- Email: [email protected]
Dynamic voltage restorer-based power quality optimization using differential evolution algorithm
Available online at www.sciencedirect.com
ScienceDirect ScienceDirect
Procedia Manufacturing 00 (2019) 000–000 Procedia Manufacturing 00 (2019) 000–000
Available online at www.sciencedirect.com
ScienceDirect
www.elsevier.com/locate/procedia www.elsevier.com/locate/procedia
Procedia Manufacturing 35 (2019) 573–579
2nd International Conference on Sustainable Materials Processing and Manufacturing 2nd International Conference on Sustainable Materials Processing and Manufacturing (SMPM 2019) (SMPM 2019)
Dynamic voltage restorer-based power quality optimization using Dynamic voltage restorer-based power quality optimization using differential evolution algorithm differential evolution algorithm John Savecaa, Zenghui Wanga*, Yanxia Sunb John Savecaa, Zenghui Wanga*, Yanxia Sunb
a Department of Electrical and Mining Engineering University of South Africa, Johannesburg 1710, South Africa a Department of Electrical and Electronic Engineering Science University of Johannesburg, Johannesburg 2006, Africa South Africa Department of Electrical and Mining Engineering University of South Africa, Johannesburg 1710, South b Department of Electrical and Electronic Engineering Science University of Johannesburg, Johannesburg 2006, South Africa b
Abstract Abstract Homes and industrial companies including electric utilities are much concerned about the study of power quality and ways to control it. is due to equipmentincluding and newelectric household’s technologies even small changes supply Homes andThis industrial companies utilities are much becoming concerned more aboutsensitive the studyto of power quality andofways to voltages.it.Therefore power quality isand onenew of the most important aspects becoming in smart grid’s of becoming a reliable andofefficient control This is due to equipment household’s technologies morevision sensitive to even small changes supply new generation power grid.quality This paper to develop a strategy overcomes poor power quality ainreliable smart grid using voltages. Therefore power is oneattempts of the most important aspects that in smart grid’s vision of becoming and by efficient Dynamic Voltage Restorer due toattempts its efficient compensation of that voltage sags andpoor swells. Employing improved Differential new generation power grid.(DVR) This paper to develop a strategy overcomes power quality in smart grid by using Evolution (DE) based on Multi-Objective Parallel Operation (MOPO) to itssags fastand convergence and ability to give high rank to Dynamic Voltage Restorer (DVR) due to its efficient compensation of due voltage swells. Employing improved Differential the population to hyper-volume defined by the (MOPO) Pareto front. Two types of simulations will be Evolution (DE) according based on Multi-Objective Parallel Operation due to its fast convergence and ability to run, give SIMULINK high rank to simulation and M.FILE to show the fast respond achieved by the DVR when Evolutionarywill Algorithm employed, the population accordingsimulation to hyper-volume defined by the Pareto front. Two types of simulations be run, isSIMULINK and also to and compare the classical DEtowith improved basedachieved on parallel optimization. simulation M.FILE simulation show the fastDE respond byoperation the DVRmulti-objective when Evolutionary Algorithm is employed, and also to compare the classical DE with improved DE based on parallel operation multi-objective optimization. © 2019 The Authors. Published by Elsevier B.V. © 2019 2019 The The Authors. Published B.V. Peer-review under responsibility of Elsevier the organizing © Authors. Published by by Elsevier B.V. committee of SMPM 2019. Peer-review under responsibility of the organizing committee of SMPM 2019. Peer-review under responsibility of the organizing committee of SMPM 2019. Keywords: Smart Grid;Parallel Operation, Multi-objective Differential Evolution; Dynamic Voltage Restorer; Power Quality; Keywords: Smart Grid;Parallel Operation, Multi-objective Differential Evolution; Dynamic Voltage Restorer; Power Quality;
1. Introduction 1. Introduction The rapid growth of smart homes and production systems technologies has led researchers to consider the power The rapid growth smart productiondevelopments. systems technologies hassmart led researchers consider the power quality needed for of those newhomes smart and technological However, technologiestoneed a quality quality neededreliable for those new smart technological However, technologies need Smart a quality power that is smart, and efficient. Smart grid isdevelopments. the best innovative idea smart to those requirements. Grid is a that is smart,automated reliable and efficient. Smart gridtoisovercome the best innovative those requirements. Smart Gridby is a modernized power grid that is set challenges idea that to have previously been experienced modernized automated power grid that is set to overcome challenges that have previously been experienced by a *Corresponding author : [email protected] *Corresponding author : [email protected] 2351-9789 © 2019 The Authors. Published by Elsevier B.V. Peer-review©under the organizing committee 2351-9789 2019responsibility The Authors. of Published by Elsevier B.V. of SMPM 2019. Peer-review under responsibility of the organizing committee of SMPM 2019.
