A real-time management and evolutionary optimization scheme for a secure and flexible smart grid towards sustainable energy

Electrical Power and Energy Systems 62 (2014) 540–548

Contents lists available at ScienceDirect

Electrical Power and Energy Systems journal homepage: www.elsevier.com/locate/ijepes

A real-time management and evolutionary optimization scheme for a secure and flexible smart grid towards sustainable energy D. Vijayakumar a,⇑, V. Malathi b a b

Electronics and Communication Engineering, Lathamathavan Engineering College, Alagarkoil, Tamil Nadu, India Electrical and Electronics Engineering, Anna University Regional Centre, Madurai, Tamil Nadu, India

a r t i c l e

i n f o

Article history: Received 10 December 2013 Received in revised form 22 April 2014 Accepted 10 May 2014

Keywords: Critical time Control center Distributed state estimator (DSE) Fault distance Genetic Algorithm (GA) Reactive power

a b s t r a c t Sustainable energy is the energy production without compromising the energy production for the future generations. The existing power grid model does not provide real-time information of transmission devices, security during emergency events, and frequency and voltage control. The proposed scheme consists of location-centric hybrid system architecture for the coordinated processing among proximate devices. The neighboring devices follow a collaboration algorithm to handle faulty and incomplete information. The proposed scheme also consists of distributed algorithm for the maintenance of the local state of the smart grid and a real-time accessibility control of transmission devices. The security of the system can be guaranteed by reconfiguration through power-electronics and switches. An embedded intelligence is inserted into the power-electronics to facilitate the reconfiguration of the system, and thereby ensuring security. A generalized optimization formulation determines the optimum location of the transmission devices. In this paper, Genetic Algorithm (GA) is used to handle the reactive power management. The use of GA decreases execution time of resource scheduling. This method performs better than the existing power grid models in terms of fault detection, degree of power saving due to power optimization, memory usage, consumer-GW (gateway) communication overhead, consumer computational overhead, and critical time. Ó 2014 Elsevier Ltd. All rights reserved.

Introduction THE SMART GRID is an evolution of the electrical grid to a flexible power network through distributed intelligence, automated control systems, and communication technologies. The power management techniques must yield sustainable energy; although there is increasing power blackouts due to the excessive power consumption. The demerits in the existing power grid models occur at the system level and local level. The control area operators in the power grid cannot, obtain the real-time information about the transmission devices, respond quickly to emergency events (or) blackouts, and perform the functions in an automated and coordinated manner. The conventional hardware used in the electrical grid lacks the frequency and voltage control according to the increasing system requirements, and cannot secure the system quickly during emergency events. A real-time coordination scheme enhances the coordination among the geographically separated devices during power blackouts.

⇑ Corresponding author. Tel.: +91 4526454125; fax: +91 0452 247 0570. E-mail address: [email protected] (D. Vijayakumar). http://dx.doi.org/10.1016/j.ijepes.2014.05.013 0142-0615/Ó 2014 Elsevier Ltd. All rights reserved.

The energy resources in the transmission systems must be efficiently utilized over the conventional high-energy resources to increase the reliability of the power system. The security of a power grid is mainly focused on dynamic and transient stability issues. The power systems can also crash via physical interconnection even when the communication system is secure. A controller can be compromised due to destabilization and higher response to minor load variations. A security control can be designed by providing a feedback for the detection and isolation of information flow targeted attacks. The information collaboration and feedback to the perfect generator units may isolate the offence unit of the system. The attacks on the power grids may partially compromise the secure communication or fully control some of the system components. Multiple lines of defense must be designed in a power grid system. It needs to fulfill the following goals:  Accurate detection of attacks depicted as current, voltage, or frequency changes.  Distributed processing at the local level and robust collaboration between the grid devices.  Protection of the system against collapse.

