2018 15th International Multi-Conference on Systems, Signals & Devices (SSD)

A Fuzzy Genetic Programming-Based Algorithm for

Eco-friendly and Economic Load Dispatch Problem
Abdellah DERGHAL Noureddine GOLEA
LGEA Laboratory, Oum el Bouaghi University, LGEA Laboratory, Oum el Bouaghi University,
04000, Oum el Bouaghi, Algeria 04000, Oum el Bouaghi, Algeria
[email protected] [email protected]

Abstract— In this paper, we attempt to apply genetic In the standard Eco-friendly and economic dispatch
algorithms to the fuzzy mathematical programming problem, operating parameters are assumed to be deterministic,
problems which involve imprecise (fuzzy) and nonlinear but in real world EED problems, input data or related
information. The principle objective in this paper is how to parameters, such as market demand, capacity, and relevant
attribute a fuzzy set in the building of the Environmental operating costs, frequently are fuzzy owing to some
economic power dispatch problem. The Eco-friendly information being incomplete or unobtainable [4]. As example
/Economic Load Dispatch problem is formulated as a of constraints/objectives with Fuzziness we have: acceptable
multiple objective problem subject to physical constraints, security risk, assessment of customer satisfaction, economic
objectives, environmental objectives, equipment loading limits,
Fuzzy mathematical programming is used to represent
normal operational limits, etc... Because of not enough
objective functions with fuzzy parameters and
information, none of these constraints or objectives is well
uncertainties in constraints set, and genetic algorithm defined. When deterministic constraints are not satisfied in the
(GA) is used to solve the reformulated problem. The conventional mathematical programming, it is difficult to learn
performance of this solution method is examined by what kind of constraints is critical to what extent [4-5].
comparing its results with that of the existing methods
through an illustrative example, these comparisons reveal The purpose of this work is to develop a solution for eco-
the efficient and robustness of the planned approach friendly/economic load dispatch (EELD) problem based on
developed in this paper. fuzzy sets formulation, taking into account the design
parameters uncertainties. The objective and constraints are
Keywords- Environmental Economic Dispatch (EED); Fuzzy modeled as fuzzy DM problem involving the minimization of
Membership Function; Multi-objective Optimization; Genetic the objective function with fuzzy parameters and uncertainties
Algorithm; Weighted Approach; Max-Min. in a set of constraints. Then, genetic algorithm is used to get a
feasible solution of the fuzzy mathematical programming
I. INTRODUCTION problem. The performance of this solution approach is
evaluated by comparing its results with that of their existing
The last decade have proof that Environmental/Economic methods.
Dispatch (EED) is really an interesting problem, in particular,
decreasing fuel cost, in order to obtain the best compromise II. ECO-FRIENDLY / ECONOMIC LOAD DISPATCH (EELD)
between the fuel cost and emission as well as the feasible PROBLEM STATEMENT
domain, to achieve the best compromise solution [1-4]. EED
calculates the cost of generation based on data relating fuel The combined economic and emission dispatch problem
cost and power output. This cost function is approximated by a can be formulated as a constrained optimization problem [1, 6]
quadratic equation with cost coefficients. composed by objective functions to be minimized under
various operating constraints.
In conventional economic load dispatch the coefficients are
assumed to be deterministic, but in real-world situations, these A. Objective functions
data are subjected to inaccuracies and uncertainties [5].
The atmospheric pollutants such as sulphur oxides SOx
Eco_friendly and economic power dispatch is a nonlinear,
and nitrogen oxides NOx caused by fossil-fuelled thermal
non-convex, large-scale, static optimization problem, with
units can be model separately [7]. However, total ton emission
both continuous and discrete control variables. Were the main
can be expressed as:
objective is to determine the optimal schedule of online
generating units so as to meet the demand power at minimum
operating cost under various system and operating constraints. ( )
C i Pg = gi + bi Pg + ai Pg2
i i i
i = 1, 2,..., n (1)

The constraints include inequality ones which are the limits Where n is the generators number; Where gi , bi , ai are
of control variables and state variables; and equality ones
which are the power flow equations. the coefficient of generating unit i .