2351-9789 © 2019 The Authors. Published by Elsevier B.V. Peer-review under responsibility of the organizing committee of SMPM 2019. 10.1016/j.promfg.2019.05.081
2 574
John Saveca, Zenghui Wang and Yanxia Sun/ Procedia Manufacturing 00 (2016) 000–000 John Saveca et al. / Procedia Manufacturing 35 (2019) 573–579
classical power grid by employing distributed energy sources, renewables and energy storage devices [1], therefore intensifying the existing distribution systems [2]. This paper concentrates on power quality, one of the challenges suffered by classical power grid [5]. Power disturbance is defined as any interruption of voltage or frequency that opposes normal operation of the system [3]. Poor power quality results in power failure that causes blowing fuses and tripping circuit breakers, damaging sensitive equipment such as computers and production line control systems, resulting in high running installations costs, leading to production being stopped and major costs being incurred [4]. Methods to improve power quality have been introduced by many researchers. This research looks to improve power quality by considering DVR method, based on improved DE Algorithm, with two objective functions, to minimize swells and to minimize sags. DVR has been frequently used for compensation of voltage sags and swells due to its ability to be controlled, high efficiency and its fast response [6]. Since in this research the focus is to minimize both voltage sag and voltage swell, therefore a multi-objective DE algorithm will be used, as they are extensively used to solve such conflicting objectives. Due to the conflicting nature of the objective functions, a number of optimal solutions known as Pareto Front will always get results for a multi-objective problem [7]. The rest of the paper is organized as follows: Section II discusses the power quality problems, DVR and multiobjective DE algorithm overview. Section III discusses the solution of improving power quality based on Improved DE. Section IV gives the results and discussion of the experiment. Section VI gives the conclusion of the research. 2. Power Quality Background Study Swells and sags are the leading contributors of poor power quality due to their frequent occurrence in power networks [8, 10]. Sags and swells are defined at [9, 10].Sags and swells are sometimes caused by power supplier during heavy demand period and switching-on heavy or reactive equipment such as motors and transformers [11]. These mentioned power disturbances result in fluctuations of production rates and incorrect operation of equipment. The most commonly used solution to overcome poor power quality is DVR due to its high efficiency and fast response. A DVR is a power electronic device that is employed to inject a dynamically controlled voltage in series and in synchronism with the operational voltages for voltage sag and swell compensation [12]. DVR is normally installed in a distribution system, located between the power supply and the sensitive load feeder at the point of common coupling (PCC) [13]. DVR consists of four major parts, Voltage Source Inverter, Injection Transformers, Passive Filters and Energy storage, discussed in details on [12]. Multi-Objective Differential Evolution Multi-objective optimization is a procedure that minimizes or maximizes objectives that are under imposed constraints. It is a mathematical or algorithmic tool that is characterized by two or more objectives [16]. In multiproblem optimization, two or more objectives are usually on conflict, therefore the evolutionary algorithm are required to search the best optimal solution [17]. Different multi-objective strategies have been applied to the problems where the correct clustering solution corresponds to a trade-off between two or more clustering metrics [18]. One strategy used in multi-objective optimization is Pareto Front. Pareto front is the set of all Pareto efficient allocations, conventionally shown in graph. It allocates resources from an impossible situation, making one individual or preference criterion best optimal without making individual or preference criterion worse optimal [19]. Pareto front is represented by Pareto-optimal solutions. With Differential Evolution Algorithm (DE) being one of the robust and effective Evolutionary Algorithm, smart grid optimization will look to employee DE in order to fulfil power quality that is efficient and reliable to the society and industries. According to [14], mutation, crossover and selection function are the three DE evolutionary stages that the objective function passes through during problem optimization. Mutation function randomly generates variations to existing individuals to present new information into the population. The functioning creates mutation vectors at each generation, based on the population of the current parent. The crossover function performs an exchange of information between different individuals in the current population. The final trial vector is formed by binomial crossover operation. The selection function passes a driving force towards the most favourable point by preferring individuals of better fitness. The selection operation selects the better one from the parent vector and the
John Saveca, Zenghui Wang and Yanxia Sun / Procedia Manufacturing 00 (2016) 000–000 John Saveca et al. / Procedia Manufacturing 35 (2019) 573–579
3 575
trial vector according to their fitness values [15]. 3. Multi-Objective Power Quality Optimization Based On Improved Differential Evolution To overcome poor power quality, two objective functions must be minimized, sag and swell, and below circuit is implemented. Figure 1 shows a power system that is assisted by DVR. The system consists of power supply, supplying sensitive RL load, masked control circuit, battery powered inverter connected to the injecting transformer for voltage compensation during fault. The other parts of the circuit are connected in such a way that voltage sags and swells are created to the system for experiment purpose. The purpose of the control circuit is to sense a type of fault from the power system circuit and be able to send a signal to the inverter. The control system uses Phase Locked Loop (PLL) system to sense fault and to synchronize the load voltage level. It can also be noticed that the injected voltage by transformer passes through the series capacitor where the voltage is filtered before being passed to the load. During normal operation of the system, the voltage across the sensitive RL load is expressed as below:
VRL = VSP x sin (2 × π × f × t)
Figure 1: DVR system for Sag and swell Compensation
(1)
where, VL = sensitive RL load voltage in volts (V), VSP = the supply voltage from the AC source in volts (V), f
= the ac voltage supply frequency in hertz (Hz)
t
= time in seconds (s).
Following is the expression during abnormal operation of the system caused by voltage sags due to heavy load switching. VLSAG = {VRL – VF × [H (t-t1) – H (t-t2)]} × sin (2 × π × f × t)
(2)
where, VLSAG
= sensitive RL load voltage during voltage sag disturbance in volts (V),
VF
= the reduction or increment voltage level in volts (V) due to disturbance’
H
= Heaviside step response,
t1
= time when the sag disturbance starts in seconds (s),
t2
= time when the sag disturbance ends in seconds (s).