D. Vijayakumar, V. Malathi / Electrical Power and Energy Systems 62 (2014) 540–548

Location-centric hybrid system architecture is chosen for the cooperative processing among neighboring devices. The electrical devices in the power grid follow a collaboration algorithm to address the faulty and incomplete information. A distributed algorithm is used for the real-time accessibility control of transmission devices and the maintenance of the local states of the smart grid. The reconfiguration of the grid via power-electronics and switches enhance the security of the electrical grid. An embedded intelligence is fixed into the power-electronics to enable the system reconfiguration. An evolutionary algorithm like Genetic Algorithm (GA) is chosen to address the reactive power management and optimum transmission device placement. The execution time of the resource scheduling is decreased by the optimization of GA. This method performs better than the existing power grid models in terms of fault detection, degree of power saving due to power optimization, memory usage, consumer-GW (gateway) communication overhead, consumer computational overhead, and critical time. The existing methods taken for comparative analysis are fault detection and classification using Functional Analysis and Computational Intelligence (FACI), Efficient and Privacy-Preserving Aggregation (EPPA) scheme, Lightweight Message Authentication (LMA) scheme, Multi-terminal DC Wind farm collection Internal Fault Analysis (MDWIFA), and traditional scheme (TRAD) without any data aggregation. The remaining part of the paper is organized as follows: Section ‘Related work’ involves the works related to the security enhancement and optimization in a power grid. Section ‘Requirements for a secure power grid’ involves the requirements for a secure power grid. Section ‘Real-time management and evolutionary optimization scheme for a secure and flexible smart grid (RMOSFS)’ involves the detailed description of the proposed real-time management and optimization scheme for a secure and flexible smart towards sustainable energy. Section ‘Performance analysis’ involves the performance analysis and comparison of the existing and proposed security enhancement and optimization techniques in a power grid. The paper is concluded in Section ‘Conclusion’.

Related work This section deals with the existing secure and optimized power grid models. A smart grid is a modern electrical grid infrastructure for higher efficiency and reliability via automated control, modern communications, sensing, high power converters, metering technologies, and energy management schemes [1]. The recent smart grid management and protection systems were surveyed by [2,3]. These smart protection systems enhanced the reliability and security of the smart grid. The suitability of Attribute Based Encryption (ABE) was analyzed for the security of smart grids [4]. A key policy based ABE was applied at the smart grid’s control center, where an encrypted message was broadcast to a defined group of users. This eliminated the requisite for multiple unicast message broadcast, which further ensured the computation and communication efficiencies. Miao and Junshan proposed a dependency graph based fault detection and localization towards secure smart grid [5]. The phasor angles across the communication links were modeled as a Markov random field (MRF), whose conditional correlation coefficients were measured in terms of physical properties of power systems. A multiscale network interference algorithm was devised to detect and localize the faults in a decentralized manner. Fouda et al. designed a Lightweight Message Authentication technique for smart grids [6]. The smart meters were distributed at various hierarchical smart grid networks. A shared session key was established between a hash-based authentication code and the smart meters.

541

Young–Jin et al. formed a data-centric, decentralized and secure information infrastructure for smart grid [7]. The secure middleware architecture can coexist with both LAN and WAN. Rodri et al. used a grid synchronization algorithm for three-phase grid-connected power networks [8]. This method is based on two adaptive filters designed using a dual second-order generalized integrator (SOGI). Chouder and Silvestre implemented automatic and supervisory fault detection technique on the Photo–Voltaic (PV) systems of the smart grid [9]. The fault detection is based on analysis of power losses. Budischak et al. modeled the cost-minimized combinations of various renewable energy sources in smart grid to power up the grid up to 99.9% of the time [10]. The essential cyber security issues in a smart grid infrastructure were dealt in [11–13]. A layered approach was introduced to evaluate the risk on physical power applications and cyber infrastructure. A classification was performed to highlight the dependencies between the cyber-physical smart grid controls. Huimin et al. designed a protection scheme for smart MVDC (Medium–Voltage DC) grid [14]. The self-healing ability against the measurement faults in power systems and protection systems enhances the fault resilience of the power grid. This method efficiently handles both measurement system fault and power system fault. Rongxing et al. proposed an Efficient and Privacy-Preserving Aggregation methodology for securing the smart grid communications [15]. This method encrypts the multi-dimensional data using a homomorphic Paillier cryptosystem technique. The authentication cost was significantly reduced by using a batch verification method. Calderaro et al. detected and localized the failures in smart grids using petri net (PN) modeling [16]. The detection of faults was modeled as matrix operations. This method enabled the fault identification the strong effect of distributed generation. Ui–Min et al. formulated open-switch fault detection in a grid-connected NPC (Neutral-Point-Clamped) inverter system [17]. This technique also determined the location of the fault besides detecting them. The fault condition was detected depending on the Concordia current pattern radius. The detection scheme does not require any additional sensors and identified the faulty switches within two fundamental periods. Jin et al. modeled a multi-terminal DC wind farm collection grid with protection design and internal fault analysis [18]. This method focused on DC faults and their transients. Lee et al. used quantum Genetic Algorithm (QGA) for solving the issue of economic dispatch in wind power generation of smart grids [19]. The QGA implemented quantum bit coding where the population of individuals contained an inherent diversity. The fittest solution for the optimization problem was easily obtained because the initial population did not have the need to be large to attain the consequence of species diversity. Lima et al. presented a control mechanism for the rotor-side converter (RSC) of wind turbines (WTs) during grid faults [20]. The power system faults in doubly fed induction generators (DFIG) were detected to control the initial overcurrent’s during the voltage saps. Katuri et al. designed an approach for reactive power optimization with voltage stability objective using Genetic Algorithm (GA) [21]. Reactive power optimization in the power system aims to maintain good voltage profile by enhancing the voltage quality other than decreasing the power loss. Ela et al. suggested a differential evolution algorithm for optimal reactive power research [22]. In this paper, a differential evolution (DE) optimization algorithm has been developed and applied to solve Reactive Power Dispatch (RPD) problem. Ali et al. [23] presented a fuzzy logic based system that monitors and controls Heating, Ventilation and Air-Conditioning Systems (HVACs) units, where demand exceeds the supplied electrical power. The proposed method performs better than these existing power grid models in terms of fault detection, degree of power

