memory with a feedforward NN in [44]. In [40], Shen et al. presented a modeless prediction scheme using an NN to dynamically estimate the varying coupling effects of practical working conditions like the variation in relative position between the transmitting and the pickup coils. The proposed methodology has the salient advantages of being simpler to calculate and convenient to measure. The work in [46] suggests a technique to detect the presence of foreign metal objects while also forecasting the misalignment distance between the primary and secondary coils in WPT systems. An NN trained using electromagnetic field simulations is the basis of the suggested approach. The differential voltages of the detecting coils and the input voltage of the primary coil make up the training data for the NN.
To overcome the frequency splitting phenomenon in a WPT system, the equivalent circuit parameters of the WPT system were extracted using a GA in [47]. A real-time active matching circuit for a biomedical WPT system was designed and optimized using a GA in [48]. In [49], a GA with a data clustering approach based on the measured S-parameters of randomly moving coils was used to design a matching circuit. In [50], a GA was used to optimally design single-, double-, and multiple-layer variable-width printed spiral coils for both the transmitter and the receiver with minimized parasitic losses. In [51], the topology for optimizing the magnetic core for a WPT device is presented, which is shown in Figure 9. Gaussian basis functions were used to represent the magnetic core shapes and the coils for the optimization process. The optimization process was carried out to maximize the coupling coefficient using the GA and a 3D finite-element analysis. In [52], a GA approach is presented for estimating the design parameters of a planar rectangular coil with unevenly spaced turns. Near-field beamforming using a two-layer planar loop array resonator for a magnetic resonance WPT system is presented in [53]. The GA was used to find the optimal capacitance of each loop for maximum PTE. The inclusion of a relay coil between the transmitting and the receiving units may significantly lower the current in the transmitter coil, boosting system efficiency and minimizing electromagnetic field leakage. In [54], a drone charging system using WPT is proposed, and the GA is used to optimize the side lengths of the transmitting coil to achieve high and stable PTE. In [55], the GA is used to optimize the trace width and spacing of a planar spiral coil by maximizing the evenness of the axial magnetic field within an area of interest above the coil. In [56], a derivative-free optimization technique based on a GA is used to determine the optimal parameters of proportional–integral–derivative controllers for bidirectional inductive power transfer systems. A multiobjective GA (MGA) approach is presented in [57] for choosing the optimal weighting parameters of an H-infinity controller for a balanced time domain and robust performance in a series–series (SS) compensated inductive WPT system.
Figure 9. The optimized fabricated WPT core [51].
In [58], the parameters of a relay coil are optimized by theoretical calculation and by using a multiobjective GA. For fast multiobjective optimization of coil design in a WPT system, the design parameters of the coil system were optimized using a nondominated sorting GA (NSGA II) in [59]. In [60], the structure of a double-D pad for inductive WPT systems was optimized using an NSGA II to ensure a good coupling coefficient while minimizing the worst-case stray leakage magnetic fields. A dynamic WPT system based on a mixed integer nonlinear model for optimizing a feeder bus transit system is presented in [61]. The decision variables that were considered during the optimization problem include bus service frequency, route networks, dynamic WPT (DWPT) device locations, and capacity of the batteries. The goal was to save overall expenditures as much as possible by minimizing the total cost. A tangible nested GA was conceptualized to attain the best solution. In [62], light field optimization approaches based on the Gerchberg–Saxton algorithm, direct binary search, and a GA for optical WPT (OWPT) are proposed to achieve maximum photoelectric efficiency. The energy utilization and robustness of the OWPT system are improved using the GA.
For a dynamic WPT system for a roadway with W-type power supply rails, a design method for adjusting parameters such as coil width and ferrite size based on a combination of sensitivity analysis and a PSO algorithm is proposed in [63]. In [64], the PSO algorithm is employed to design and optimize transmitter pads for dynamic WPT systems for in-motion charging of electric vehicles. Multiobjective hybrid PSO and multiobjective real-numbered PSO algorithms for a circular coupler design are presented in [65]. Two different multiobjective function pairs are considered: the coil coupling coefficient and the maximum leakage field magnetic flux density and the product between the quality factor and the coil coupling coefficient combined with the maximum leakage field magnetic flux density. Further, in [63], a DWPT system is presented based on an automated guided vehicle and a W-type roadway. For the DWPT system with a W-type roadway, a design method for coil and ferrite based on a combination of sensitivity analysis and a PSO algorithm is proposed.