In the case of voltage swells disturbance, VF is added for a period of (t1 – t2), as expressed in the following equation, VLSWELL = {VRL + VF × [H (t-t3) – H (t-t4)]} × sin (2 × π × f × t)
(3)
John Saveca, Zenghui Wang and Yanxia Sun/ Procedia Manufacturing 00 (2016) 000–000 John Saveca et al. / Procedia Manufacturing 35 (2019) 573–579
4 576
where, VLSWELL = sensitive RL load voltage during voltage swell disturbance in volts (V), Following are equations for DVR’s action during the voltage compensation for voltage sags 0 = VF – V C
(4)
VF = – V C
(5)
VCOMP = {-VC × [H (t-t1) – H (t-t2)]} × sin (2 × π × f × t)
(6)
where, VC
= the added voltage for compensation or counterbalancing of operational voltage level during DVR action.
VCOMP = DVR voltage compensation for voltage sags Therefore VC is variable during DVR action in order to certify the objective requirements. The load voltage after compensation becomes as follows, VL = VLSAG – VCOMP
(7)
For voltage swell correction on the system, the similar action as of voltage sag compensation is performed by DVR for operational voltage level correction during voltage swell disturbance. In this case the action happens in opposite polarity from the voltage sag compensation. 0 = – VF + VC
(8)
VF = – V C
(9)
VCORR = {VC × [H (t-t3) – H (t-t4)]} × sin (2 × π × f × t)
where, VCORR = DVR operational voltage level correction for voltage swells.
(10)
Therefore the load voltage after correction becomes as follows. VL = VLSWELL – VCORR
(11)
For voltage sag and swell optimization based on DE, the equations becomes as follows, VSAGmin = VL – VLSAG
(12)
J1 = VSAGmin+ VCOMP
(13)
VSWELLmin = VL - VLSAG
(14)
J2 = VSWELLmin+ VCORR
(15)
where
VSAGmin
= voltage sag occurrence
VSWELLmin
= voltage swell occurrence
J1
= First objective function
J2
= Second objective function
J1 and J2 are multi-objectives to be minimized in smart grid based on proposed DE, VC is the variable to be controlled during voltage sag compensation and voltage swell counterbalancing based on [14]. For optimization of J1 and J2, a multi-objective DE based on parallel operation strategy will be used. The advantage of the strategy is that, two or more objective functions don’t conflict with each other because they run parallel once they reach mutation function up until selection phase, therefore creating a room for one objective function to be optimized effectively to the maximum or to the minimum set target. During operation, the objective functions are initialized with the same control parameters and same limitations. The functions are parallelized before mutation operation is introduced. The functions run parallel until they reach selection operation, therefore resulting in a
John Saveca, Zenghui Wang and Yanxia Sun / Procedia Manufacturing 00 (2016) 000–000 John Saveca et al. / Procedia Manufacturing 35 (2019) 573–579
5 577
separate Pareto set graphic presentation. Formulas for mutation, crossover and selection are discussed on [14]. 4. Results and Discussion Following are results taken during simulations of multi-objective power quality optimization in smart grid based on the proposed DE. The first results are taken before the proposed DE algorithm is implemented. Output
Output
Figure 3: Load voltage behavior with DVR connected.
Figure 2: Load voltage behavior with DVR disconnected
Figure 2 and 3 shows three graphs each, top graph shows the load terminal voltage behavior, middle graph shows the DVR response to the load voltage and the bottom graph shows the load voltage behavior due to system voltage. Figure 12: Load voltage after correction.
2 3 4 5
18 21 10 14
1.052e-11 1.055e-11 1.052e-11 1.056e-11
-6.935e-12 -6.954e-12 -6.959e-12 -6.959e-12
Iteration number 10 11 12 9 10
Sags fitness 5.234e-12 5.142e-12 5.376e-12 5.322e-12 5.334e-12
Swells fitness -5.988e-12 -6.837e-12 -6.309e-12 -6.731e-12 -6.745e-12
Iteration number 5 6 7 3 5
Sags fitness 4.902e-12 4.717e-12 4.873e-12 4.744e-12 4.903e-12
Swells fitness -6.625e-12 -6.322e-12 -6.249e-12 -5.995e-12 -6.790e-12
DE/Modi/2
Figure 4: Sag and swell effects after voltage correction from DVR.
On Figure 4, the red painted part represents voltage sags effects while the blue part represents voltage swells effect.
Run number 1 2 3 4 5 DE/Modi/3
Figure 5: Multi-objective minimization based on DE/Modi/2.