542

D. Vijayakumar, V. Malathi / Electrical Power and Energy Systems 62 (2014) 540–548

saving due to power optimization, memory usage, consumer-GW (gateway) communication overhead, consumer computational overhead, and critical time.

Control-theoretic Adaptation Contingency Analysis (CA)

Visualization (Vis.)

Requirements for a secure power grid State Estimation (SE)

The power grid must be designed securely at system level and device level. The following are the requirements for a secure power grid.

Sub-station 1

Sub-station 2 DSE

DSE

FACTS

FACTS

Real-time behavior

RTUs

RTUs

The high frequency (per 0.1–10 s) of event occurrence in the power grid makes the control devices incapable for real-time control and reconfiguration. The conventional data collection and contingency analysis are performed only for every few seconds and the state estimator (SE) executes only every 5 min, which cannot fulfill the ‘‘real-time condition’’. The factors which hinder the implementation of a real-time management scheme for smart grid are communication delay, computational overhead, and large amount of raw data. Real-time scheduling and routing algorithms are required to guarantee end-to-end real-time behavior. The fault detection and mitigation processes can be accelerated via distributed processing. Grid sensitivity Sensitivity can be implemented at both device level and system level, where the device-level response is quicker than the systemlevel response. Device-level response attributes to the equilibrium of the local power flow during emergencies, whereas system-level response focuses on local level responses and higher level stability. Adaptation-control techniques can be designed to enhance the dynamic sensitiveness to irregular system level variations. Fault resiliency Two types of faults namely, hardware faults and data faults can occur in a smart grid. Hardware faults occur due to system attacks when a section of the power grid requires to be executed even when separated. Data faults occur when the data are missing, incomplete or false due to the point attacks on the communication link. The design algorithms need to cope up with the variable data inputs to result in a correct output. Real-time management and evolutionary optimization scheme for a secure and flexible smart grid (RMOSFS) A secure, resilient and reconfigurable electrical grid is designed using fault-tolerant real-time controls to handle spontaneous and intentional attacks in the grid. The power grid is integrated with centralized system-level adaptation and embedded intelligence based load actuation to form an efficient smart grid. The sensor nodes (SNs) and the base station (BS) serve the purpose of smart meters in the grid. Location-centric hybrid system architecture A distributed control mechanism is considered, where the distributed controllers communicate the information with the primary peers to guarantee the reliability and safety of local operations. Distributed control is required because the higher distributed power generation, transmission, and distribution necessitates the distributed command and control. The location-centric

DG

DG

DG

DG

DG

DG

Fig. 1. Location-centric hybrid architecture.

hybrid system architecture is shown in Fig. 1. The filled circles and squares represent the embedded controllers. This hybrid architecture supports distributed control which differs from the conventional power grid in the following perspectives:  The generator and transmission devices are integrated with a controller to make executions based on collaborative information.  The collaboration levels at generator and transmission sides prevent isolated and coordinated attacks.  A distributed state estimator (DSE) is used at each substation to decrease SE execution time and furnish the feedback on local faults. The threshold value for the collaboration algorithm is 30 W.  The grid stability in terms of security is enhanced by an adaptation technique.