The mutual inductance of an SS-compensated WPT system is monitored through an adaptive differential evolution (ADE) algorithm in [67]. The same algorithm has also been used for optimizing the parameters of the receiving resonators used in a WPT system. In [68], a multiobjective genetic programming algorithm is used to generate low-complexity behavioral analytical models for the characterization and design optimization of WPT systems. In [69], a reinforcement learning (RL) algorithm is applied to develop a ferromagnetic core structure for a WPT system that can allow optimal coupling between the transmitting and receiving coils. In [70], a DNN-based WPT beam scheduling approach is proposed for a system that contains multiple IoT devices and a power beacon. The nonstatic behavior of IoT devices restricts the application of conventional beam scheduling schemes in these devices. In [71], the impedance mismatching problem of a WPT system is addressed by implementing the Sugeno fuzzy inference system. The proposed CI technique maps the load impedance of the circuit to a matched capacitor set for establishing an automatic impedance-matching method. This approach improves the efficiency and also helps to maintain the PTE at 86.35% within 4.5–18 cm, as can be seen in Figure 7(b). To minimize the field density and maximize the efficiency of a WPT link, an optimization approach is developed in [45]. The GA is used to determine the optimal design of the WPT link. The computational time of the optimization process is reduced using an ANN model based on a feedforward architecture consisting of an input layer; two hidden layers with twenty and five neurons in the first and second layers, respectively; and an output layer. In [66], a power and channel allocation method for WPT systems is presented for potential implementation in cognitive radio networks. The authors aimed to optimize the total wireless transmitted power in both the primary and the secondary systems with a probabilistic constraint. The probabilistic constraint was transformed using a support vector machine and the PSO used to find an optimal solution to the transformed resource allocation optimization problem.
CI techniques have emerged as one of the most popular approaches among the research community for the design and performance optimization of different microwave components. In keeping with this trend, CI techniques are being extensively implemented for the modeling and optimization of RFEH and WPT systems to minimize design complexity and enhance performance. While considerable research has been conducted on RFEH and WPT systems, research on the application of CI techniques is comparatively scarce. A pictorial representation of the amount of research conducted in the past three years in different areas is represented in Figure 10(a). It is quite evident that the popularity of CI techniques for developing efficient RFEH and WPT systems has grown significantly in recent years. This review article has therefore attempted to showcase the available research and development related to CI-based modeling and optimization of RFEH and WPT systems. In the literature, widely used CI techniques include ANNs, GAs, and PSO. Therefore, this article presents a comprehensive survey of advances in this field of research with a special focus on ANNs, GAs, and PSO. The distribution of more than 60 articles is shown in Figure 10(b). Most of the research work utilized GAs and PSO for the design and optimization of RFEH systems and their components. For WPT systems, a majority of the published work utilized GAs followed by ANNs and PSO.
Figure 10. (a) A comparison of the number of articles published in previous years. (b) The distribution of published articles on CI-based RFEH and WPT systems.
Tables 2 and 3 summarize the CI techniques employed in the literature for RFEH and WPT systems, respectively. For RFEH systems, antenna optimization is achieved through a GA in a majority of the work [23], [24], [25], [26], [27], where different parameters of the antenna are optimized to achieve the design requirements. For antenna structure optimization, it is necessary to find the parameters that significantly affect the performance of the antenna. For example, in [26], a rectangular cell is divided into 180 pixels, and a GA is used to find the best configuration out of 2180 different configurations, which yield the minimum reflection coefficient and the maximum gain. Both a GA and PSO are also used to optimize the matching circuit of an RFEH system in [28] and [30]. MLP is used to model the rectification element of the RFEH system in [16], whereas in [17], the harvesting behavior of the harvester is modeled using MLP. A DNN is used to model the rectifier to achieve maximum power conversion efficiency in [20] and [21]. The challenge of finding the different parameter values for WPT systems can be effectively reduced by CI techniques such as ANNs, GAs, and PSO. In [38] and [39], the matched capacitor set of MRC–WPT systems are optimized using an MLP to improve the PTE. The computation time for estimating and extracting the best parameter can be significantly reduced using ANNs. A GA is used in MRC and IC WPT systems to optimize and predict the values of various parameters in [47], [48], [49], [50], [51], [52], [53], and [54]. CI techniques such as ANNs, DNNs, CNNs, and adaptive neuro-fuzzy inference systems (ANFISs) are modeling algorithms that are used to make logical estimations of a system. However, optimization algorithms like GAs and PSO are used to optimize a structure or to fine-tune the parameters of modeling algorithms.