Table 1: DE/Rand/1 fitness for Sags and Swells DE/Rand/1 Run number 1
Iteration number 14
Sags fitness 1.057e-11
Swells fitness -6.959e-12
Run number 1 2 3 4 5
Table 2: Statistical Data Analysis Mutation Best Average Worse strategy DE/Rand/1 2 2 1 DE/Modi/2 4 1 0 DE/Modi/3 4 1 0
578
John Saveca et al. / Procedia Manufacturing 35 (2019) 573–579
5. Conclusion Based on results, it is noticed that the DVR does not take part in the power network during normal operational, it only takes part when it senses operational disturbance from the sensitive load, therefore it does not waste electric power unnecessary and instead it operates efficiently. It is also noticed that Improved Differential Evolution is able to do multi-objective optimization efficiently and without experiencing stagnation and conflict amongst the two objective functions. The proposed DE achieved robust and effective convergence in less than 30 iterations. That indicates efficiency, reliability and fast convergence speed of the improved DE. During voltage disturbance optimization, it is noticed that the voltage disturbances were minimized up to as minimum as 5.234e-11 volts for voltage sags and up to as minimum as -6.873 volts for voltage swells and the mutation strategies were consistence with regards to iterations were convergence occurs, therefore making the Improved DE suitable for power quality optimization in smart grid. It can be concluded that robust, efficient, sustainable, economical and reliable power quality optimization in smart grid is achievable when DVR and improved DE are included in smart grid’s mission. Acknowledgments This research is supported partially by South African National Research Foundation Grants (No. 112108&112142), South African National Research Foundation Incentive Grant (No. 114911), and Tertiary Education Support Program (TESP) of South African ESKOM. References F. Y. Melhem , O. Grunder, Z. Hammoudan, and N. Moubayed, “Energy Management in Electrical Smart Grid Environment Using Robust Optimization Algorithm”, IEEE Transactions on Industry Applications, Vol. 54, no. 3, p. 2714-2726 , 2018. [2] G. M. Casolino, A. R. Di Fazio, A. Losi, M. Russo and M. De Santis,” A Voltage Optimization Tool for Smart Distribution Grids with Distributed Energy Resources”, IEEE International Conference on Environment and Electrical Engineering, p. 1-6, 2017. [3] T. Horsley, and J. Seymour, “The seven types of power problems”,Journal of Engineer IT, p. 60–65, 2008. [4] F. Delincé, & K. Schipman, “The importance of good power quality”, ABB, Report number: 1., 2010. [5] V. Agarwal, and L.H. Tsoukalas, “Smart Grids: Importance of Power Quality”, In: Hatziargyriou, N. et al. (Eds.) E-Energy 2010. Institute for Computer Sciences, Social Informatics and Telecommunications Engineering 2011. LNICST, vol. 54. p. 136–143, 2011. [6] M. A. Messiha, C. F. Maraket, A. M. Massoud, A. Iqbal and R Soliman, “ Voltage Sag Mitigation Employing Dynamic Voltage Restorer with Minimum Eneregy Requirements: Analysis and Implementation”,p. 1-6, 2017. [7] S.T. Suganthi, D.Devaraj, S.HosiminThilagar and K. Ramar, “Improved Multi Objective Differential Evolution Algorithm for Congestion Management in Restructured Power Systems”, p. 203-210, 2016 [8] M. Hojabri, and A. Toudeshki, “ Power Quality Consideration for Off-Grid Renewable Energy Systems”, Journal of Energy and Power Engineering, Vol .5, p. 377–383., 2013. [9] B. K. Anand, Y. T. Mahamadnayeem R. Atre, " Compensation of Voltage sag/swell and harmonics by DVR", International journal of advanced electronics and communication systems, February 2014. [10] P. Thakur, A. K. Singh, “A novel method for joint characterization of unbalanced voltage sags and swells”, Int Trans Electr Energ Syst. 2017;27:e2370, John Wiley & Sons, Ltd., p. 1-10, 2017. [1]
*Corresponding author : [email protected] 2351-9789 © 2019 The Authors. Published by Elsevier B.V. Peer-review under responsibility of the organizing committee of SMPM 2019.
John Saveca, Zenghui andetYanxia Sun / Procedia Manufacturing 00 573–579 (2016) 000–000 JohnWang Saveca al. / Procedia Manufacturing 35 (2019)
7 579
[11] N. Edomah, “ Effects of Voltage Sags, Swell and Other Disturbances on Electrical Equipment and Their Economic Implications”, 20th International Conference on Electricity Distribution, CIRED, p. 1-4, 2009. [12] A. K. Ramasamy, R. K. Iyer, V. K Ramachandaramuthy, R.N. Mukerjee, “Dynamic Voltage Restorer For Voltage Sag Compensation”, IEEE International Conference on Power Electronics and Drive Systems(IEEE PEDS 2005), p. 1289-1294, 2005. [13] A. K. Jena, B. Mohapatra, K. Pradhan and K.R. Subhashini, “Modeling and Simulation of a Dynamic Voltage Restorer” Department of Electrical Engineering National Institute of Technology Rourkela, Odisha -769008, p 1-50, 2011. [14] J. Saveca, Z. Wang and Y. Sun,” Improved Differential Evolution Based on Mutation Strategies”, Lecture Note in Computer Science, Springer,Cham, vol. 10941. p. 233-242, 2018 [15] M. P. Ganbavale, “Differential Evolution Using Matlab”. Birla Institute of Technology and Science, Pilani. Hyderabad Campus, 2014 [16] A. R. Adekoya, M. Helbig, “A Differential Evolution Algorithm for Dynamic Multi-Objective Optimization” Congress on Evolutionary Computation (CEC), p. 1-10, 2017. [17] M. Reyes-Sierra and C. C. Coello, "Multi-objective particle swarm optimizers: survey of the state-of-the-art," International journal of computational intelligence research, Vol. 2, no. 3, p. 287-308, 2006. [18] F. Lezama, A. Y. Rodrıguez-Gonz´alez and E. M. de Cote, “Load Pattern Clustering Using Differential Evolution with Pareto Tournament” Congress on Evolutionary Computation (CEC), p. 241- 248, 2016. [19] J. M. Chaves-Gonz´alez and M. A. P´erez-Toledano, “Differential evolution with pareto tournament for the multi-objective next release problem,” Applied Mathematics and Computation, Vol. 252, p. 1–13, 2015.