Security through reconfiguration of system The power grid is transformed into a reconfigurable grid by the incorporation of power electronic devices with control and sensing functions. Reactive power control and management is necessary for efficient voltage control/stability. Reactive power is more useful when it is supplied locally. The aim of intelligent controllers is to determine the quantity of real and reactive power to be employed in the grid under specific variations in detection of current, frequency, or voltage. The power grid is partitioned into several microgrids containing generators and loads separated from the primary grid. The microgrids can provide power at least to sensitive loads during contingency and system attacks. Normally, the generators operate in a current control mode where they are tightly synchronized with the primary grid. But, when there is power failure due to component attacks or the microgrid is disconnected from the primary grid, the generators need to transform to voltage control mode for providing constant voltage to at least the local sensitive loads. The need for an intelligent load shedding scheme relies on the determination of local voltage and frequency transients for each generator. Fig. 2 illustrated a microgrid with 3 DGs and 4 critical loads. Consider a microgrid with four generators (G1, G2, G3, G4) four sensitive loads (L1, L2, L3, L4). The total quantity of active and

D. Vijayakumar, V. Malathi / Electrical Power and Energy Systems 62 (2014) 540–548

543

(1) The overlap function possesses its maximum values from [ns–nf, ns], where ns represents the total number of sensors and f represents the total number of faulty device inputs. (2) The mid-value of the integration has to be equal to or greater than the median of the estimated interval. (3) Both the conditions (1) and (2) have to be fulfilled in two consecutive integrations excluding the first round so that decision stability is enhanced.

Grid

Feeder 1

Feeder 4

Distributed algorithm

Critical Load 1

Critical Load 4

DG 1

DG 4

Fig. 2. A microgrid with 4 DGs.

reactive power required by the load is given in (1) and (2) respectively.

X X

PL ¼ P L1 þ PL2 þ PL3 þ PL4

ð1Þ

Q L ¼ Q L1 þ Q L2 þ Q L3 þ Q L4

ð2Þ

The total quantity of active and reactive power released by the generators is given in (3) and (4) respectively.

X X

PG ¼ P G1 þ PG2 þ PG3 þ PG4

ð3Þ

Q G ¼ Q G1 þ Q G2 þ Q G3 þ Q G4

ð4Þ

The microgrid operates constantly even after power failure by balancing the conditions given in (5).

X

PL ¼

X

PG ;

X

QL ¼

X

QG

ð5Þ

The normal range for the operation of microgrids according to IEEE Std. 1547 is 0.88–1.1 pu. The switching of the generator inverter system to voltage control mode is a major issue. When the rate of voltage variation can be detected after the power failure, the amount of load shedding or reduction in generation can be determined before switching to voltage control mode. The local frequency and voltage is monitored to accelerate the responsiveness to changes in the system state. The controllers in the multiple generators crash because there is no information exchange between the generators, which does not accurately estimate the state of the microgrid. This problem is issued by the following collaboration algorithm. Collaboration algorithm When a controller is integrated with a power electronic device, the communication links to the controllers from the field remote terminal units (RTUs) can be damaged. This results in loss of data or false data. The local controllers need to execute correct decisions that do not affect the grid reliability or stability. A generic algorithm is developed to attain fault resiliency based on multiple controller coordination. The faults are effectively detected, prevented, and mitigated. A distributed interval integration algorithm [24] is used, where the generator is chosen to collect the device output and develop an overlap function. This algorithm guarantees that the mere variations in the input intervals result only in mere variations of the integrated result. The significance of this algorithm is the decreased width of the output interval for a case of a large number of devices. The following conditions are considered to determine the collaboration of the local controller with its neighbors:

The time consumption of the existing power grid models is quite high. The computation time can be decreased by using distributed and parallel processing, where the remote processor executes the local state estimation with the output transmitted to a control center to filter the computations. The output of the DSE can be utilized by the local controllers for effective analysis of the power grid state and feedback upon fault detection when the anticipated state is dissimilar to the estimated state. An asynchronous distributed algorithm is developed to reduce the processing time of the state estimation process. Here, the network is divided into numerous overlapping areas, where each area contains its own local processor which associates with the local measurements and estimates an approximate local state. The areas will have a particular overlap with each other for the following two purposes: 1. Two local estimators on the overlapping region can be utilized for result consolidation and discrepancy reduction. 2. To avoid the constraints in false data detection and identification. Each area composes its own individual state estimation with spatially and sequentially occurring iterations until the product converges to a desired tolerance level. Resource scheduling and reactive power management An important need for a power grid system is the adaptability to security attacks and unpredictable variations. A control-theoretic module is used to achieve the dynamicity to quality-of-service (QoS) attacks. The power grid system is modeled as controlled variables based dynamic system. The QoS attributes like real-time response time and throughput need to monitor the QoS specifications. When disturbances occur due to the QoS attacks, the devices may get damaged due to abnormal voltage, current, or frequency. A real-time decentralized scheduling algorithm is designed based on control theory for scheduling the energy resources in the grid and assuring the end-to-end real-time constraints. The merit of decentralized state estimator is that nodes can be easily added or deleted in the network without requiring drastic changes to the overall topology. The goal of the short-term energy resource management is to minimize the operational costs. When the grid is under attack, functions such as contingency analysis and state estimation are assigned high priorities, while the functions of data collection and backup are suspended or executed at a much lower task rates. Utilization control is used to employ the desired utilizations on multiple processors besides the uncertainties in the workload. There is a need for utilization control in real-time systems since the processor tasks are required to fulfill their deadlines. Utilization control can be modeled as a variable condition optimization problem. A utilization set point vector, U = [U1. . .Us]T, is considered, where Ux denotes the ideal utilization on processor Px. A rate constraint [Rmin,x, Rmax,x] is chosen for each task Tx and the controller can dynamically select the task invocation rate rx(i)

544

D. Vijayakumar, V. Malathi / Electrical Power and Energy Systems 62 (2014) 540–548

for each task to reduce the difference between the utilization ux(i) for the processors in the system and the set point Ux, under the rate constraint given in (6) and (7). ns X 2 ðU x  ux ðiÞÞ

min

ð6Þ

fr y ðiÞ1 6 y 6 ns g x¼1

Equality constraints

So that Rmin;y 6 r y ðiÞ 6 Rmax;y

ð7Þ

The rate constraints confirm that each function is within their acceptable ranges. The formulation of the optimization enhances the task rates by altering the utilization of each processor approximate to its set point as governed by the constraints. A decentralized end-to-end utilization algorithm is used to sustain the ideal CPU utilizations on multiple processors besides the uncertainties in task execution times. It consists of a centralized MIMO (multi–input– multi–output) controller based on MPC (model predictive control) theory. This is applied for the coordination and management of the adaptation process under the limitations of task rates. The overhead of a controller depends only on the neighborhood. The resource scheduling is designed according to the anticipation times. The reactive power management process is modeled as a constrained, large-scale, mixed, non-linear, and optimization problem as depicted in (8) and (9).

min

f ðs; cÞ

s 2 Rns



So that

ð8Þ

gðs; cÞ ¼ 0

ð9Þ

hðs; cÞ 6 0

In (8) and (9), f(s,c) represents the objective function, g(s,c) consists of the equality constraints, and h(s,c) consists of the inequality constraints. Mathematical model The primary factor for voltage instability is the inefficiency in the maintenance of an appropriate voltage level and reactive power management. The energy resource scheduling is modeled according to the following mathematical model identified as a non-linear mixed-integer problem. The objective function f(s,c) in (10) is formulated to determine the minimum operational costs in each period t. The constraint for consideration is given in (11).

min f ðs; cÞ ¼ min

" nT 1 X

PE1 ðT 1 ;tÞ :g E1 ðT 1 ;tÞ þ

T 1 ¼1

þ

nT m X

PE2 ðT 2 ;tÞ :g E2 ðT 2 ;tÞ þ . . .

T 2 ¼1

#

PEm ðT m ;tÞ :g Em ðT m ;tÞ

nT 2 X

8t 2 f1; . . . ; Tg

ð10Þ

T m ¼1

In (10), m represents the number of energy resources, E1, E2 . . . Em represents the various energy resources, T1, T2 . . . Tm represents the unit energy value of various energy resources, nT 1 represents the number of type-1 resources, represents the number of type-2 resources, nT m represents the number of typem resources, P represents the active power generation and g represents the generation cost. The power in (10) must be balanced in each period t. nT 1 X

PE1 ðT 1 ;tÞ :g E1 ðT 1 ;tÞ þ

T 1 ¼1

þ

nT 2 X

PE2 ðT 2 ;tÞ :g E2 ðT 2 ;tÞ þ . . .

T 2 ¼1 nT m X

P Em ðT m ;tÞ :g Em ðT m ;tÞ

T m ¼1

¼

nL X LoadðL;tÞ þ Ploss 8t 2 f1; . . . ; Tg L¼1

In (11), Load(L, t) represents the active power requirement of load L in period t and Ploss represents the total power losses in the distribution lines which is the 5% value of the former parameter. nL represents the number of loads.