Table 2. A summary of CI modeling of RFEH systems.
Table 3. A summary of CI modeling of WPT systems.
In the conventional method of designing an RFEH or a WPT system, an initial design is conceptualized by the perception, expertise, and creativity of the designer. Following this initial process, the initial design is subjected to multiple rounds of simulation and experiments until the desired performance criteria are satisfied. The difficulty associated with this approach is that it is still unclear if the first design, conceptualized by the designer’s expertise, will provide the necessary performance for a given application requirement. Additionally, designing and optimizing the initial structure for RFEH and WPT systems takes longer because the nonlinear characteristic of the constituent components of the rectennas increases their complicacy. Complex problems that demand repetitive and repeated analysis with high precision and a higher pace of convergence can be solved using CI techniques. CI approaches can speed up the design process while retaining high accuracy, reducing errors, and saving time, as well as predicting circuit or system behavior, improving computing efficiency, and reducing the number of simulations required. It can also automate analytical model building. As microwave components are becoming more and more complex, designers can take advantage of CI techniques to generate mathematical models for their physical designs and perform fast and intelligent optimization on these mathematical models. However, amid all the advantages, CI techniques are still an expanding domain for the field of RFEH and WPT systems. There are multiple research gaps in the existing literature related to the implementation of CI techniques that may be taken up by young researchers for investigation and to provide state-of-the-art solutions. First, the application of CI approaches is primarily restricted to antenna design. Their implementation in designing other associated components, like rectifiers, matching circuits, or the even the complete rectenna system, is still almost nonexistent and therefore requires research attention. Lately, common rectenna systems that can satisfy the requirements of both WPT and RFEH applications are being extensively explored. However, designing and performing detailed investigations of these structures by means of numerical solvers involves a huge computational cost. On the other hand, these shared systems exhibit extremely nonlinear behavior during their operation. Thus, there is a great research opportunity in extending these intelligent optimization techniques to design future common nonlinear systems. In addition, the existing configurations can be improved upon to achieve further miniaturization using CI techniques with accelerated design processes and high accuracy. The authors are optimistic that many of these challenges will be resolved in the coming years.
This work was supported by the Science and Engineering Research Board, Government of India, under the Visiting Advanced Joint Research Scheme (Grant VJR/2019/000009, dated 22 July 2020).
[1] J. C. Bezdek, “On the relationship between neural networks, pattern recognition and intelligence,” Int. J. Approx. Reasoning, vol. 6, no. 2, pp. 85–107, Feb. 1992, doi: 10.1016/0888-613X(92)90013-P.
[2] A. P. Engelbrecht, Computational Intelligence: An Introduction. Chichester, U.K.: Wiley, 2002.
[3] D. Sarkar, T. Khan, F. A. Talukdar, and Y. M. M. Antar, “Computational intelligence paradigms for UWB antennas: Comprehensive review of analysis, synthesis and optimization,” Artif. Intell. Rev., vol. 56, no. 1, pp. 655–684, Jan. 2022, doi: 10.1007/s10462-022-10181-w.
[4] J. M. Zurada, Introduction to Artificial Neural Systems. St. Paul, MN, USA: West Publishing, 1992.
[5] J. M. Johnson and Y. R. Samii, “Genetic algorithms in engineering electromagnetics,” IEEE Antennas Propag. Mag., vol. 39, no. 4, pp. 7–21, Aug. 1997, doi: 10.1109/74.632992.
[6] J. Robinson and Y. R. Samii, “Particle swarm optimization in electromagnetics,” IEEE Trans. Antennas Propag., vol. 52, no. 2, pp. 397–407, Feb. 2004, doi: 10.1109/TAP.2004.823969.