978-1-5386-5305-0/18/$31.00 ©2018 IEEE 738

Authorized licensed use limited to: ULAKBIM UASL - Dumlupinar University. Downloaded on November 23,2022 at 18:29:56 UTC from IEEE Xplore. Restrictions apply.
 2018 15th International Multi-Conference on Systems, Signals & Devices (SSD)

The fuel cost can be expressed as a summation of all control, security assessment, stabilization control (PSS) and
generating units’ operating cost function: unit commitment. A large number of research articles
appeared recently indicate applicability of fuzzy systems to
fi (Pg ) = ai + bi Pg + ci Pg2 i = 1,..., n (2) power systems for wider operating conditions under
i i i
uncertainties [8, 9].
Where ai, bi and ci are the ith generator operating cost The literature contains many proposed formulations, but
coefficients. few have been implemented in practice. To deal with this
problem many papers have been published. Before discussing
B. Equality and Inequality Constraints the details of application of fuzzy sets in power flow models,
The environmental/economic dispatch problem has to take some theoretic background is given in [10-11-12-13]; all of
into account the equality and inequality constraints. these papers have made interesting achievements on the
subject.
• Power Balance Constraint
The purpose of this work is to develop a solution for multi-
n objective EED problem based on fuzzy sets formulation, taking
åPg = PD + Ploss
i
(3) into account the design parameters uncertainties.
i =1

Where PD is the total power demand, Ploss is the total IV. GENETIC ALGORITHM IN ENVIRONMENTAL/ECONOMIC
DISPATCH (EED)
transmission network loss given by:
The GAs (genetic algorithms) have been applied to solve
Ploss = P T BP (4) unit commitment problem [14], Optimal reactive power
dispatch (ORPD) problem [15, 16] and for economic load
Where B is the transmission loss coefficients matrix. dispatch problem [17, 18]. Miranda et al. [19] have provided a
survey of three branches of evolutionary programming (EP),
• Generating Capacity Constraints genetic algorithms (GAs) and discusses their relative merits
Generation power of each generation unit is restricted by and demerits.
their lower and upper limits as follows:
The Genetic Algorithm Optimal Power Flow (GAOPF)
Pg min
£ Pg £ Pg , i = 1, 2,..., n max
(5) problem solved by Yuryevich et al., [20], based on the use of a
i i i genetic algorithm load flow, and to accelerate the concepts it
min propose the use of gradient information by the use of the
Where: Pgi and Pgimax are the unit i minimum and steepest decent method. The method is not sensitive to the
maximum generating limits, respectively. starting points and capable to determining the global optimum
solution to the OPF for range of constraints and objective
Where B is the transmission loss coefficients matrix.
functions. But GAOPF requires two load flow to be performed
per individual, per iteration because all controllable variables
• Ramp-Rate Limits Constraint are included in the fitness.
max(Pgmin , Pg (t - 1) - DRi ) £ Pg (t ) £ The combination of Genetic algorithm and fuzzy set theory
i i i
(6)
min(P max
gi
, Pg (t - 1) + URi ) will overcome the conflict among the objectives and the
i
problems of determining the aspiration levels via GA, and at
the same time fuzzy set theory will treat the vagueness of the
Where URi and DRi are the i th unit up and down ramp
extracted information from the DM. Here, the link between the
limits (MW/time period), respectively. two approaches is performed through implementation of the
aspiration levels. In this study, the basic GA has been
III. FUZZY SETS IN THE UNCERTAINTY ECO- developed in order to make it suitable for solving fuzzy
FRIENDLY/ECONOMIC LOAD DISPATCH PROBLEM mathematical nonlinear constrained optimization problem.
A large class of problems within power system planning
and operation is decision-making (DM) and multiobjective V. FUZZY EED PROBLEM STATEMENT
optimization, which includes transmission planning, security In the classical multi-objective decision making model, all
analysis, optimal load flow, state estimation, and unit data and information are assumed to be absolutely accurate,
commitment, among others. and the objectives and constraints are all assumed to be well
Fuzzy set theory applications have received increasing expressed by mathematical formations. However, it is difficult
attention in various areas of power systems such as: to clearly describe the objective functions and constraints by
contingency analysis, monitoring, distribution planning, load mathematical equations for many realistic problems, thus, the
frequency control, maintenance scheduling, load flow multi-objective decision making model with certain
computations, forecasting, management, reactive power parameters cannot deal with all real-life problems. So the