ScienceDirect ScienceDirect
Procedia Manufacturing 00 (2019) 000–000 Procedia Manufacturing 00 (2019) 000–000
Available online at www.sciencedirect.com
ScienceDirect
www.elsevier.com/locate/procedia www.elsevier.com/locate/procedia
Procedia Manufacturing 35 (2019) 573–579
2nd International Conference on Sustainable Materials Processing and Manufacturing 2nd International Conference on Sustainable Materials Processing and Manufacturing (SMPM 2019) (SMPM 2019)
Dynamic voltage restorer-based power quality optimization using Dynamic voltage restorer-based power quality optimization using differential evolution algorithm differential evolution algorithm John Savecaa, Zenghui Wanga*, Yanxia Sunb John Savecaa, Zenghui Wanga*, Yanxia Sunb
a Department of Electrical and Mining Engineering University of South Africa, Johannesburg 1710, South Africa a Department of Electrical and Electronic Engineering Science University of Johannesburg, Johannesburg 2006, Africa South Africa Department of Electrical and Mining Engineering University of South Africa, Johannesburg 1710, South b Department of Electrical and Electronic Engineering Science University of Johannesburg, Johannesburg 2006, South Africa b
Abstract Abstract Homes and industrial companies including electric utilities are much concerned about the study of power quality and ways to control it. is due to equipmentincluding and newelectric household’s technologies even small changes supply Homes andThis industrial companies utilities are much becoming concerned more aboutsensitive the studyto of power quality andofways to voltages.it.Therefore power quality isand onenew of the most important aspects becoming in smart grid’s of becoming a reliable andofefficient control This is due to equipment household’s technologies morevision sensitive to even small changes supply new generation power grid.quality This paper to develop a strategy overcomes poor power quality ainreliable smart grid using voltages. Therefore power is oneattempts of the most important aspects that in smart grid’s vision of becoming and by efficient Dynamic Voltage Restorer due toattempts its efficient compensation of that voltage sags andpoor swells. Employing improved Differential new generation power grid.(DVR) This paper to develop a strategy overcomes power quality in smart grid by using Evolution (DE) based on Multi-Objective Parallel Operation (MOPO) to itssags fastand convergence and ability to give high rank to Dynamic Voltage Restorer (DVR) due to its efficient compensation of due voltage swells. Employing improved Differential the population to hyper-volume defined by the (MOPO) Pareto front. Two types of simulations will be Evolution (DE) according based on Multi-Objective Parallel Operation due to its fast convergence and ability to run, give SIMULINK high rank to simulation and M.FILE to show the fast respond achieved by the DVR when Evolutionarywill Algorithm employed, the population accordingsimulation to hyper-volume defined by the Pareto front. Two types of simulations be run, isSIMULINK and also to and compare the classical DEtowith improved basedachieved on parallel optimization. simulation M.FILE simulation show the fastDE respond byoperation the DVRmulti-objective when Evolutionary Algorithm is employed, and also to compare the classical DE with improved DE based on parallel operation multi-objective optimization. © 2019 The Authors. Published by Elsevier B.V. © 2019 2019 The The Authors. Published B.V. Peer-review under responsibility of Elsevier the organizing © Authors. Published by by Elsevier B.V. committee of SMPM 2019. Peer-review under responsibility of the organizing committee of SMPM 2019. Peer-review under responsibility of the organizing committee of SMPM 2019. Keywords: Smart Grid;Parallel Operation, Multi-objective Differential Evolution; Dynamic Voltage Restorer; Power Quality; Keywords: Smart Grid;Parallel Operation, Multi-objective Differential Evolution; Dynamic Voltage Restorer; Power Quality;
1. Introduction 1. Introduction The rapid growth of smart homes and production systems technologies has led researchers to consider the power The rapid growth smart productiondevelopments. systems technologies hassmart led researchers consider the power quality needed for of those newhomes smart and technological However, technologiestoneed a quality quality neededreliable for those new smart technological However, technologies need Smart a quality power that is smart, and efficient. Smart grid isdevelopments. the best innovative idea smart to those requirements. Grid is a that is smart,automated reliable and efficient. Smart gridtoisovercome the best innovative those requirements. Smart Gridby is a modernized power grid that is set challenges idea that to have previously been experienced modernized automated power grid that is set to overcome challenges that have previously been experienced by a *Corresponding author : [email protected] *Corresponding author : [email protected] 2351-9789 © 2019 The Authors. Published by Elsevier B.V. Peer-review©under the organizing committee 2351-9789 2019responsibility The Authors. of Published by Elsevier B.V. of SMPM 2019. Peer-review under responsibility of the organizing committee of SMPM 2019.