[7] J. Huang, Y. Zhou, Z. Ning, and H. Gharavi, “Wireless power transfer and energy harvesting: Current status and future prospects,” IEEE Wireless Commun., vol. 26, no. 4, pp. 163–169, Aug. 2019, doi: 10.1109/MWC.2019.1800378.
[8] X. Gu, P. Burasa, S. Hemour, and K. Wu, “Recycling ambient RF energy: Far-field wireless power transfer and harmonic backscattering,” IEEE Microw. Mag., vol. 22, no. 9, pp. 60–78, Sep. 2021, doi: 10.1109/MMM.2021.3086335.
[9] A. Ali, M. N. Mohd Yasin, M. Jusoh, N. A. M. Ahmad Hambali, and S. R. Abdul Rahim, “Optimization of wireless power transfer using artificial neural network: A review,” Microw. Opt. Technol. Lett., vol. 62, no. 2, pp. 651–659, Feb. 2020, doi: 10.1002/mop.32089.
[10] W. Saeed, N. Shoaib, H. M. Cheema, and M. U. Khan, “RF energy harvesting for ubiquitous, zero power wireless sensors,” Int. J. Antennas Propag., vol. 2018, Apr. 2018, Art. no. 8903139, doi: 10.1155/2018/8903139.
[11] L. G. Tran, H. K. Cha, and W. T. Park, “RF power harvesting: A review on designing methodologies and applications,” Micro Nano Syst. Lett., vol. 5, no. 1, pp. 1–6, Dec. 2017, doi: 10.1186/s40486-017-0051-0.
[12] S. K. Divakaran, D. Das Krishna, and Nasimuddin, “RF energy harvesting systems: An overview and design issues,” Int. J. RF Microw. Comput.-Aided Eng., vol. 29, no. 1, pp. 1–15, Jan. 2019, doi: 10.1002/mmce.21633.
[13] B. W. Flynn and K. Fotopoulou, “Rectifying loose coils: Wireless power transfer in loosely coupled inductive links with lateral and angular misalignment,” IEEE Microw. Mag., vol. 14, no. 2, pp. 48–54, Mar./Apr. 2013, doi: 10.1109/MMM.2012.2234634.
[14] C. R. Valenta and G. D. Durgin, “Harvesting wireless power: Survey of energy-harvester conversion efficiency in far-field, wireless power transfer systems,” IEEE Microw. Mag., vol. 15, no. 4, pp. 108–120, Jun. 2014, doi: 10.1109/MMM.2014.2309499.
[15] M. A. Ullah et al., “A review on antenna technologies for ambient RF energy harvesting and wireless power transfer: Designs, challenges and applications,” IEEE Access, vol. 10, pp. 17,231–17,267, Feb. 2022, doi: 10.1109/ACCESS.2022.3149276.
[16] B. Milosevic, M. Radovanovic, B. Jokanovic, and Z. Marinkovic, “Artificial neural network model of zero-bias Schottky diode for energy harvesting,” in Proc. 14th Int. Conf. Adv. Technol. Syst. Serv. Telecommun. (TELSIKS), 2019, pp. 319–332, doi: 10.1109/TELSIKS46999.2019.9002357.
[17] M. B. Lam, N. T. Dang, T. H. Nguyen, and W. Y. Chung, “A neural network-based model of radio frequency energy harvesting characteristics in a self-powered food monitoring system,” IEEE Sens. J., vol. 19, no. 19, pp. 8813–8823, Oct. 2019, doi: 10.1109/JSEN.2019.2922346.
[18] Y. Qian, Y. Xing, and L. Dong, “Deep learning for radio-frequency energy harvesting with multiple wireless transmitters,” in Proc. IEEE 88th Veh. Technol. Conf. (VTC-Fall), 2018, pp. 1–2, doi: 10.1109/VTCFall.2018.8690766.
[19] Y. Xing, Y. Qian, and L. Dong, “Deep learning for optimized wireless transmission to multiple RF energy harvesters,” in Proc. IEEE 88th Veh. Technol. Conf. (VTC-Fall), 2018, pp. 1–5, doi: 10.1109/VTCFall.2018.8690775.