739

Authorized licensed use limited to: ULAKBIM UASL - Dumlupinar University. Downloaded on November 23,2022 at 18:29:56 UTC from IEEE Xplore. Restrictions apply.
 2018 15th International Multi-Conference on Systems, Signals & Devices (SSD)

uncertain multiobjective decision making models are ìï

ïï 0 if fi ³ fi max
proposed. In 1965, after L.A. Zadeh [21] proposed the fuzzy
ïï f - fi
set theory, it was rapidly and widely applied in the filed of
operations, management science, control theory and so on. In mf Pg
i
( )
i
= ïí
i max i

ïï fi max - fi min
if fi min < fi < fi max (8)
1970, R. E. Bellman and L. A. Zadeh [22] collaborated to ïï
propose the fuzzy decision making model based on multi- ïï 1 if fi £ fi min
î
objective programming. That’s the original work which is the
basis of this monograph. Where fi min and fi max are the ith objective function
The following is a fuzzy formulation for the minimum and maximum values, respectively.
environmental/economic load dispatch problem.
μfi (Pg )
1
A. Uncertainty Power Load Demand
For load demand, it is assumed that all load demands are
fuzzy in nature, and can be expressed by triangular fuzzy
numbers. As shown in fig. 1, Pdd will never be below Pd - L
fi
or above Pd + L .
0
fi min fi max
m(Pdd )
Figure 2. Fuzzy objective membership function.
1
VI. IMPLEMENTATION OF THE SUGGESTED METHOD TO
SOLVE THE EED PROBLEM
The suggested method consists of seven steps (all of them
identified a sub problem): The step by step algorithm for fuzzy
Pdd multi objective EED with power losses is as follows:
0
Pd -L Pd Pd +L Step 1: Specify the lower and upper bound generation power
f f
of each unit, calculate min and max for all objective functions
Figure 1. Membership function for load demand. to set the initial value of reference membership functions
levels.
The best estimate of Pdd is Pd . In the present study L is Step 2: The DM subjectively defines the linear membership
taken as 5% of the forecasted peak loads. functions.
The degree of satisfaction can be defined as: Step 3: Formulate the problem with fuzzy modeling to create a
fuzzy multi objective mathematical programming.
ìï0, Pdd £ Pd - L & Pdd ³ Pd + L
ïï
ïïP - P Step 4: Introduce the genetic algorithms into the fuzzy
ïï dd d
, Pd - L < Pdd £ Pdd mathematical programming problem.
ï (7)
m (Pdd ) = í L
ïï1, Pd = Pdd The basic procedure of GA is summarized in [23].
ïï
ïïPdd - Pd Step 5: Solve the fuzzy multi objective EED problem using
ïï , Pd < Pdd £ Pd + L
î L GA based algorithm with ramp-rate limits and with
Where, m (Pd d ) is the membership function with respect to considering of the prohibited zones to obtain a feasible
solution that satisfies the power balance equality constraints.
th
the i constraint and Pd  L is the maximum (minimum) If the decision maker satisfies the present solution, end
tolerable error in the forecasted peak loads. here, otherwise, proceed to step 6.
Step 6: Update the reference membership function values
B. Fuzzy Objective Function
based on the current values of objective functions. Inside the
We can introduce the membership function that represent program, the membership function is updated by using the
the fuel cost objective function (fig. 2) as follows: information given by the DM [24], and go to step 7.
Step 7: Obtain the power flow solution with update
membership functions. It means: Solve the fuzzy EED
problem to get the optimum dispatch for all the units. If the
solution satisfies the termination conditions and DM accepts

740

Authorized licensed use limited to: ULAKBIM UASL - Dumlupinar University. Downloaded on November 23,2022 at 18:29:56 UTC from IEEE Xplore. Restrictions apply.
 2018 15th International Multi-Conference on Systems, Signals & Devices (SSD)

it, then stop the calculation process, and the solution is Until the preferred compromise solution is obtained. The
determined to be a satisfactory solution. Otherwise, return to DMs may reject the solution or determine a new upper and
step 6. lower bounds based on the current solution to update the
membership function and aspiration levels to resolve.
Start
The diagram (flowchart) block of these step by step is
shown in fig.3.
Input Power System information
VII. TEST SYSTEM
The proposed algorithm is also tested on a 6-unit system.
Fig. 4 shows the test system for the study case. The system has
Modelling the membership functions
three plants and six generating units. The fuel cost equations,
emission equations and the transmission loss coefficients of
Calculate objective fonctions fmin and fmax the generating units are given in Table II and III respectively.
Only one type of pollutant (NOx) is considered for simplicity.
For the test system, the parameter settings to execute GA-
FMOPF are probability of crossover = 0.5, probability of
Specify the degree of membership for each mutation = 0.7 and population size = 60.
objective and constraint