2351-9789 © 2019 The Authors. Published by Elsevier B.V. Peer-review under responsibility of the organizing committee of SMPM 2019. 10.1016/j.promfg.2019.05.081
2 574
John Saveca, Zenghui Wang and Yanxia Sun/ Procedia Manufacturing 00 (2016) 000–000 John Saveca et al. / Procedia Manufacturing 35 (2019) 573–579
classical power grid by employing distributed energy sources, renewables and energy storage devices [1], therefore intensifying the existing distribution systems [2]. This paper concentrates on power quality, one of the challenges suffered by classical power grid [5]. Power disturbance is defined as any interruption of voltage or frequency that opposes normal operation of the system [3]. Poor power quality results in power failure that causes blowing fuses and tripping circuit breakers, damaging sensitive equipment such as computers and production line control systems, resulting in high running installations costs, leading to production being stopped and major costs being incurred [4]. Methods to improve power quality have been introduced by many researchers. This research looks to improve power quality by considering DVR method, based on improved DE Algorithm, with two objective functions, to minimize swells and to minimize sags. DVR has been frequently used for compensation of voltage sags and swells due to its ability to be controlled, high efficiency and its fast response [6]. Since in this research the focus is to minimize both voltage sag and voltage swell, therefore a multi-objective DE algorithm will be used, as they are extensively used to solve such conflicting objectives. Due to the conflicting nature of the objective functions, a number of optimal solutions known as Pareto Front will always get results for a multi-objective problem [7]. The rest of the paper is organized as follows: Section II discusses the power quality problems, DVR and multiobjective DE algorithm overview. Section III discusses the solution of improving power quality based on Improved DE. Section IV gives the results and discussion of the experiment. Section VI gives the conclusion of the research. 2. Power Quality Background Study Swells and sags are the leading contributors of poor power quality due to their frequent occurrence in power networks [8, 10]. Sags and swells are defined at [9, 10].Sags and swells are sometimes caused by power supplier during heavy demand period and switching-on heavy or reactive equipment such as motors and transformers [11]. These mentioned power disturbances result in fluctuations of production rates and incorrect operation of equipment. The most commonly used solution to overcome poor power quality is DVR due to its high efficiency and fast response. A DVR is a power electronic device that is employed to inject a dynamically controlled voltage in series and in synchronism with the operational voltages for voltage sag and swell compensation [12]. DVR is normally installed in a distribution system, located between the power supply and the sensitive load feeder at the point of common coupling (PCC) [13]. DVR consists of four major parts, Voltage Source Inverter, Injection Transformers, Passive Filters and Energy storage, discussed in details on [12]. Multi-Objective Differential Evolution Multi-objective optimization is a procedure that minimizes or maximizes objectives that are under imposed constraints. It is a mathematical or algorithmic tool that is characterized by two or more objectives [16]. In multiproblem optimization, two or more objectives are usually on conflict, therefore the evolutionary algorithm are required to search the best optimal solution [17]. Different multi-objective strategies have been applied to the problems where the correct clustering solution corresponds to a trade-off between two or more clustering metrics [18]. One strategy used in multi-objective optimization is Pareto Front. Pareto front is the set of all Pareto efficient allocations, conventionally shown in graph. It allocates resources from an impossible situation, making one individual or preference criterion best optimal without making individual or preference criterion worse optimal [19]. Pareto front is represented by Pareto-optimal solutions. With Differential Evolution Algorithm (DE) being one of the robust and effective Evolutionary Algorithm, smart grid optimization will look to employee DE in order to fulfil power quality that is efficient and reliable to the society and industries. According to [14], mutation, crossover and selection function are the three DE evolutionary stages that the objective function passes through during problem optimization. Mutation function randomly generates variations to existing individuals to present new information into the population. The functioning creates mutation vectors at each generation, based on the population of the current parent. The crossover function performs an exchange of information between different individuals in the current population. The final trial vector is formed by binomial crossover operation. The selection function passes a driving force towards the most favourable point by preferring individuals of better fitness. The selection operation selects the better one from the parent vector and the
John Saveca, Zenghui Wang and Yanxia Sun / Procedia Manufacturing 00 (2016) 000–000 John Saveca et al. / Procedia Manufacturing 35 (2019) 573–579
3 575
trial vector according to their fitness values [15]. 3. Multi-Objective Power Quality Optimization Based On Improved Differential Evolution To overcome poor power quality, two objective functions must be minimized, sag and swell, and below circuit is implemented. Figure 1 shows a power system that is assisted by DVR. The system consists of power supply, supplying sensitive RL load, masked control circuit, battery powered inverter connected to the injecting transformer for voltage compensation during fault. The other parts of the circuit are connected in such a way that voltage sags and swells are created to the system for experiment purpose. The purpose of the control circuit is to sense a type of fault from the power system circuit and be able to send a signal to the inverter. The control system uses Phase Locked Loop (PLL) system to sense fault and to synchronize the load voltage level. It can also be noticed that the injected voltage by transformer passes through the series capacitor where the voltage is filtered before being passed to the load. During normal operation of the system, the voltage across the sensitive RL load is expressed as below:
VRL = VSP x sin (2 × π × f × t)
Figure 1: DVR system for Sag and swell Compensation
(1)
where, VL = sensitive RL load voltage in volts (V), VSP = the supply voltage from the AC source in volts (V), f
= the ac voltage supply frequency in hertz (Hz)
t
= time in seconds (s).
Following is the expression during abnormal operation of the system caused by voltage sags due to heavy load switching. VLSAG = {VRL – VF × [H (t-t1) – H (t-t2)]} × sin (2 × π × f × t)
(2)
where, VLSAG
= sensitive RL load voltage during voltage sag disturbance in volts (V),
VF
= the reduction or increment voltage level in volts (V) due to disturbance’
H
= Heaviside step response,
t1
= time when the sag disturbance starts in seconds (s),
t2
= time when the sag disturbance ends in seconds (s).