ð11Þ

The primary equality constraints relate to the power flow equations in every communication link.

DPl ¼ P Gl  P Ll  P l

ð12Þ

DQ l ¼ Q Gl  Q Ll  Q l

ð13Þ

In (12) and (13), P Gl and Q Gl represent the real and reactive powers of generator G at the communication link l, respectively; PLl and Q Ll the real and reactive load powers, respectively. Pl and Ql are the power inputs at the node. Inequality constraints These constraints compose the operational constraints or physical boundaries of the components in the system. The following are the inequality constraints:  Voltage limits at communication links.  The limits of the loading factor k.  Generator input power limits. Genetic algorithm – optimal solution for energy resource management The meta-heuristic approach to optimization problems yields lower processing time. Genetic Algorithm (GA) [25] is used for the determination of optimal solution to combinatorial problems with the least amount of computational resources. The numerous generator sources are optimally located that enable reactive power ability spread over the network. Smart grids need a dynamic reactive power management system to enable reactive power regulation and voltage control. GA is applied in reactive power management. The initial step in the optimization process is to gather the number of available energy resources. The simple generational pseudo code for GA is given in the following textbox.

 Choose the initial population of individuals.  Evaluate the fitness of each individual in that population.  Repeat on this generation until termination (time limit achieved).  Select the best fit individuals for reproduction.  Breed new individuals through cross over and mutation.  Evaluate the individual fitness of new individuals.

The initial solution is obtained by selecting the generators that will possess lower costs. The initial population is chosen to the minimum cost level and the variables that are affiliated with this cost are set to ‘1’ while the others are set to ‘0’. GA is initiated with an initial population and it evolves to the further generation of individuals by executing reproduction, crossover, and mutation, between the individuals. The parameters for GA in this work are given in Table 1. The merits of GA [25] are:  This algorithm does not require any initial system information for the initiation of searching process.  Vastness in exploring the search space using iterative properties and multiple points of the population.

D. Vijayakumar, V. Malathi / Electrical Power and Energy Systems 62 (2014) 540–548

545

Mutation

Table 1 Choice of GA parameters. Parameter

Value

Size of population Number of generations Fitness scaling Probability of crossover Selection function Crossover function Mutation function Elitism

20 10,000 Rank 0.01 Tournament Cut and splice Boundary –

The goal of mutation is to introduce heterogeneity in the population of individuals by modifying the properties of the individuals from the reproduction. This process employs new information in the evolution process. In this work, boundary approach is applied to mutation operation. Update A new population is created in each generation. The individuals produced in the intermediate generations are updated in the population.

 The best fit individuals will be chosen among the parents and the offspring generation process delimits the convergence to a global minimum. The reactive power output will be limited by the amount of input active power and the rating of the coupling power converter. GA consists of the following steps as shown in Fig. 3.

Encoding

The convergence condition is checked till which the evolution process is continued and the best solution is finally obtained. The conventional convergence criteria include the maximum number of generations or the convergence tolerance between two sequential populations. Performance analysis

The genetic details of each individual are encoded in a chromosome, which is represented as a string of real numbers that contains:  Location of the communication link.  Input reactive power.  Loadability factor. Suppose, there are ng generators, the chromosome consists of 2ng genes composing the location of the communication link and one gene composing the optimal loadability factor.

Evaluation A fitness function (F) is assigned to each individual of the population to assess the goodness value. The chromosomes are initiated with a fitness value which estimates the individual adaptation ability and the survival probability in the further generations. In this work, F is equal to k. The scaling process is applied to the evaluated individuals using the range operation as given in (14).

pffi Range ¼ 1= l

Final solution

ð14Þ

Selection The individuals are chosen for reproduction process, where the probability of a specific solution being selected is proportional to its fitness value. The individuals with the highest fitness value are chosen as parents because they possess higher survival likelihood. In this work, tournament selection is applied to selection operation.

Crossover This operation is performed to recombine the properties of the selected solutions to generate new individuals from a couple of parents. A random point is chosen to perform the crossover. In this work, cut and splice approach is applied to crossover operation.