TABLE I. PARAMETER SETTINGS OF GENETIC ALGORITHM (GA) USED

Select the best FOR COMPUTATION

compromise solution
Update membership functions
using Genetic Control parameters Values
Algorithm (GA) Population size 60
Probability of mutation 0.7
Is the solution acceptable ?
Crossover 0.5
Satisfy the Decision Maker’s
Optimal Solution
NO
TABLE II. FUEL COST AND NOX EMISSION COEFFICIENTS AND
YES OPERATING LIMITS.

Report the Dispatch Results

Unit ai bi ci gi bi ai Pimin Pimax
1 756.79886 38.53973 0.15247 13.85932 0.32767 0.004119 10 125
End
2 451.32513 46.15916 0.10587 13.85932 0.32767 0.004119 10 150

3 1049.99770 40.39655 0.02803 40.26690 -0.54551 0.00683 35 225

Figure 3. Overall Flow Chart of the Developed Method.
4 1243.53110 38.30553 0.03546 40.26690 -0.54551 0.00683 35 210
The condition in step by step algorithm is: the power
outputs of all the units are within limits, fall out of the 5 1658.56960 36.32782 0.02111 42.89553 -0.51116 0.00461 130 325
prohibited operating zones and the optimal dispatch is
6 1356.65920 38.27041 0.01799 42.89553 -0.51116 0.00461 125 315
obtained for given total system demand.
In Step 5, when the power system operator made the
judgment that it cannot be achieved because the present no
inferior solution and the target point are too far apart. We
remember that higher the membership function value better is
the solution. It is interesting to note that when the membership
function increases the degree of satisfaction will be improved.
It’s naturally to assume that the DM may have fuzzy or
imprecise goals for each objective function.

741

Authorized licensed use limited to: ULAKBIM UASL - Dumlupinar University. Downloaded on November 23,2022 at 18:29:56 UTC from IEEE Xplore. Restrictions apply.
 2018 15th International Multi-Conference on Systems, Signals & Devices (SSD)

TABLE III. LOSS B-COEFFICIENTS TABLE IV. BEST OPTIMAL SOLUTION FOR STUDY CASE (PD=700MW)

0.002022 -0.000286 -0.000534 -0.000565 -0.000454 -0.000103

Fuzzy Weighted
-0.000286 0.003243 0.00016 -0.000307 -0.000422 -0.000147 Unit power output, minmax method
GA-FMOPF FCGA NSGA-II
MW
-0.000533 0.000016 0.002085 0.000831 0.000023 -0.000270 Case I Case II

-0.000565 -0.000307 0.000831 0.001129 0.000113 -0.000295 P1 95.1642 80.16 86.286 85.924 89.098
-0.000454 -0.000422 0.000023 0.000113 0.000460 -0.000153 29.6395 53.71 60.288 60.963 65.475
P2
0.000103 -0.000147 -0.000270 -0.000925 -0.000153 0.000898
P3 83.2099 40.93 73.064 53.909 69.713

P4 83.8974 116.23 109.036 107.124 116.079

VIII. DISCUSSION
P5 158.4641 251.20 223.448 250.503 223.596
To implement GA to unravel the uncertainty in the
environmental/economic load dispatch problem using the 235.2981 190.62 184.111 176.504 175.709
proposed method codes in C++ and Matlab 7.8 are P6
developed and executed on a Processor Core (TM) i3 CPU. Total power, MW 685.6732* 732.85 736.234 734.9270 739.6700
The proposed method is applied for the EED problem with 46.1573 32.85* 36.234 34.927 39.669
load demand of 700 MW, and it is compared with FCGA and P loss, MW
NSGA-II [25, 26] and weighted minimax technique simulation Cost, $/h 36332.95* 38408.82 38671.813 39070.74 39010.74
[27], Table IV presents a comparison of the generation cost
and emission. The best cost and best emission solutions Emission Kg/h 450.490* 527.46 484.931 528.447 493.977
obtained out of 20 runs with the proposed method are given
also in Table IV.
4
We try to get some numerical tests which have multi- x10
10 2000
objective functions, and imprecise nonlinear information, in
order to demonstrate the effectiveness of the proposed method;
we study the performance in executing the optimization
functions for this case system.

