 Original research
 Open Access
 Published:
Dynamic reconfiguration for TEG systems under heterogeneous temperature distribution via adaptive coordinated seeker
Protection and Control of Modern Power Systems volume 7, Article number: 38 (2022)
Abstract
A thermoelectric generation (TEG) system has the weakness of relatively low thermoelectric conversion efficiency caused by heterogeneous temperature distribution (HgTD). Dynamic reconfiguration is an effective technique to improve its overall energy efficiency under HgTD. Nevertheless, numerous combinations of electrical switches make dynamic reconfiguration a complex combinatorial optimization problem. This paper aims to design a novel adaptive coordinated seeker (ACS) based on an optimal configuration strategy for largescale TEG systems with series–parallel connected modules under HgTDs. To properly balance global exploration and local exploitation, ACS is based on ‘divideandconquer’ parallel computing, which synthetically coordinates the local searching capability of tabu search (TS) and the global searching capability of a pelican optimization algorithm (POA) during iterations. In addition, an equivalent reoptimization strategy for a reconfiguration solution obtained by metaheuristic algorithms (MhAs) is proposed to reduce redundant switching actions caused by the randomness of MhAs. Two case studies are carried out to assess the feasibility and superiority of ACS in comparison with the artificial bee colony algorithm, ant colony optimization, genetic algorithm, particle swarm optimization, simulated annealing algorithm, TS, and POA. Simulation results indicate that ACS can realize fast and stable dynamic reconfiguration of a TEG system under HgTDs. In addition, RTLAB platformbased hardwareintheloop experiments are carried out to further validate the hardware implementation feasibility.
Introduction
Coinciding with rapid economic growth, over the past few decades, global demand for electrical energy has increased dramatically [1,2,3]. However, traditional fossil fuels currently acting as the main source of electricity generation, e.g., oil, coal, natural gas [4], face everincreasing depletion while causing severe environmental deterioration [5]. Hence, it is crucial to reform the current energy structure and utilization patterns. Renewable energy is envisaged as a revolutionary and promising candidate to ease the continued tension and pressure in the global energy crisis [6].
In addition to wind, hydro, solar, and hydrogen [7, 8], thermoelectric generation (TEG) also possesses remarkable application potential and value [9] because of its prominent advantages of simple structure, sturdiness, being noiseless, long service life, etc. TEG has been applied in some industrial production, such as in automobile engines [10], geothermal energy exploitation [11], natural gas boilers [12], solar thermoelectric cooling systems [13], combined heat and power generation [14], wearable devices [15], energyautonomous sensors [16], etc.
However, the low energy conversion efficiency of TEG systems given unpredictable and ineluctable heterogeneous temperature distribution (HgTD) is the main obstacle that limits its larger scale and wider range of application [17]. To solve this issue, efficient thermoelectric materials and maximum power point tracking (MPPT) technology are considered as two promising solutions [18]. Unfortunately, the former is hard for largescale applications because of its high cost and long research process [19].
In contrast, MPPT is a more feasible and practical solution. Thus far, a large variety of MPPT approaches have been exploited and employed to capture the global maximum power point (GMPP) of TEG systems under HgTD. Perturb and observe (P&O) [20] and incremental conductance (INC) [21] are two of the most commonly used MPPT strategies because of their simple operational mechanisms. Nevertheless, their performance tends to degrade when multiple local maximum power points (MLMPPs) [22] appear caused by HgTD. Thus, a series of metaheuristic algorithm (MhAs)based methods have been proposed to find the GMPP based on their superior global searching ability, e.g., the fast atom search optimizer (FASO) [4], interacted collective intelligence (ICI) [17], adaptive compass search [23], equilibrium optimizer (EO) [22], arithmetic optimization algorithm (AOA) [24], etc. In addition, a novel adaptive rapid neural optimization approach is designed in [25] to realize an available approximation of duty cycle and output power curves for centralized TEG systems through a generalized regression neural network. This significantly reduces power oscillations and energy losses.
It is noteworthy that the aforementioned MPPT methods are all accomplished by controlling the centralized TEG converter to match the impedances of the source and load [26]. This cannot improve the potential GMPP of TEG systems. To fundamentally address this issue, TEG system reconfiguration can be an alternative and effective tool. It is inspired by photovoltaic (PV) system reconfiguration in partial shade conditions (PSC). Specifically, reconfiguration strategies for TEG systems can be divided into two types that are similar to PV system reconfiguration [27], i.e., static and dynamic. Static reconfiguration disperses the temperature difference area by changing the physical position of TEG modules rather than changing electrical connection, while dynamic reconfiguration is capable of dynamically altering the electrical interconnection through switching matrices against various temperature variations. This method generally possesses higher flexibility and lower complexity.
To alleviate adverse effects (e.g., module mismatch, difficulty in tracking the maximum power point, extreme power loss, hot spots, etc. [28]) of partial shading caused by meteorological or geographic factors on PV systems, numerous intensive studies have been done to develop sophisticated technologies on the reconfiguration of PV systems. For instance, a dragonfly algorithm (DA)based reconfiguration technique for PV systems is proposed in [29]. This is capable of achieving up to 22% output power improvement over conventional methods (e.g., electrical reconfiguration, Sudoku, etc.) under (3 × 3) and (9 × 9) arrays. Distinct from MhAsbased techniques, reconfiguration mechanisms enlightened by the magic square puzzle are conceived in [30, 31] to substantially reduce mismatch loss and enhance the GMPP of PV systems. These can offer an appropriate shade dispersion via relocation of PV modules. In addition, reference [32] applies reinforcement learning to the optimal reconfiguration of PV systems under PSC to provide frequency support for the power system.
However, there are not many current studies on the TEG system reconfiguration. In [10], a dynamic reconfiguration algorithm is proposed for a TEG array installed on vehicle radiators, one which can change interconnections (series or parallel) of TEG modules. Nevertheless, it assumes that the temperature difference between the TEG modules connected in parallel in the same row is similar, which is not always guaranteed in largescale TEG systems. Reference [33] achieves TEG array reconfiguration via adjusting switches with wireless sensors. This is suitable for occasions where wiring is inconvenient.
Although using electrical switchgear is feasible, how to obtain an appropriate switch adjustment scheme is complicated and challenging since dynamic reconfiguration of TEG systems is a nonlinear, discrete, and constrained optimization problem. Furthermore, traditional approaches, e.g., branch and bound as well as enumeration methods, face the risk of combination explosion with increasing dimensionality. Therefore, traditional methods are normally not applicable to the reconfiguration of largescale TEG systems. To address the above limitations, a novel adaptive coordinated seeker (ACS) is presented here for dynamic reconfiguration of TEG systems under different HgTDs.
The major contributions of this paper can be summarized as follows:

Dynamic reconfiguration of TEG systems under HgTDs is developed based on MhAs, for the first time.

ACS is first designed to coordinate the local searching capability of a tabu search (TS) and the global searching capability of a pelican optimization algorithm (POA). This can efficiently overcome drawbacks of traditional MhAs that are highly randomized and heavily dependent on certain parameters.

An equivalent reoptimization strategy (EROS) for reconfiguration solutions is developed to reduce unnecessary switching actions.

The practical and specific performance of an ACS is comprehensively evaluated under two typical scenarios, i.e., symmetrical and asymmetrical, via comparison with seven typical MhAs, i.e., artificial bee colony (ABC) algorithm, ant colony optimization (ACO), genetic algorithm (GA), particle swarm optimization (PSO), simulated annealing (SA) algorithm, TS, and POA.
The rest of this paper is structured as follows: TEG system modeling under HgTD is established in Sect. 2, while Sect. 3 introduces the optimization framework of the ACS. In Sect. 4, the execution process of TEG system reconfiguration and EROS are provided, respectively. Case studies, statistical analysis, and hardwareintheloop experiments are carried out in Sect. 5. Finally, main findings and conclusions are drawn in Sect. 6.
TEG system modelling
TEG module modelling
As illustrated in Fig. 1, a TEG module consists of two main components packaged together, i.e., two ceramic plates located on the cold and hot sides, respectively, while multiple thermoelectric couple units are connected in series with conductive metal strips between them. The thermoelectric couple unit is a fundamental and crucial element for a TEG module.
In general, the equivalent circuit of a TEG module is modeled as a voltage source V_{oc} connected with a series resistance R_{TEG} [4]. Here, the opencircuit voltage V_{oc} (V) is directly determined by the temperature difference between the two ceramic plates as [34]:
where n_{np} = 488 is the number of thermoelectric units, α_{sb} is the Seeback coefficient (V/K), T_{hs} (K) and T_{cs} (K) represent the temperature of the hot and cold sides, respectively.
The relationship between α_{sb} and Thomson coefficient τ (V/K) is given by [22]:
where T_{av} (K) is the average value of T_{hs} and T_{cs}.
For practical materials, τ is not zero. Thus, α_{sb} can be deduced from (2) as [4]:
where α_{sb0} = 210 μV/K and α_{sb1} = 120 μV/K. These are the constant portion and variation rate of α_{sb}, respectively [17, 21, 35]. T_{0} = 300 K denotes the temperature reference.
In addition, R_{TEG} is affected by the operational temperature of the TEG module [36]. This can be described as:
TEG system modelling under HgTD
In general, multiple TEG modules are interconnected in an SP electrical connection to obtain the expected voltage level and power density. As shown in Fig. 2, the series diodes are used to avoid circulating current between different strings (columns) within the TEG system when the total voltages of different strings mismatch, while the bypass diodes guarantee that the string with damaged TEG modules still operates normally.
According to the equivalent circuit of a single TEG module illustrated in Fig. 1, the total voltage V_{oc,m} and total internal resistance R_{m} at the mth string can be expressed as:
where N is the number of rows in the TEG system, β_{n,m} stands for the state of the bypass diode in the nth row and the mth column. This is either 0 (forward conduction) or 1 (reverse cutoff). V_{oc,n,m} and R_{TEG,n,m} represent the opencircuit voltage and internal resistance of the TEG module in the nth row and the mth column, respectively.
Therefore, the mth string is equalized to a voltage resource V_{oc,m} in series with an internal resistance R_{m}. Based on the Thevenin theorem, the total current I_{total} and total resistance R_{total} of the TEG system can be obtained by:
where M is the number of columns in the TEG system, γ_{m} represents the state of the mth series diode, which is determined by the terminal voltage V_{t} of the TEG system and V_{oc,m}.
A voltage vector [V_{0}, V_{1}, V_{2}, …, V_{k}, …, V_{K}] (1 ≤ K ≤ M) is acquired by sorting the unique values of [V_{0}, V_{oc,1}, V_{oc,2}, …, V_{oc,m}, …, V_{oc,M}] in ascending order, while V_{0} = 0 V. According to the maximum power transfer theorem [37], the maximum output power P_{max,k} (1 ≤ k ≤ K) can be written when V_{t} falls into the interval of (V_{k1}, V_{k}], as:
where K stands for the conductive number of TEG strings.
Hence, the potential GMPP of a TEG system under HgTD can be calculated by Eq. (11) and approached by the aforementioned MPPT techniques.
Design of adaptive coordinated seeker
Tabu search
The TS [38, 39] developed by F. Glover is a widely applied and representative algorithm in combinatorial optimization. It has powerful local searching capability. Its fundamental principle is demonstrated in Table 1, in which neighbourhood structure is defined as the interchange of any two TEG modules within the same row of the TEG system.
Pelican optimization algorithm (POA)
The swarmbased POA mimics the hunting process of the pelican [40], and has the merits of simple structure and easy implementation. POA contains two principal phases, i.e., moving towards prey and winging on the water surface, as described by (12) and (13), respectively. It should be noted that results are discretized using integer coding with an ascending rule [41], which makes POA applicable to the dynamic reconfiguration of TEG systems.
where x_{i}(it) and x_{i}(it + 1) indicate the status of the ith pelican in the itth and (it + 1)th iteration, respectively. x_{p} represents a randomly selected prey from remaining pelicans, while r_{1} and r_{2} are both random values distributed in the interval of [0, 1]. It stands for the maximum iteration, while F_{p} and F_{i} are objective function values of x_{p} and x_{i}(it), respectively.
Adaptive coordinated seeker