In the case of voltage swells disturbance, VF is added for a period of (t1 – t2), as expressed in the following equation, VLSWELL = {VRL + VF × [H (t-t3) – H (t-t4)]} × sin (2 × π × f × t)
(3)
John Saveca, Zenghui Wang and Yanxia Sun/ Procedia Manufacturing 00 (2016) 000–000 John Saveca et al. / Procedia Manufacturing 35 (2019) 573–579
4 576
where, VLSWELL = sensitive RL load voltage during voltage swell disturbance in volts (V), Following are equations for DVR’s action during the voltage compensation for voltage sags 0 = VF – V C
(4)
VF = – V C
(5)
VCOMP = {-VC × [H (t-t1) – H (t-t2)]} × sin (2 × π × f × t)
(6)
where, VC
= the added voltage for compensation or counterbalancing of operational voltage level during DVR action.
VCOMP = DVR voltage compensation for voltage sags Therefore VC is variable during DVR action in order to certify the objective requirements. The load voltage after compensation becomes as follows, VL = VLSAG – VCOMP
(7)
For voltage swell correction on the system, the similar action as of voltage sag compensation is performed by DVR for operational voltage level correction during voltage swell disturbance. In this case the action happens in opposite polarity from the voltage sag compensation. 0 = – VF + VC
(8)
VF = – V C
(9)
VCORR = {VC × [H (t-t3) – H (t-t4)]} × sin (2 × π × f × t)
where, VCORR = DVR operational voltage level correction for voltage swells.
(10)
Therefore the load voltage after correction becomes as follows. VL = VLSWELL – VCORR
(11)
For voltage sag and swell optimization based on DE, the equations becomes as follows, VSAGmin = VL – VLSAG
(12)
J1 = VSAGmin+ VCOMP
(13)
VSWELLmin = VL - VLSAG
(14)
J2 = VSWELLmin+ VCORR
(15)
where
VSAGmin
= voltage sag occurrence
VSWELLmin
= voltage swell occurrence
J1
= First objective function
J2
= Second objective function
J1 and J2 are multi-objectives to be minimized in smart grid based on proposed DE, VC is the variable to be controlled during voltage sag compensation and voltage swell counterbalancing based on [14]. For optimization of J1 and J2, a multi-objective DE based on parallel operation strategy will be used. The advantage of the strategy is that, two or more objective functions don’t conflict with each other because they run parallel once they reach mutation function up until selection phase, therefore creating a room for one objective function to be optimized effectively to the maximum or to the minimum set target. During operation, the objective functions are initialized with the same control parameters and same limitations. The functions are parallelized before mutation operation is introduced. The functions run parallel until they reach selection operation, therefore resulting in a
John Saveca, Zenghui Wang and Yanxia Sun / Procedia Manufacturing 00 (2016) 000–000 John Saveca et al. / Procedia Manufacturing 35 (2019) 573–579
5 577
separate Pareto set graphic presentation. Formulas for mutation, crossover and selection are discussed on [14]. 4. Results and Discussion Following are results taken during simulations of multi-objective power quality optimization in smart grid based on the proposed DE. The first results are taken before the proposed DE algorithm is implemented. Output
Output
Figure 3: Load voltage behavior with DVR connected.
Figure 2: Load voltage behavior with DVR disconnected
Figure 2 and 3 shows three graphs each, top graph shows the load terminal voltage behavior, middle graph shows the DVR response to the load voltage and the bottom graph shows the load voltage behavior due to system voltage. Figure 12: Load voltage after correction.
2 3 4 5
18 21 10 14
1.052e-11 1.055e-11 1.052e-11 1.056e-11
-6.935e-12 -6.954e-12 -6.959e-12 -6.959e-12
Iteration number 10 11 12 9 10
Sags fitness 5.234e-12 5.142e-12 5.376e-12 5.322e-12 5.334e-12
Swells fitness -5.988e-12 -6.837e-12 -6.309e-12 -6.731e-12 -6.745e-12
Iteration number 5 6 7 3 5
Sags fitness 4.902e-12 4.717e-12 4.873e-12 4.744e-12 4.903e-12
Swells fitness -6.625e-12 -6.322e-12 -6.249e-12 -5.995e-12 -6.790e-12
DE/Modi/2
Figure 4: Sag and swell effects after voltage correction from DVR.
On Figure 4, the red painted part represents voltage sags effects while the blue part represents voltage swells effect.
Run number 1 2 3 4 5 DE/Modi/3
Figure 5: Multi-objective minimization based on DE/Modi/2.