[20] W. W. Y. Lau, H. W. Ho, and L. Siek, “Deep neural network (DNN) optimized design of 2.45 GHz CMOS rectifier with 73.6% peak efficiency for RF energy harvesting,” IEEE Trans. Circuits Syst. I, Reg. Paper, vol. 67, no. 12, pp. 4322–4333, Dec. 2020, doi: 10.1109/TCSI.2020.3022280.
[21] H. W. Ho and W. W. Y. Lau, “Automated design optimization for CMOS rectifier using deep neural network (DNN),” in Proc. IEEE Wireless Power Transf. Conf. (WPTC), 2019, pp. 599–603, doi: 10.1109/WPTC45513.2019.9055537.
[22] Y. Liu, T. Wang, Y. Jiang, and B. Chen, “Harvesting ambient RF for presence detection through deep learning,” IEEE Trans. Neural Netw. Learn. Syst., vol. 33, no. 4, pp. 1571–1583, Apr. 2022, doi: 10.1109/TNNLS.2020.3042908.
[23] S. Shen and R. D. Murch, “Designing dual-port pixel antenna for ambient RF energy harvesting using genetic algorithm,” in Proc. IEEE Int. Symp. Antennas Propag. USNC/URSI Nat. Radio Sci. Meeting, 2015, pp. 1286–1287, doi: 10.1109/APS.2015.7305032.
[24] B. S. Dhaliwal, A. Kaur, S. Pattnaik, and S. S. Pattnaik, “Optimization of hybrid fractal antenna using curve fitting and GA approach for RF energy harvesting application,” in Proc. IEEE Indian Conf. Antennas Propog. (InCAP), 2019, pp. 3–6, doi: 10.1109/InCAP47789.2019.9134573.
[25] S. Vijay Gokul, M. Suba Lakshmi, and T. Swetha, “Design of RF energy harvesting antenna using optimization techniques,” Int. J. Eng. Res., vol. 9, no. 3, pp. 595–601, 2020.
[26] M. C. Derbal and M. Nedil, “A high gain dual band rectenna for RF energy harvesting applications,” Prog. Electromagn. Res. Lett., vol. 90, pp. 29–36, Feb. 2020, doi: 10.2528/PIERL19122604.
[27] D. Masotti, A. Costanzo, M. Del Prete, and V. Rizzoli, “Genetic-based design of a tetra-band high-efficiency radio-frequency energy harvesting system,” IET Microw., Antennas Propag., vol. 7, no. 15, pp. 1254–1263, Dec. 2013, doi: 10.1049/iet-map.2013.0056.
[28] J. Bito, M. M. Tentzeris, and A. Georgiadis, “A hybrid heuristic design technique for real-time matching optimization for wearable near-field ambient RF energy harvesters,” in Proc. IEEE MTT-S Int. Microw. Symp. (IMS), 2016, pp. 1–4, doi: 10.1109/MWSYM.2016.7540074.
[29] M. M. Ababneh, S. Perez, and S. Thomas, “Optimized power management circuit for RF energy harvesting system,” in Proc. IEEE 18th Wireless Microw. Technol. Conf. (WAMICON), 2017, pp. 1–4, doi: 10.1109/WAMICON.2017.7930238.
[30] S. Ulker, “Modified patch antenna with matching network for energy harvesting at multiple frequencies,” in Proc. 18th Mediterranean Microw. Symp. (MMS), 2018, pp. 283–286, doi: 10.1109/MMS.2018.8611983.
[31] Y. Zhao, V. C. M. Leung, C. Zhu, H. Gao, Z. Chen, and H. Ji, “Energy-efficient sub-carrier and power allocation in cloud-based cellular network with ambient RF energy harvesting,” IEEE Access, vol. 5, pp. 1340–1352, Feb. 2017, doi: 10.1109/ACCESS.2017.2667678.
[32] M. M. Salim, D. Wang, H. A. El Atty Elsayed, Y. Liu, and M. A. Elaziz, “Joint optimization of energy-harvesting-powered two-way relaying D2D communication for IoT: A rate-energy efficiency tradeoff,” IEEE Internet Things J., vol. 7, no. 12, pp. 11,735–11,752, Dec. 2020, doi: 10.1109/JIOT.2020.2999618.