Based on the idea of parallelism, the entire population is divided equally and randomly into two subpopulations to execute different operations separately. All individuals for the first subpopulation perform algorithm #1 in Table 1. The neighbourhood size of each individual is adaptively adjusted according to its objective function value, as indicated in (14). Therefore, the drawback of traditional TS with heavy dependence on starting solution and search efficiency can be effectively lessened.
where NS_{j} is the neighbourhood size of the jth individual, with j = 1, 2, …, NoSP. NS_{0} is the initial neighbourhood size of each individual, and NoSP stands for the individual number per subpopulation, which is equal to NP/2.
For the second subpopulation, two different operations can be selected with the same probability, i.e., the discretized POA and TS without tabu. In addition, similar to the first subpopulation, the concept of neighbourhood search with adaptive adjustment is used for each individual of the second subpopulation.
To ensure communication between the two subpopulations, the top 20% of the row vectors with the largest difference between the optimal individuals of the two subpopulations are swapped in order after certain iterations (0.04It). For instance, the differences between the rows for the two solutions in Fig. 3 are 2/3 and 1, respectively. An effective updating mechanism [40] is adopted during the optimization of ACS to obtain a better solution. The overall ACS implementation procedure for the TEG system dynamic reconfiguration is provided in Fig. 4.
As the discretized POA has strong global search ability, it can effectively avoid the problem of premature convergence. In contrast, the mutation mechanism is employed in [42] to improve the PSO for parameter optimization of a transient stability prediction model, which can ensure population diversity. TS is biased to local exploitation, which can reasonably prevent ACS from blind search. Similarly, reference [43] constructs the memetic algorithm using a binary differential evolution algorithm and TS to extract optimal feature subsets for transient stability assessment of the power system. This can retain valid information to the greatest extent.
Dynamic reconfiguration of TEG system via ACS
Coding and switching matrix
For dynamic reconfiguration, each solution represents one arrangement mode while variables are the column serial number of each module in the TEG system, and the change of module locations is achieved by switching matrices. As shown in Fig. 2, reconfiguration within the same column has no impact for P_{GMPP} as the TEG modules in the same column are connected in series. Therefore, consecutive integer coding from 1 to M (column serial number of modules) is only adopted separately for each row of the TEG system. For instance, the initial solution and a possible solution, and their corresponding configuration of the switching matrices are shown in Fig. 3 for a (2 × 3) TEG system.
Objective function and constraints
Dynamic reconfiguration of the TEG system aims to maximize its potential GMPP, while a highquality solution corresponds to a high potential GMPP. Therefore, the objective function is designed as:
where X means the feasible solution space and x^{*} represents the optimal solution.
Moreover, dynamic reconfiguration is performed between different TEG modules in the same row. Thus, a feasible solution x should satisfy the following constraints:
where x(n,m) is the variable value of the nth row and the mth column in the feasible solution x.
Equivalent reoptimization of reconfiguration solution
Reconfiguration solutions obtained by various MhAs may not be in an optimal form under the same P_{GMPP} because of the inherent randomness of MhAs and the nonuniqueness of the solution. For example, assuming that the temperature conditions for modules (1,2) and (1,3) in Fig. 3 are identical, the two modules are completely equivalent according to (1) and (4). Thus, the switching states for modules (1,2) and (1,3) of the possible solution can also be consistent with the initial states. However, if the operation is performed directly following the possible solution, unnecessary switching actions will be caused. Therefore equivalent reoptimization for reconfiguration solutions obtained by various MhAs is essential and meaningful based on the temperature distribution of the TEG system. Its detailed procedure is tabulated in Table 2.
Implementation process of dynamic reconfiguration for TEG system using ACS
Figure 4 explicitly demonstrates the overall implementation procedure of dynamic reconfiguration for the TEG system using ACS. rand stands for a random value distributed in the interval of [0, 1]. It is worth noting that the initial arrangement of the TEG system needs to be included in the population during random initialization to ensure an effective optimization. Additionally, a basic precondition for this work is that the operating temperatures of each module in the TEG system can be collected by measuring devices.
Case studies
To validate the dynamic reconfiguration performance of ACS for the TEG systems under various HgTDs, two typical scenarios are developed, i.e., symmetrical (15 × 15) and asymmetrical configuration (20 × 15). Furthermore, four typical temperature distributions are studied for each configuration, including diagonal, outer, inner, and random [28]. Seven conventional MhAs, e.g., ABC [44], ACO [45], GA [28], PSO [28], SA [41], TS [38, 39], and POA [40], are used for comparison with ACS. For a fair comparison, NP and It are uniformly set as (NP = 30, It = 500) and (NP = 30, It = 1000) under symmetrical and asymmetrical scenarios, respectively. All algorithms are independently executed 50 times to obtain the statistical results, while all simulations are executed in the MATLAB/Simulink 2019b environment with a personal computer with Intel(R) Core(TM) i5 CPU at 2.9 GHz and 16 GB of RAM, and ode45 (DormandPrince) with variable step chosen as the solver.
Symmetrical configuration