Table 1: DE/Rand/1 fitness for Sags and Swells DE/Rand/1 Run number 1
Iteration number 14
Sags fitness 1.057e-11
Swells fitness -6.959e-12
Run number 1 2 3 4 5
Table 2: Statistical Data Analysis Mutation Best Average Worse strategy DE/Rand/1 2 2 1 DE/Modi/2 4 1 0 DE/Modi/3 4 1 0
578
John Saveca et al. / Procedia Manufacturing 35 (2019) 573–579
5. Conclusion Based on results, it is noticed that the DVR does not take part in the power network during normal operational, it only takes part when it senses operational disturbance from the sensitive load, therefore it does not waste electric power unnecessary and instead it operates efficiently. It is also noticed that Improved Differential Evolution is able to do multi-objective optimization efficiently and without experiencing stagnation and conflict amongst the two objective functions. The proposed DE achieved robust and effective convergence in less than 30 iterations. That indicates efficiency, reliability and fast convergence speed of the improved DE. During voltage disturbance optimization, it is noticed that the voltage disturbances were minimized up to as minimum as 5.234e-11 volts for voltage sags and up to as minimum as -6.873 volts for voltage swells and the mutation strategies were consistence with regards to iterations were convergence occurs, therefore making the Improved DE suitable for power quality optimization in smart grid. It can be concluded that robust, efficient, sustainable, economical and reliable power quality optimization in smart grid is achievable when DVR and improved DE are included in smart grid’s mission. Acknowledgments This research is supported partially by South African National Research Foundation Grants (No. 112108&112142), South African National Research Foundation Incentive Grant (No. 114911), and Tertiary Education Support Program (TESP) of South African ESKOM. References F. Y. Melhem , O. Grunder, Z. Hammoudan, and N. Moubayed, “Energy Management in Electrical Smart Grid Environment Using Robust Optimization Algorithm”, IEEE Transactions on Industry Applications, Vol. 54, no. 3, p. 2714-2726 , 2018. [2] G. M. Casolino, A. R. Di Fazio, A. Losi, M. Russo and M. De Santis,” A Voltage Optimization Tool for Smart Distribution Grids with Distributed Energy Resources”, IEEE International Conference on Environment and Electrical Engineering, p. 1-6, 2017. [3] T. Horsley, and J. Seymour, “The seven types of power problems”,Journal of Engineer IT, p. 60–65, 2008. [4] F. Delincé, & K. Schipman, “The importance of good power quality”, ABB, Report number: 1., 2010. [5] V. Agarwal, and L.H. Tsoukalas, “Smart Grids: Importance of Power Quality”, In: Hatziargyriou, N. et al. (Eds.) E-Energy 2010. Institute for Computer Sciences, Social Informatics and Telecommunications Engineering 2011. LNICST, vol. 54. p. 136–143, 2011. [6] M. A. Messiha, C. F. Maraket, A. M. Massoud, A. Iqbal and R Soliman, “ Voltage Sag Mitigation Employing Dynamic Voltage Restorer with Minimum Eneregy Requirements: Analysis and Implementation”,p. 1-6, 2017. [7] S.T. Suganthi, D.Devaraj, S.HosiminThilagar and K. Ramar, “Improved Multi Objective Differential Evolution Algorithm for Congestion Management in Restructured Power Systems”, p. 203-210, 2016 [8] M. Hojabri, and A. Toudeshki, “ Power Quality Consideration for Off-Grid Renewable Energy Systems”, Journal of Energy and Power Engineering, Vol .5, p. 377–383., 2013. [9] B. K. Anand, Y. T. Mahamadnayeem R. Atre, " Compensation of Voltage sag/swell and harmonics by DVR", International journal of advanced electronics and communication systems, February 2014. [10] P. Thakur, A. K. Singh, “A novel method for joint characterization of unbalanced voltage sags and swells”, Int Trans Electr Energ Syst. 2017;27:e2370, John Wiley & Sons, Ltd., p. 1-10, 2017. [1]
*Corresponding author : [email protected] 2351-9789 © 2019 The Authors. Published by Elsevier B.V. Peer-review under responsibility of the organizing committee of SMPM 2019.
John Saveca, Zenghui andetYanxia Sun / Procedia Manufacturing 00 573–579 (2016) 000–000 JohnWang Saveca al. / Procedia Manufacturing 35 (2019)
7 579
[11] N. Edomah, “ Effects of Voltage Sags, Swell and Other Disturbances on Electrical Equipment and Their Economic Implications”, 20th International Conference on Electricity Distribution, CIRED, p. 1-4, 2009. [12] A. K. Ramasamy, R. K. Iyer, V. K Ramachandaramuthy, R.N. Mukerjee, “Dynamic Voltage Restorer For Voltage Sag Compensation”, IEEE International Conference on Power Electronics and Drive Systems(IEEE PEDS 2005), p. 1289-1294, 2005. [13] A. K. Jena, B. Mohapatra, K. Pradhan and K.R. Subhashini, “Modeling and Simulation of a Dynamic Voltage Restorer” Department of Electrical Engineering National Institute of Technology Rourkela, Odisha -769008, p 1-50, 2011. [14] J. Saveca, Z. Wang and Y. Sun,” Improved Differential Evolution Based on Mutation Strategies”, Lecture Note in Computer Science, Springer,Cham, vol. 10941. p. 233-242, 2018 [15] M. P. Ganbavale, “Differential Evolution Using Matlab”. Birla Institute of Technology and Science, Pilani. Hyderabad Campus, 2014 [16] A. R. Adekoya, M. Helbig, “A Differential Evolution Algorithm for Dynamic Multi-Objective Optimization” Congress on Evolutionary Computation (CEC), p. 1-10, 2017. [17] M. Reyes-Sierra and C. C. Coello, "Multi-objective particle swarm optimizers: survey of the state-of-the-art," International journal of computational intelligence research, Vol. 2, no. 3, p. 287-308, 2006. [18] F. Lezama, A. Y. Rodrıguez-Gonz´alez and E. M. de Cote, “Load Pattern Clustering Using Differential Evolution with Pareto Tournament” Congress on Evolutionary Computation (CEC), p. 241- 248, 2016. [19] J. M. Chaves-Gonz´alez and M. A. P´erez-Toledano, “Differential evolution with pareto tournament for the multi-objective next release problem,” Applied Mathematics and Computation, Vol. 252, p. 1–13, 2015.