[33] A.-N. Nguyen, D.-B. Ha, V. N. Vo, V.-T. Truong, D.-T. Do, and C. So-In, “Performance analysis and optimization for IoT mobile edge computing networks with RF energy harvesting and UAV relaying,” IEEE Access, vol. 10, pp. 21,526–21,540, Feb. 2022, doi: 10.1109/ACCESS.2022.3150046.
[34] N. A. Eltresy et al., “RF energy harvesting IoT system for museum ambience control with deep learning,” Sensors, vol. 19, no. 20, Oct. 2019, Art. no. 4465, doi: 10.3390/s19204465.
[35] M. Mussetta, H. Shadmehr, F. Grimaccia, A. Gandelli, and R. E. Zich, “Optimization of a radio frequency energy harvesting device,” in Proc. IEEE Congr. Evol. Comput., 2012, pp. 1–5, doi: 10.1109/CEC.2012.6256141.
[36] M. Roscia, D. Zaninelli, G. C. Lazaroiu, and H. Shadmehr, “Using design optimization in radio frequency energy harvesting systems,” in Proc. 4th Int. Conf. Clean Elect. Power (ICCEP), 2013, pp. 740–744, doi: 10.1109/ICCEP.2013.6586939.
[37] A. D. Boursianis, M. S. Papadopoulou, S. Nikolaidis, and S. K. Goudos, “Dual-band single-layered modified E-shaped patch antenna for RF energy harvesting systems,” in Proc. 24th Eur. Conf. Circuit Theory Des. (ECCTD), 2020, pp. 1–4, doi: 10.1109/ECCTD49232.2020.9218354.
[38] Y. Li, W. Dong, Q. Yang, J. Zhao, L. Liu, and S. Feng, “An automatic impedance matching method based on the feedforward-backpropagation neural network for a WPT system,” IEEE Trans. Ind. Electron., vol. 66, no. 5, pp. 3963–3972, May 2019, doi: 10.1109/TIE.2018.2835410.
[39] S. Jeong, T. H. Lin, and M. M. Tentzeris, “A real-time range-adaptive impedance matching utilizing a machine learning strategy based on neural networks for wireless power transfer systems,” IEEE Trans. Microw. Theory Tech., vol. 67, no. 12, pp. 5340–5347, Dec. 2019, doi: 10.1109/TMTT.2019.2938753.
[40] S. Shen, Z. Zhang, Y. Wu, R. Wang, and Z. Liang, “Modeless prediction of variable coupling effect for multiple-load wireless power transfer,” IEEE Trans. Magn., vol. 58, no. 2, pp. 1–5, Feb. 2022, doi: 10.1109/TMAG.2021.3089784.
[41] H. N. Bui, J. S. Kim, and J. W. Lee, “Design of tunable metasurface using deep neural networks for field localized wireless power transfer,” IEEE Access, vol. 8, pp. 194,868–194,878, Oct. 2020, doi: 10.1109/ACCESS.2020.3033527.
[42] D. He, C. Liu, H. Wang, and T. Q. S. Quek, “Learning-based wireless powered secure transmission,” IEEE Wireless Commun. Lett., vol. 8, no. 2, pp. 600–603, Apr. 2019, doi: 10.1109/LWC.2018.2881976.
[43] T. Bai, B. Mei, L. Zhao, and X. Wang, “Machine learning-assisted wireless power transfer based on magnetic resonance,” IEEE Access, vol. 7, pp. 109,454–109,459, Aug. 2019, doi: 10.1109/ACCESS.2019.2933679.
[44] J. M. Kang, C. J. Chun, I. M. Kim, and D. I. Kim, “Deep RNN-based channel tracking for wireless energy transfer system,” IEEE Syst. J., vol. 14, no. 3, pp. 4340–4343, Sep. 2020, doi: 10.1109/JSYST.2020.2975188.
[45] A. O. Hariri, T. Youssef, A. Elsayed, and O. Mohammed, “A computational approach for a wireless power transfer link design optimization considering electromagnetic compatibility,” IEEE Trans. Magn., vol. 52, no. 3, pp. 1–4, Mar. 2016, doi: 10.1109/TMAG.2015.2492922.