Figures 5 and 6 illustrate the initial and optimal temperature distributions acquired by ACS in a symmetrical TEG system (15 × 15) under four different scenarios. Boxplots of the maximum output power obtained by eight algorithms under the four different scenarios of the (15 × 15) TEG system are shown in Fig. 7. This reveals the statistical distribution of the maximum output power for each algorithm in 50 runs. Clearly ACS can obtain the highest average value of the maximum output power of the TEG system. In particular, the distribution range of the maximum output power acquired from ACS is the smallest in the outer, inner, and random conditions. Thus, it can be seen that ACS has superior performance in terms of convergence stability and expectation of the maximum output power.
Figure 8 draws the convergence curves of various algorithms for the four different TEG systems, where ACS can converge within 350 iterations along with optimum output power under the four conditions. As depicted in Fig. 8b, c, although ABC has a faster convergence speed in the early iterations, it falls into a local optimum solution in the late stages, which further proves the superior global searching capability of ACS.
The powervoltage (P–V) and current–voltage (I–V) curves of the TEG systems before and after optimization using ACS are shown in Fig. 9. It can be seen that the potential GMPPs of the TEG systems are notably enhanced by 3.4%, 4.5%, 2.7%, and 22.8% via ACS under diagonal, outer, inner, and random scenarios, respectively. In addition, the total voltage difference between the different strings in the TEG system is significant when the temperature is regionally and intensively distributed as shown in Fig. 5. Hence, MLMPPs phenomena are inevitably induced because of the presence of series diodes [4, 22, 23]. More specifically, the voltage discrepancy among different strings will be decreased on account of uniform temperature distributions in the row direction obtained by ACS as in Fig. 6, and thus MLMPPs phenomena of P–V curves are significantly reduced, as shown in Fig. 9.
Asymmetrical configuration
To further validate the superiority of ACS under various HgTDs, an asymmetrical TEG configuration (20 × 15) is designed and comprehensively evaluated. The initial and optimized temperature distributions based on ACS are shown in Figs. 10 and 11, respectively. It can be easily seen that temperatures in Fig. 11 are uniformly distributed in the row direction under all tested HgTDs.
As delineated in Fig. 12, the distribution of the output power generated from 50 independent reconfigurations of the (20 × 15) TEG system by ACS is more centralized than others, where the acquired lower and upper bounds of output power are higher.
It is seen in Fig. 13 that faster convergence and higher output power are achieved by ACS. In contrast, other algorithms, particularly ACO, PSO, and POA, tend to suffer from worse reconfiguration solutions with inferior convergence rates. Based on the above analysis and discussion, it is shown that ACS can rapidly provide reconfiguration solutions for TEG systems with high desired output power.
Figure 14 shows the output characteristics of the TEG system under the four different scenarios. This indicates that potential GMPPs corresponding to the optimal solutions are respectively boosted by 164.0 W, 122.8 W, 34.7 W, and 47.4 W, compared to those of the initial solutions. The output characteristics are also significantly smoothed via ACS, which significantly alleviates the burden and complexity of MPPT.
Analysis and discussion for statistical results
Tables 3 and 4 provide statistical results obtained by the eight methods in 50 independent runs under the symmetrical and asymmetrical scenarios, respectively, where the optimal results are shown in bold. Although all methods can improve the output power of the TEG systems because of their original arrangements being used as the initial solutions of each method, ACS can harvest the highest output power and maximum average output power under all cases. In particular, in the symmetrical TEG system with random temperature distribution, the maximum output power of ACS is 103.1%, 116.3%, 114.3%, 110.7%, 108.4%, 104.7%, and 113.7% of that of ABC, ACO, GA, PSO, SA, TS, and POA, respectively. In addition, GMPPs of the optimized P–V curves in Fig. 14 are 16.9 W, 23.1 W, 18.7 W, and 19.8 W lower than the best results obtained by ACS in Table 4 under the four scenarios, respectively. This result is mainly caused by the forward conduction losses on the diodes.
As illustrated in Tables 3 and 4, both ACS and TS can obtain reconfiguration solutions with similar maximum output power under two HgTDs (i.e., outer and inner), which indicates that they have similar performance in these scenarios. However, as mentioned above, ACS is superior to TS in terms of convergence speed and stability, as displayed in Figs. 7, 8, 12, and 13. Overall, it is evident that ACS is successfully applied to the dynamic reconfiguration of the TEG systems with high accuracy, fast convergence speed, and excellent stability.
Table 5 tabulates the action module number reduced by EROS for the optimal solution of each algorithm in both symmetrical and asymmetrical scenarios. It is apparent that EROS can noticeably reduce the number of action modules in all cases. Hence, unnecessary switching actions and reconfiguration costs can be considerably reduced.
RTLAB based hardwareintheloop experiments
RTLAB platformbased hardwareintheloop experiments are carried out to further validate the hardware implementation feasibility, as shown in Fig. 15, with a sampling frequency of 1 kHz. Figures 16 and 17 demonstrate the output characteristics of each TEG system under various HgTDs optimized by ACS. It can be seen from Figs. 16 and 17 that the responses of the simulation and experiment acquired by the two configurations under several HgTDs are largely similar.
Conclusions
In this work, a novel ACS is designed to realize an effective and efficient dynamic reconfiguration of two different TEG systems, i.e., symmetrical and asymmetrical under various HgTDs. The main conclusions are:

1.
An MhAbased dynamic reconfiguration strategy for largescale TEG systems is presented and accomplished for the first time. It aims to reuse waste heat generated in daily life and industrial production.

2.
Compared with traditional MhAs, ACS is able to execute a reasonable equilibrium between global exploration and local exploitation via simultaneous execution of TS and discretized POA. Therefore, ACS can accurately and rapidly converge to optimal reconfiguration solutions for TEG systems under various HgTDs with a higher probability.

3.
Comprehensive case studies verify that ACS has a better performance in dynamic reconfiguration than other MhAs. In particular, the maximum output power of ACS is 103.1%, 116.3%, 114.3%, 110.7%, 108.4%, 104.7%, and 113.7% of that of ABC, ACO, GA, PSO, SA, TS, and POA under symmetrical scenario with random temperature distribution, respectively. The expected output power is enhanced by 34.6 W247.8 W via ACS compared with that of initial solutions under eight HgTDs, which further validates that ACS can significantly improve the potential GMPP of TEG systems.

4.
Statistical results show that EROS can significantly reduce the number of action modules required under various cases. This can boost the practicability and economy of TEG systems with low generation efficiency.
However, it is necessary to reasonably set the number of individuals and iterations of ACS according to the scale of the TEG system in practical application in order to obtain an acceptable computational cost. The completion of the dynamic reconfiguration of the TEG system relies highly on accurate and reliable temperature measurement equipment, which increases the cost. Therefore, the deployment of the thermometer and operator of ACS can be further optimized, while ACS can be extended to the reconfiguration of PV and PVTEG systems in future studies.
Availability of data and materials
The datasets used or analyzed during the current study are available from the corresponding author on reasonable request.
Abbreviations
 α _{sb} :

Seeback coefficient (V/K)
 α _{sb0} :

Constant portion of α_{sb} (V/K)
 α _{sb1} :

Variation rate of α_{sb} (V/K)
 β _{ n } _{,} _{ m } :

State of bypass diode in the nth row and the mth column
 γ _{ m } :

State of the mth series diode
 τ :

Thomson coefficient (V/K)
 it :

Current iterations of pelican optimization algorithm
 It :

Maximum iterations of pelican optimization algorithm
 I _{total} :

Total current of TEG system (A)
 M :

Number of columns in TEG system
 N :

Number of rows in TEG system
 NP :

Total number of individuals of each algorithm
 n _{np} :