The real-time management and evolutionary optimization scheme for a secure and flexible smart grid (RMOSFS) towards sustainable energy is analyzed in terms of fault detection, degree of power saving due to power optimization, memory usage, consumer-GW (gateway) communication overhead, consumer computational overhead, and critical time. The voltage of the transmission line is considered as 35 kV. The existing methods considered for comparative analysis are fault detection and classification using Functional Analysis and Computational Intelligence (FACI) [26], Efficient and Privacy-Preserving Aggregation (EPPA) scheme [15], Lightweight Message Authentication (LMA) scheme [6], Multi-terminal DC Wind farm collection Internal Fault Analysis (MDWIFA) [18], and traditional scheme (TRAD) without any data aggregation [12]. Fault detection The occurrence of fault in the transmission is shown in Fig. 4. The X-axis represents time in seconds and the Y-axis represents the amplitude of voltage (mV) and current (mA). The pink colored waveform depicts the current, yellow colored waveform depicts voltage, and the blue colored waveform represents the fault reference line. The gap between the waveforms (0.02–0.11 s) is depicted as a fault. The voltage and current almost remain constant during the fault occurrence. The fault detection time for FACI and RMOSFS is analyzed and compared in Fig. 5(a). When the ratio of fault detection is equal to ‘1’, it depicts that the fault is fully depicted and when the ratio of fault detection is equal to ‘0’, it depicts that the fault is not at all depicted. FACI takes around 0.04 s to fully detect the fault, whereas RMOSFS detect the fault instantly but only at around t = 0.2 s. Fitness function evolution The fitness function evolution, i.e. Variation of fitness value with respect to the increasing number generation is shown in Fig. 4(b). The X-axis represents the generation and the Y-axis represents the fitness value. The green colored waveform represents the fitness value for voltage with respect to the generations and the blue colored waveform represents the fitness value for current with respect to the generations.

546

D. Vijayakumar, V. Malathi / Electrical Power and Energy Systems 62 (2014) 540–548

1

√

Data network formation Encoding and Initialization process

Encoding

Fitness evaluation for each individual So that Crossover

Mutation Evolution process Reproduction

Selection process

No Reached last generation? Yes Output the best solution

Fig. 3. Optimization process based on GA.

Fig. 4. Occurrence of fault in the transmission line, the gap between the waveforms (0.02–0.11 s) is depicted as fault.

Fig. 5. (a) Fault detection time for FACI and RMOSFS, and (b) fitness function evolution.

D. Vijayakumar, V. Malathi / Electrical Power and Energy Systems 62 (2014) 540–548

Fig. 6. Degree of power saving due to power optimization with respect to clock pulse.

Fig. 7. (a) Memory usage of LMA and RMOSFS, and (b) consumer-GW communication overhead of EPPA, TRAD, and RMOSFS.

Fig. 8. (a) Consumer computational overhead of EPPA, TRAD, and RMOSFS, and (b) critical time of MDWIFA and RMOSFS.

547

548

D. Vijayakumar, V. Malathi / Electrical Power and Energy Systems 62 (2014) 540–548

Power optimization and power saving The degree of power saving due to the power optimization with respect to clock pulse is shown in Fig. 6. The X-axis represents the time in seconds and the Y-axis represents the amplitude of the power (105 W). As the power is optimized, the power is saved relatively with respect to time.

GA. This technique performs better than the existing power grid models in terms of fault detection, degree of power saving due to power optimization, memory usage, consumer-GW (gateway) communication overhead, consumer computational overhead, and critical time. The future work consists of real-time implementation using hardware toolkits to verify the practical energy consumption and optimization performance.

Memory usage

References

The memory usage of LMA scheme and RMOSFS is analyzed and compared with respect to time in Fig. 7(a). It is observed that RMOSFS consumes lesser memory than LMA scheme.