E m is s io n
Plant 1
G1 G2 G3 Fuel Cost
Fuel C ost

Emission

1000

G4

Plant 2

0
G5 0 50 100 150 200 250
Iteration
Plant 3

Figure 5. The GA-FMOPF algorithm’s performance for study case,

G6

Figure 4. Single-line diagram of case 2 test system

742

Authorized licensed use limited to: ULAKBIM UASL - Dumlupinar University. Downloaded on November 23,2022 at 18:29:56 UTC from IEEE Xplore. Restrictions apply.
 2018 15th International Multi-Conference on Systems, Signals & Devices (SSD)

The problem has been formulated as multi-objective [8] R. C. Bansal, “Biogeography on the fuzzy set theory applications in
power systems (1994-2001),” IEEE Transactions on Power Systems,
optimization problem with competing fuel cost and vol. 18, No. 4 , pp. 1291-1299. 2003.
environmental impact objectives. [9] A. Derghal, N. Goléa, Multi-Objective Generation Scheduling Using
It is clearly shown that fuzzy genetic programming-based Genetic-Based Fuzzy Mathematical Programming Technique. In I.
Management Association (Ed.), Nature-Inspired Computing: Concepts,
algorithm obtains the best cost and best emission. It is quite Methodologies, Tools, and Applications, Hershey, PA: IGI Global, pp.
evident that the suggested approach yields satisfactory 1131-1160, 2017.
solutions. The effectiveness of the proposed method for EED [10] B. Borkowska, “Probabilistic load flow,” IEEE Trans. Power Apparatus
problem demonstrated with load uncertainty and System, vol. PAS-93, No.3, pp. 752-755, 1974.
[11] J. F. Dopazo, O. A. Klitin, and A. M Sasson, “Stochastic load flow,”
The results also show the superiority of the proposed IEEE Trans. Power Apparatus and Systems, vol. PAS-94, No. 2, pp.
approach and confirm its potential to solve the multi-objective 299-309, 1975.
EED problem with much less time (Only 5 to 7 iterations are [12] F. Wang, and L. Alvarado, “Interval arithmetic in power flow analysis,”
IEEE Trans. Power Systems, vol. 7, No. 3, pp. 1341-1349, 1992.
needed to reach the desired tolerance) and the results are quite
[13] V. Miranda, and J. T. Saraiva, “Fuzzy modeling of power system
accurate comparing to other existing methods. optimal load flow,” IEEE Trans. Power Systems, vol. 7, No. 2, pp. 843-
849, 1992.
IX. CONCLUSION [14] S. A. Kazarlis, A. G. Bakirtzis, and V. Petridis, “A genetic algorithm
solution to the unit commitment problem. IEEE Transactions on Power
In this paper, Genetic algorithms solve a Fuzzy Systems, vol. 11(1), pp. 83–92, 1996.
mathematical programming problem with respect to both the [15] S. N. Singh, A. Chandramouli, P. K. Kalra, S. C. Srivastava, and D. K.
accuracy and the precision of solutions. Mishra, “Optimal reactive power dispatch using genetic algorithms.
Proceedings of the International Symposium on Electricity Distribution
Application of Genetic approach to fuzzy Multi objective and Energy Management, Singapore, pp. 464–469, 1993.
EED problem has been explored and tested. However, this [16] K. S. Swarup, M. Yoshimi, S. Shimano, and Y. Izni, “Optimization
approach could also be applied in other method based on methods using genetic algorithms for reactive power planning in power
systems, Proceedings of 12th PSCC, Dresden, Germeny, vol. 1, pp. 483–
similar principles as PSO and evolutionary strategy. 491, 1996.
Moreover, the proposed method provides the facility to [17] D. C. Walters, and G. B. Sheble, “Genetic algorithm solution of
consider the uncertainties in the multi objective EED formula. economic dispatch with valve point loading. IEEE Transactions on
Power Systems, vol. 8(3), pp. 1325– 1332, 1993.
The results also show the superiority of the proposed [18] G. B. Sheble, and K. Britigg, “Refined genetic algorithm-economic