[46] Y. Gong, Y. Otomo, and H. Igarashi, “Neural network for both metal object detection and coil misalignment prediction in wireless power transfer,” IEEE Trans. Magn., vol. 58, no. 9, pp. 1–4, Sep. 2022, doi: 10.1109/TMAG.2022.3176908.
[47] J. Han, Y. D. Kim, and N. H. Myung, “Efficient performance optimisation of wireless power transmission using genetic algorithm,” Electron. Lett., vol. 50, no. 6, pp. 462–464, Mar. 2014, doi: 10.1049/el.2014.0030.
[48] J. Bito, S. Jeong, and M. M. Tentzeris, “A real-time electrically controlled active matching circuit utilizing genetic algorithms for wireless power transfer to biomedical implants,” IEEE Trans. Microw. Theory Tech., vol. 64, no. 2, pp. 365–374, Feb. 2016, doi: 10.1109/TMTT.2015.2513765.
[49] J. Bito, S. Jeong, and M. M. Tentzeris, “A novel heuristic passive and active matching circuit design method for wireless power transfer to moving objects,” IEEE Trans. Microw. Theory Tech., vol. 65, no. 4, pp. 1094–1102, Apr. 2017, doi: 10.1109/TMTT.2017.2672544.
[50] S. Mehri, A. C. Ammari, J. B. H. Slama, and M. Sawan, “Design optimization of multiple-layer PSCs with minimal losses for efficient and robust inductive wireless power transfer,” IEEE Access, vol. 6, pp. 31,924–31,934, May 2018, doi: 10.1109/ACCESS.2018.2831785.
[51] Y. Otomo and H. Igarashi, “A 3-D topology optimization of magnetic cores for wireless power transfer device,” IEEE Trans. Magn., vol. 55, no. 6, pp. 1–5, Jun. 2019, doi: 10.1109/TMAG.2019.2900744.
[52] Y. Zhang, L. Wang, Y. Guo, and Y. Zhang, “Optimisation of planar rectangular coil achieving uniform magnetic field distribution for EV wireless charging based on genetic algorithm,” IET Power Electron., vol. 12, no. 10, pp. 2706–2712, Aug. 2019, doi: 10.1049/iet-pel.2018.6202.
[53] Y. H. Nam and J. H. Lee, “Efficient near-field beamforming using two-layer planar loop array for magnetic resonance wireless power transfer,” IEEE Microw. Wireless Compon. Lett., vol. 30, no. 8, pp. 818–820, Aug. 2020, doi: 10.1109/LMWC.2020.3000768.
[54] C. Rong et al., “Optimization design of resonance coils with high misalignment tolerance for drone wireless charging based on genetic algorithm,” IEEE Trans. Ind. Appl., vol. 58, no. 1, pp. 1242–1253, Jan./Feb. 2022, doi: 10.1109/TIA.2021.3057574.
[55] Q. Xu, Q. Hu, H. Wang, Z. H. Mao, and M. Sun, “Optimal design of planar spiral coil for uniform magnetic field to wirelessly power position-free targets,” IEEE Trans. Magn., vol. 57, no. 2, pp. 1–9, Feb. 2021, doi: 10.1109/TMAG.2020.3042131.
[56] M. J. Neath, A. K. Swain, U. K. Madawala, and D. J. Thrimawithana, “An optimal PID controller for a bidirectional inductive power transfer system using multiobjective genetic algorithm,” IEEE Trans. Power Electron., vol. 29, no. 3, pp. 1523–1531, Mar. 2014, doi: 10.1109/TPEL.2013.2262953.
[57] M. Yang, Y. Li, H. Du, C. Li, and Z. He, “Hierarchical multiobjective H-infinity robust control design for wireless power transfer system using genetic algorithm,” IEEE Trans. Control Syst. Technol., vol. 27, no. 4, pp. 1753–1761, Jul. 2019, doi: 10.1109/TCST.2018.2814589.
[58] F. Wen, X. Chu, Q. Li, R. Li, L. Liu, and F. Jing, “Optimization on three-coil long-range and dimension-asymmetric wireless power transfer system,” IEEE Trans. Electromagn. Compat., vol. 62, no. 5, pp. 1859–1868, Oct. 2020, doi: 10.1109/TEMC.2020.2976652.