[1] Gungor VC et al. Smart grid technologies: communication technologies and standards. Ind Inf, IEEE Trans 2011;7:529–39. [2] Fang X et al. Smart grid – the new and improved power grid: a survey. Commun Surveys Tutorials, IEEE 2012;14:944–80. [3] Pagani GA, Aiello M. The power grid as a complex network: a survey. Phys A: Stat Mech Appl 2013;392:2688–700. [4] Fadlullah ZM et al. Toward secure targeted broadcast in smart grid. Commun Mag, IEEE 2012;50:150–6. [5] Miao H, Junshan Z. A dependency graph approach for fault detection and localization towards secure smart grid. Smart Grid, IEEE Trans 2011;2:342–51. [6] Fouda MM et al. A lightweight message authentication scheme for smart grid communications. Smart Grid, IEEE Trans 2011;2:675–85. [7] Young-Jin K et al. A secure decentralized data-centric information infrastructure for smart grid. Commun Mag, IEEE 2010;48:58–65. [8] Rodri et al. A stationary reference frame grid synchronization system for threephase grid-connected power converters under adverse grid conditions. Power Electron, IEEE Trans 2012;27:99–112. [9] Chouder A, Silvestre S. Automatic supervision and fault detection of PV systems based on power losses analysis. Energy Convers Manage 2010;51: 1929–37. [10] Budischak C et al. Cost-minimized combinations of wind power, solar power and electrochemical storage, powering the grid up to 99.9% of the time. J Power Sources 2013;225:60–74. [11] Ericsson GN. Cyber security and power system communication-essential parts of a smart grid infrastructure. Power Delivery, IEEE Trans 2010;25:1501–7. [12] Xu L et al. Securing smart grid: cyber attacks, countermeasures, and challenges. Commun Mag, IEEE 2012;50:38–45. [13] Sridhar S et al. Cyber-physical system security for the electric power grid. Proc IEEE 2012;100:210–24. [14] Huimin L et al. Design of smart MVDC power grid protection. Instrum Meas, IEEE Trans 2011;60:3035–46. [15] Rongxing L et al. EPPA: an efficient and privacy-preserving aggregation scheme for secure smart grid communications. Parallel Distributed Syst, IEEE Trans 2012;23:1621–31. [16] Calderaro V et al. Failure identification in smart grids based on petri net modeling. Ind Electron, IEEE Trans 2011;58:4613–23. [17] Ui-Min C et al. Method for detecting an open-switch fault in a grid-connected NPC inverter system. Power Electron, IEEE Trans 2012;27:2726–39. [18] Jin Y et al. Multi-terminal DC wind farm collection grid internal fault analysis and protection design. Power Delivery, IEEE Trans 2010;25:2308–18. [19] Lee J-C et al. Quantum genetic algorithm for dynamic economic dispatch with valve-point effects and including wind power system. Int J Electrical Power Energy Syst 2011;33:189–97. [20] Lima FKA et al. Rotor voltage dynamics in the doubly fed induction generator during grid faults. Power Electron, IEEE Trans 2010;25:118–30. [21] Katuri R et al. Genetic algorithm optimization of generator reactive power. AASRI Procedia 2012;2:192–8. [22] Ela A et al. Differential evolution algorithm for optimal reactive power dispatch. Electric Power Syst Res 2011;81:458–64. [23] Al-Ali AR, et al., Smart grid controller for optimizing HVAC energy consumption. In: Computer Systems and Industrial Informatics (ICCSII), 2012 International Conference on; 2012. p. 1–4. [24] Hairong Q et al. A resilient real-time system design for a secure and reconfigurable power grid. Smart Grid, IEEE Trans 2011;2:770–81. [25] Alonso M et al. Integration of renewable energy sources in smart grids by means of evolutionary optimization algorithms. Expert Syst Appl 2012;39:5513–22. [26] de Souza Gomes A et al. Detection and classification of faults in power transmission lines using functional analysis and computational intelligence. Power Delivery, IEEE Trans 2013;28:1402–13.

Consumer-GW communication overhead The consumer-GW communication overhead of EPPA scheme, TRAD scheme and RMOSFS is analyzed and compared with respect to the number of consumers in Fig. 7(b). It is observed that RMOSFS consumes lesser overhead in terms of communication than EPPA and TRAD schemes. Consumer computational overhead The consumer computational overhead of EPPA scheme, TRAD scheme and RMOSFS is analyzed and compared with respect to the number of data dimensions in Fig. 8(a). It is observed that RMOSFS consumes lesser overhead in terms of computations than EPPA and TRAD schemes. Influence of fault distance on critical time Critical time is the limit for the total switchgear operation time. It should be small for the faster operation of the power grid. The critical time for MDWIFA and RMOSFS is analyzed and compared with respect to fault distance in Fig. 8(b). It is seen that RMOSFS performs the switchgear operations even when the faults are detected increasing distances from the control center. Conclusion Sustainable energy is very important for the future energy generation and management technologies. A real-time management and evolutionary optimization scheme for a secure and flexible smart grid towards sustainable energy is designed in this paper. The existing power grid models cannot obtain the real-time information about the transmission devices, perform the functions in an automated and coordinated manner, and respond quickly to emergency events (or) blackouts. This method consists of a location centric hybrid architecture following a collaboration algorithm to address the faulty and incomplete information. A distributed algorithm and system reconfiguration via power-electronics and switches is used for the real-time accessibility control, local state maintenance, and security enhancement. In this paper, Genetic Algorithm (GA) is chosen to address the reactive power management and optimum transmission device placement. The execution time of the resource scheduling is decreased by the optimization of