approach and confirm its potential to solve the multiobjective dispatch example. IEEE Transactions on Power Systems, vol. 10(1), pp.
EED problem. So the method is capable of representing 117–124, 1995.
practical situations in power system operation where the limits [19] V. Miranda, D. Srinivasan, and L. M. Proenca, “Evolutionary
on specific variables are soft and the small violations of these computation in power systems, A Survey Paper, Proceedings of 12th
PSCC, Dresden, Germany , pp. 25–40, 19–23 August 1996.
limits may be tolerable.
[20] J. Yuryevich, K. P. Wong, “Evolutionary programming based optimal
power flow algorithm,’’ IEEE transactions on Power Systems, vol. 14,
REFERENCES pp. 1245-1250, 1999
[1] S. C. Parti, D. P. Kothari, and P. V. Gupta, “Economic thermal power [21] L. A. Zadeh, “Fuzzy sets,” Inform. Control Journal, vol. 8, pp. 338-353,
dispatch,” Institution of Engineers (India), Journal-EL, vol. 64, pp. 126- 1965.
132, 1983. [22] R. Belmann, and L. A. Zadeh, “Decision making in a fuzzy
[2] C. H. Fujisawa, M. F. Carvalho, A. T. Azevedo, S. Soares and T. Ohishi, environment. Management Science, vol. 17B , pp. 141-164, 1970.
“Optimal power flow models using network flow method,” Sixth [23] A. Derghal, N. Golea, “Genetic algorithm to solving large practical
IEEE/PES Transmission and Distribution: Latin America Conference fuzzy economic load dispatch with prohibited operating zones”, The 2nd
and Exposition (T&D-LA), vol. 1, pp.1-3, Montevideo, Uruguai, 2012 IEEE International Renewable and Sustainable Energy Conference.
[3] S. R. Naarayan, “Optimization principles practical applications to the IRSEC’14 , Ouarzazate, Morocco – October 17-19, 2014.
operation and markets of the electric power industry,” IEEE Press Series [24] J. Kubokawa, Y. Okubo, H. Sasaki, and R. Yokoyama, “A study of an
on Power Engineering, 2003. interactive fuzzy multiobjective optimal power flow. Electrical
[4] F. D. Galiana, and A. J. Conejo, “Economics of Electricity Generation, Engineering in Japan, vol. 130 (1), 2000.
Electricity Energy Systems: Analysis and Operation,” Taylor & Francis [25] H. G. S. Rughooputh, R. T. F. Ah King, “Environmental/economic
group, LLC, 2009. dispatch of thermal units using an elitist multiobjective evolutionary
[5] D. P. Kothari, and K. P. Singh Parmar, “A novel approach for Eco- algorithm” IEEE International Conference in Industrial Technology, vol.
friendly and economic power dispatch using Matlab,” International 1, pp. 48–53 , 2003.
Conference on Power Electronics, Drives and Energy Systems [26] B. H. Dinh, T. T. Nguyen, New Solutions to Modify the Differential
(PEDES’06), pp. 1-6, 2006. Evolution Method for Multi-objective Load Dispatch Problem
[6] S. Espinosa, D. Arias Cazco and M. Yanez Salcedo, “Economic Considering Quadratic Fuel Cost Function. In Recent Advances in
Dispatch hydrothermal system with CO2 Emissions Constraints,” IEEE Electrical Engineering and Related Sciences. Lecture Notes in Electrical
Latin America Transactions, vol. 15, no. 11, pp. 2090-2096, Nov. 2017. Engineering, Springer, vol. 415, 2017.
[7] V. N. Dieu, and W. Ongsakul, “Hopfield lagrange network for economic [27] J. S. Dhillon, S. C. Parti, and D. P. Kothari, “Stochastic economic
load dispatch. In P. Vasant, N. Barsoum, & J. Webb (Eds.), Innovation emission load dispatch. Electric Power Systems Research, vol. 26,
in Power, Control, and Optimization: Emerging Energy Technologies pp.179–186. 1993.
Hershey, PA: Engineering Science Reference, pp. 57-94. 2012.

743

Authorized licensed use limited to: ULAKBIM UASL - Dumlupinar University. Downloaded on November 23,2022 at 18:29:56 UTC from IEEE Xplore. Restrictions apply.