Number of thermoelectric units
 P _{GMPP} :

Global maximum output power of TEG system (W)
 R _{TEG} :

Internal resistance of TEG system (Ω)
 R _{TEG,} _{ n } _{,} _{ m } :

Internal resistance of TEG module in the nth row and the mth column (Ω)
 R _{ m } :

Total internal resistance of the mth module (Ω)
 R _{total} :

Total resistance of TEG system (Ω)
 T _{hs} :

Temperature of hot side (K)
 T _{cs} :

Temperature of cold side (K)
 T _{av} :

Average value of T_{hs} and T_{cs} (K)
 T _{0} :

Temperature reference (K)
 V _{oc} :

Opencircuit voltage of TEG module (V)
 V _{oc,} _{ m } :

Total opencircuit voltage of the mth module (V)
 V _{oc,} _{ n } _{,} _{ m } :

Opencircuit voltage of TEG module in the nth row and the mth column (V)
 V _{t} :

Terminal voltage of TEG system (V)
 X :

Feasible solution space
 x :

A solution in X
 x _{ i } :

Status of the ith pelican
 x ^{*} :

Optimal solution
 ACS:

Adaptive coordinated seeker
 AOA:

Arithmetic optimization algorithm
 ABC:

Artificial bee colony
 ACO:

Ant colony optimization
 DA:

Dragonfly algorithm
 EO:

Equilibrium optimizer
 EROS:

Equivalent reoptimization strategy
 FASO:

Fast atom search optimizer
 GMPP:

Global maximum power point
 GA:

Genetic algorithm
 HgTD:

Heterogeneous temperature distribution
 INC:

Incremental conductance
 ICI:

Interacted collective intelligence
 MPPT:

Maximum power point tracking
 MLMPPs:

Multiple local maximum power points
 MhAs:

Metaheuristic algorithms
 P&O:

Perturb and observe
 PV:

Photovoltaic
 PSC:

Partial shading conditions
 POA:

Pelican optimization algorithm
 PSO:

Particle swarm optimization
 SP:

Series–parallel
 SA:

Simulated annealing
 TS:

Tabu search
 TEG:

Thermoelectric generation
References
Papież, M., Śmiech, S., & Frodyma, K. (2019). Effects of renewable energy sector development on electricity consumptionGrowth nexus in the European Union. Renewable and Sustainable Energy Reviews, 113, 109276.
Cheng, S., Wang, Y.Q., Liao, W.L., Zuo, X.W., & Dai, J. (2022). Bilevel multiobjective optimization of a new energy microgrid with electric vehicles. Power System Protection and Control, 50(12), 63–71. https://doi.org/10.19783/j.cnki.pspc.211149
Wang, Z., Li, L.L., Li, Z.H., Cheng, Z.Y., Yao, W.F., & Liu, R.S. (2021). The evolution characteristics of power grid frequency probability distribution. Power System Protection and Control, 49(20), 65–73. https://doi.org/10.19783/j.cnki.pspc.201654
Yang, B., Zhang, M.T., Zhang, X.S., Wang, J.B., Shu, H.C., Li, S.N., He, T.Y., Yang, L., & Yu, T. (2020). Fast atom search optimization based MPPT design of centralized thermoelectric generation system under heterogeneous temperature difference. Journal of Cleaner Production, 248, 119301.
Guo, Y.F., Gao, H.L., & Wu, Q.W. (2017). A combined reliability model of VSCHVDC connected offshore wind farms considering wind speed correlation. IEEE Transactions on Sustainable Energy, 8(4), 1637–1646.
Li, G.Q., Shittu, S., Diallo, T. M. O., Yu, M., Zhao, X.D., & Ji, J. (2018). A review of solar photovoltaicthermoelectric hybrid system for electricity generation. Energy, 158, 41–58.
Fu, X. (2022). Statistical machine learning model for capacitor planning considering uncertainties in photovoltaic power. Protection and Control of Modern Power Systems, 7, 5.
Abd ElKareem, A. H., Abd Elhameed, M., & Elkholy, M. M. (2021). Effective damping of local low frequency oscillations in power systems integrated with bulk PV generation. Protection and Control of Modern Power Systems, 6, 41.
Tohidi, F., Holagh, S. G., & Chitsaz, A. (2022). Thermoelectric generators: A comprehensive review of characteristics and applications. Applied Thermal Engineering, 201, 117793.
Kim, J., Baek, D., Ding, C.W., Lin, S., Shin, D., Lin, X., Wang, Y.Z., HooCho, Y., Park, S. H., & Chang, N. (2018). Dynamic reconfiguration of thermoelectric generators for vehicle radiators energy harvesting under locationdependent temperature variations. IEEE Transactions on Very Large Scale Integration (VLSI) Systems, 26(7), 1241–1253.
Catalan, L., Aranguren, P., Araiz, M., Perez, G., & Astrain, D. (2019). New opportunities for electricity generation in shallow hot dry rock fields: A study of thermoelectric generators with different heat exchangers. Energy Conversion and Management, 200, 112061.
Zhao, Y.L., Wang, S.X., Ge, M.H., Li, Y.Z., & Liang, Z.J. (2017). Analysis of thermoelectric generation characteristics of flue gas waste heat from natural gas boiler. Energy Conversion and Management, 148, 820–829.
Shen, Z.H., Ni, H., Ding, C., Sui, G.R., Jia, H.Z., Gao, X.M., & Wang, N. (2021). Improving the energyconversion efficiency of a PVTE system with an intelligent powertrack switching technique and efficient thermalmanagement scheme. IEEE Transactions on Components, Packaging and Manufacturing Technology, 11(6), 963–973.
Montecucco, A., Siviter, J., & Knox, A. R. (2017). Combined heat and power system for stoves with thermoelectric generators. Applied Energy, 185(Part 2), 1336–1342.
Shiotsu, Y., Seino, T., Kondo, T., & Sugahara, S. (2020). Modeling and design of thinfilm pitype micro thermoelectric generator using vacuum/insulatorhybrid isolation for selfpowered wearable devices. IEEE Transactions on Electron Devices, 67(9), 3834–3842.
Chen, J., Klein, J., Wu, Y.J., Xing, S.X., Flammang, R., Heibel, M., & Zuo, L. (2016). A thermoelectric energy harvesting system for powering wireless sensors in nuclear power plants. IEEE Transactions on Nuclear Science, 63(5), 2738–2746.
Yang, B., Zhang, M.T., Wang, J.B., Zeng, K.D., Zhang, Z.A., Shu, H.C., Zhang, X.S., & Yu, T. (2021). Interacted collective intelligence based energy harvesting of centralized thermoelectric generation systems under nonuniform temperature gradient. Sustainable Energy Technologies and Assessments, 48, 101600.
Liu, Y.H., Chiu, Y.H., Huang, J.W., & Wang, S.C. (2016). A novel maximum power point tracker for thermoelectric generation system. Renewable Energy, 97, 306–318.
Zhang, X.S., Yang, B., Yu, T., & Jiang, L. (2020). Dynamic surrogate model based optimization for mppt of centralized thermoelectric generation systems under heterogeneous temperature difference. IEEE Transactions on Energy Conversion, 35(2), 966–976.
Bijukumar, B., Raam, A. G. K., Ganesan, S. I., & Nagamani, C. (2018). A linear extrapolationbased MPPT algorithm for thermoelectric generators under dynamically varying temperature conditions. IEEE Transactions on Energy Conversion, 33(4), 1641–1649.
Twaha, S., Zhu, J., Yan, Y.Y., Li, B., & Huang, K. (2017). Performance analysis of thermoelectric generator using dc–dc converter with incremental conductance based maximum power point tracking. Energy for Sustainable Development, 37, 86–98.
Mansoor, M., Mirza, A. F., Duan, S.H., Zhu, J., Yin, B.Q., & Ling, Q. (2021). Maximum energy harvesting of centralized thermoelectric power generation systems with nonuniform temperature distribution based on novel equilibrium optimizer. Energy Conversion and Management, 246, 114694.
Yang, B., Wang, J.T., Zhang, X.S., Zhang, M.T., Shu, H.C., Li, S.N., He, T.Y., Yang, L., & Yu, T. (2019). MPPT design of centralized thermoelectric generation system using adaptive compass search under nonuniform temperature distribution condition. Energy Conversion and Management, 199, 111991.
Mirza, A. F., Mansoor, M., Zerbakht, K., Javed, M. Y., Zafar, M. H., & Khan, N. M. (2021). Highefficiency hybrid PVTEG system with intelligent control to harvest maximum energy under various nonstatic operating conditions. Journal of Cleaner Production, 320, 128643.