[59] L. Tan, Z. Tang, R. Zhong, X. Huang, H. Liu, and C. Chen, “An optimization strategy based on dimension reduction method in wireless charging system design,” IEEE Access, vol. 7, pp. 151,733–151,745, Oct. 2019, doi: 10.1109/ACCESS.2019.2948196.
[60] Z. Luo, X. Wei, M. G. S. Pearce, and G. A. Covic, “Multiobjective optimization of inductive power transfer double-D pads for electric vehicles,” IEEE Trans. Power Electron., vol. 36, no. 5, pp. 5135–5146, May 2021, doi: 10.1109/TPEL.2020.3029789.
[61] G. Chen, D. Hu, and S. Chien, “Optimizing battery-electric-feeder service and wireless charging locations with nested genetic algorithm,” IEEE Access, vol. 8, pp. 67,166–67,178, Apr. 2020, doi: 10.1109/ACCESS.2020.2985168.
[62] P. Xu, W. Zhang, and Z. He, “Light field optimization for optical wireless power transfer,” IEEE Photon. J., vol. 13, no. 1, pp. 1–9, Feb. 2021, doi: 10.1109/JPHOT.2020.3044948.
[63] D. Wang, F. Chaowei, Q. Zhao, and T. Hu, “A PSO-based optimization design of W-type noncontact transformer for stable power transfer in DWPT system,” IEEE Trans. Ind. Appl., vol. 58, no. 1, pp. 1211–1221, Jan./Feb. 2022, doi: 10.1109/TIA.2021.3104265.
[64] R. Tavakoli, E. M. Dede, C. Chou, and Z. Pantic, “Cost-efficiency optimization of ground assemblies for dynamic wireless charging of electric vehicles,” IEEE Trans. Transport. Electrific., vol. 8, no. 1, pp. 734–751, Mar. 2022, doi: 10.1109/TTE.2021.3105573.
[65] T. Yilmaz, N. Hasan, R. Zane, and Z. Pantic, “Multi-objective optimization of circular magnetic couplers for wireless power transfer applications,” IEEE Trans. Magn., vol. 53, no. 8, pp. 1–12, Aug. 2017, doi: 10.1109/TMAG.2017.2692218.
[66] A. P. Shrestha and S. J. Yoo, “Optimal resource allocation using support vector machine for wireless power transfer in cognitive radio networks,” IEEE Trans. Veh. Technol., vol. 67, no. 9, pp. 8525–8535, Jun. 2018, doi: 10.1109/TVT.2018.2850799.
[67] Y. Yang, S. C. Tan, and S. Y. R. Hui, “Front-end parameter monitoring method based on two-layer adaptive differential evolution for SS-compensated wireless power transfer systems,” IEEE Trans. Ind. Informat., vol. 15, no. 11, pp. 6101–6113, Nov. 2019, doi: 10.1109/TII.2019.2924926.
[68] G. Di Capua et al., “Mutual inductance behavioral modeling for wireless power transfer system coils,” IEEE Trans. Ind. Electron., vol. 68, no. 3, pp. 2196–2206, Mar. 2021, doi: 10.1109/TIE.2019.2962432.
[69] B. G. Choi, E. S. Lee, and Y. S. Kim, “Optimal structure design of ferromagnetic cores in wireless power transfer by reinforcement learning,” IEEE Access, vol. 8, pp. 179,295–179,306, Sep. 2020, doi: 10.1109/ACCESS.2020.3027765.
[70] H. S. Lee and J. W. Lee, “Adaptive wireless power transfer beam scheduling for non-static IoT devices using deep reinforcement learning,” IEEE Access, vol. 8, pp. 206,659–206,673, Nov. 2020, doi: 10.1109/ACCESS.2020.3037323.
[71] Y. Li, W. Dong, Q. Yang, S. Jiang, X. Ni, and J. Liu, “Automatic impedance matching method with adaptive network based fuzzy inference system for WPT,” IEEE Trans. Ind. Informat., vol. 16, no. 2, pp. 1076–1085, Feb. 2020, doi: 10.1109/TII.2019.2913792.
Digital Object Identifier 10.1109/MMM.2023.3284764