Li, F.S., Lin, D., Yu, T., Li, J.W., Wang, K.Y., Zhang, X. S., Yang, B., & Wu, Y.F. (2021). Adaptive rapid neural optimization: A datadriven approach to MPPT for centralized TEG systems. Electric Power Systems Research, 199, 107426.
Chen, M. (2014). Adaptive removal and revival of underheated thermoelectric generation modules. IEEE Transactions on Industrial Electronics, 61(11), 6100–6107.
Belhachat, F., & Larbes, C. (2021). PV array reconfiguration techniques for maximum power optimization under partial shading conditions: A review. Solar Energy, 230, 558–582.
Yang, B., Ye, H.Y., Wang, J.B., Li, J.L., Wu, S.C., Li, Y.L., Shu, H.C., Ren, Y.X., & Ye, H. (2021). PV arrays reconfiguration for partial shading mitigation: Recent advances, challenges and perspectives. Energy Conversion and Management, 247, 114738.
Aljafari, B., Satpathy, P. R., & Thanikanti, S. B. (2022). Partial shading mitigation in PV arrays through dragonfly algorithm based dynamic reconfiguration. Energy, 257, 124795. https://doi.org/10.1016/j.energy.2022.124795
Pachauri, R. K., Thanikanti, S. B., Bai, J., Yadav, V. K., Aljafari, B., Ghosh, S., & Alhelou, H. H. (2022). Ancient Chinese magic squarebased PV array reconfiguration methodology to reduce power loss under partial shading conditions. Energy Conversion and Management, 253, 115148.
Reddy, S. S., & Yammani, C. (2020). A novel MagicSquare puzzle based onetime PV reconfiguration technique to mitigate mismatch power loss under various partial shading conditions. Optik, 222, 165289.
Zhang, X.S., Li, C.Z., Li, Z.L., Yin, X.Q., Yang, B., Gan, L.X., & Yu, T. (2021). Optimal mileagebased PV array reconfiguration using swarm reinforcement learning. Energy Conversion and Management, 232, 113892.
Chen, M. (2014). Reconfiguration of sustainable thermoelectric generation using wireless sensor network. IEEE Transactions on Industrial Electronics, 61(6), 2776–2783.
Zhang, X.S., Tan, T., Yang, B., Wang, J.B., Li, S.N., He, T.Y., Yang, L., Yu, T., & Sun, L.M. (2020). Greedy search based datadriven algorithm of centralized thermoelectric generation system under nonuniform temperature distribution. Applied Energy, 260, 114232.
Chakraborty, A., Saha, B. B., Koyama, S., & Ng, K. C. (2006). Thermodynamic modelling of a solid state thermoelectric cooling device: Temperatureentropy analysis. International Journal of Heat and Mass Transfer, 49(19–20), 3547–3554.
Han, H. S., Kim, Y. H., Kim, S. Y., Um, S., & Hyun, J. M. (2010). Performance measurement and analysis of a thermoelectric power generator. In IEEE intersociety conference on thermal thermomechanical phenomena electronics systems, Las Vegas, NV, USA, pp. 1–7.
Kong, C.S. (1995). A general maximum power transfer theorem. IEEE Transactions on Education, 38(3), 296–298.
Glover, F. (1989). Tabu searchpart I. INFORMS Journal on Computing, 1(3), 190–206.
Glover, F. (1990). Tabu searchpart II. INFORMS Journal on Computing, 2(1), 4–32.
Trojovský, P., & Mohammad, D. (2022). Pelican optimization algorithm: A novel natureinspired algorithm for engineering applications. Sensors, 22(3), 855.
Ding, H., Sun, G.Y., Hao, L.J., Wu, B., & Wu, Y.C. (2020). A loading pattern optimization method based on discrete differential evolution. Annals of Nuclear Energy, 137, 107057.
Zhang, Y.J., Li, T., Na, G.Y., Li, G.Q., & Li, Y. (2015). Optimized extreme learning machine for power system transient stability prediction using synchrophasors. Mathematical Problems in Engineering, 2015, 529724.
Gu, X.P., Li, Y., & Jia, J.H. (2015). Feature selection for transient stability assessment based on kernelized fuzzy rough sets and memetic algorithm. International Journal of Electrical Power & Energy Systems, 64, 664–670.
Karaboga, D., & Ozturk, C. (2011). A novel clustering approach: Artificial bee colony (ABC) algorithm. Applied Soft Computing, 11(1), 652–657.
Dorigo, M., Birattari, M., & Stutzle, T. (2006). Ant colony optimization. IEEE Computational Intelligence Magazine, 1(4), 28–39.
Acknowledgements
The authors appreciatively acknowledge the support of by National Natural Science Foundation of China (61963020).
Funding
National Natural Science Foundation of China (61963020).
Author information
Authors and Affiliations
Contributions
YC: Conceptualization, methodology, writing, and software; BY: Writingreview and editing; ZG: Investigation, and visualization; JW: Supervision, writingreview, and editing; MZ: Writingreviewing, editing, and methodology; ZL: Experiment and software; TY: Supervision, project administration, resources, and data curation. All authors read and approved the final manuscript.
Authors' Information
Yijun Chen (1996–), male, Master Degree Candidate, Major in the renewable energy optimization and control.
Bo Yang (1988–), male, PHD and Professor, Major in the intelligent optimization and control of renewable energy and energy storage system.
Zhengxun Guo (1998–), male, Master Degree Candidate, Major in the renewable energy optimization and control.
Jingbo Wang (1996–), male, Master Degree, Major in the optimization operation of photovoltaic system.
Mengmeng Zhu (1986–), male, PHD Candidate and Engineer, Major in the voltage and current transformers.
Zilin Li (1988–), male, PHD, Major in the electric power electronic power system stability and operation.
Tao Yu (1974–), PHD and Professor, Major in the nonlinear control theory, optimization for complex power systems, and machine learning.
Corresponding author
Ethics declarations
Competing interests
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Chen, Y., Yang, B., Guo, Z. et al. Dynamic reconfiguration for TEG systems under heterogeneous temperature distribution via adaptive coordinated seeker. Prot Control Mod Power Syst 7, 38 (2022). https://doi.org/10.1186/s41601022002596
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s41601022002596
Keywords
 Thermoelectric generation systems
 Dynamic reconfiguration
 Heterogeneous temperature distribution
 Adaptive coordinated seeker