Microsoft Word - ETASR_V11_N1_pp6680-6686 Engineering, Technology & Applied Science Research Vol. 11, No. 1, 2021, 6680-6686 6680 www.etasr.com Phung & Le: Load Shedding in a Microgrid with Consideration of Voltage Quality Improvement Load Shedding in Microgrids with Consideration of Voltage Quality Improvement Bao Long Nguyen Phung Electrical and Electronics Department HCMC University of Technology and Education Ho Chi Minh City, Vietnam 16142018@student.hcmute.edu.vn Trong Nghia Le Electrical and Electronics Department HCMC University of Technology and Education Ho Chi Minh City, Vietnam trongnghia@hcmute.edu.vn Abstract-Microgrids have become more and more popular their usefulness as a renewable energy resource has been recognized. The core ability and promise of microgrids is addressing the environmental concerns due to climate change that have been growing during recent years. The innovation of microgrids is that they are designed to operate either in island mode or interconnected with the main grid system. However, when the microgrid operates in islanded mode, faults may occur which can cause system collapse or even blackout. Load curtailment schemes can be utilized to decrease the quantity of associated load to a level that can be securely supported by accessible generation in isolated mode. The main goal of this research is to evaluate the optimal amount of shedding power considering sustainable power sources, with the help of primary and secondary adjustments of the generator to restore the frequency to the allowed range. Particle Swarm Optimization algorithm is applied in this paper to determine the distributed shedding power on each demand load bus which can improve the voltage quality of the isolated microgrid system. The effectiveness of the proposed method is demonstrated through the simulation of IEEE 16- bus microgrid. Keywords-load shedding; islanded microgrid; primary and secondary adjustment; particle swarm optimization I. INTRODUCTION The imbalance between the power supply and demand loads, when a microgrid (MG) is separated from the main grid by fault or by automatically switching to islanded mode may lead to frequency and voltage droop. In this case, the MG may be overloaded or even collapse and bring the whole system to blackout [1, 2]. Load shedding schemes can operate to decrease the quantity of associated load to a level that can be securely supported by accessible generation during an emergency. The calculation of the amount of load shedding power is generally based on frequency droop [3]. If the calculation is deficient, it will not be able to restore the frequency to a permissible value and will cause over shedding. Researches on load shedding mainly calculate the amount of load shedding power based on the rotation motion of the rotor [4]. However, these strategies do not consider the actual operating conditions such as the primary and secondary control of the generating sets. The techniques of load shedding are divided into three main areas of study [5]: conventional load shedding, adaptive load shedding, and intelligent load shedding. The conventional load shedding is a method of load shedding by using Under Frequency Load Shedding (UFLS) [6], or Under Voltage Load Shedding (UVLS) relays. The UFLS method is known as one of the most used strategies. UFLS technique has a few disadvantages: the load is arbitrarily chosen on the grid when the frequency arrives at the load shedding limit. This technique is not effective when the load shedding areas are spread over the grid and in some cases over shedding or under shedding occurs [7]. Besides the researches on conventional techniques and their development, many intelligent algorithms are used in studying and making new load shedding schemes, for example fuzzy logic, Genetic Algorithm (GA), Artificial Neutral Networks (ANNs), Analytic Hierarchy Priority (AHP), Particle Swarm Optimization (PSO), etc. The fuzzy logic based load shedding scenario is efficient in restoring the frequency of an islanded MG system quickly [6, 8-12]. In some researches, fuzzy – AHP logic algorithm was used in load shedding strategies to measure the weight of load nodes in the system in various load levels [13]. The ANN-based load curtailment method can optimize the amount of load shedding in order to stabilize the power system [16]. A load curtailment based on a combination between ANN and neuro-fuzzy was proposed in [15]. The results show that the ANN can compute the total amount of active power that needs to be shed but it does not determine the number of loads or the distributed load shedding power that need to be shed in each demand load. Authors in [16, 17] proposed a load shedding method based on GA gain that can be used in the calculation of the optimal amount of shedding power. In fact, the GA can help in making the calculation speed for load shedding scenario robust, however the limitation of this algorithm is that it cannot help in optimal distributed load shedding for each demand load bus. The efficiency of PSO has been applied in load shedding techniques in [15, 18]. The studies show that PSO can forecast the frequency droop and optimize the amount of the required load demand. However, in this strategy, there’s still no limitation in the optimal shedding power. In [19], an UFLS technique based on PSO was studied, and the main objective was to calculate the optimal shedding power in order to stabilize the system frequency. However, the proposed method did not consider the voltage quality criteria in the system. Despite the fact that there are many studies that Corresponding author: Trong Nghia Le Engineering, Technology & Applied Science Research Vol. 11, No. 1, 2021, 6680-6686 6681 www.etasr.com Phung & Le: Load Shedding in a Microgrid with Consideration of Voltage Quality Improvement show the effectiveness of PSO in optimal load curtailment scheme, most of the methods are used in general or islanded grid systems, not on islanded MGs with renewable sources. The main objective of this paper is to plan an optimal load shedding strategy for islanded MGs with renewable energy sources during the islanded operation. The minimum shedding power will be computed based on the ability of primary control of the generators' governor and the reserve power of the generators for the secondary adjustment of the generators. The application of PSO algorithm will determine the optimal distributed shedding power of each load bus. And from that, it can optimize the voltage droop and improve the voltage quality. The effectiveness of this method is tested by the simulations of the IEEE 16-bus system, and the results are compared with a conventional under-frequency load-shedding scheme. Calculated and simulated results showed that the proposed method has less amount of shedding power than the UFLS relay method. Furthermore, the recovery of frequency was still held within the allowable values and maintained the stability of the power system. The current research can be used as a reference for MG system operators and electrical engineering graduate students when considering the issue of optimizing load shedding in the MG system when there is a disconnection from the main grid. II. THEORETICAL BACKGROUND A. Primary and Secondary Frequency Adjustment Under ordinary conditions and in the case of power losses in the framework, the power balance of the MG is introduced by: _ _ P P P P P main grid Gi G storage RES Lj + + + =∑ (1) where Pmain_grid is the power from the bulk grid, PGi is the power of the i th generator that has a speeder or a distributed generator with Auto Frequency Controller (AFC), ∑PLj is the real power of the j th load in normal operation, and PG_storage and PRES are the capacity of energy storage system and power of Renewable Energy Sources (RES). When the MG disconnects from the main grid, the first reaction of the system will be the primary frequency adjustment. It’s a reaction of a power source that has a speeder or an AFC. These frequency controllers will increase the power of generators in relation to the frequency [20, 21]. In Figure 1, we see the characteristics of the generators under normal operation before and after the connection loss of the MG from the main grid. Characteristic lines (F), (G) illustrate the characteristic of the load power in the normal operating conditions and during load shedding respectively. Characteristic lines (B), (D) show the output power of a generator without a governor corresponding to the normal condition, after disconnection from the main grid. The (A), (C) and (E) characteristics are the power. PGn and PGn-1 are the total value of the output power in the normal operating conditions and during the outage generator respectively, f0 is the rated frequency in the normal operating conditions, f1 is the frequency of the MG during the islanded operation (in the case that the generators have a governor and the distributed generators have a droop controller), f1’ is the frequency of the outage generator (in case of the generator without a governor), f2 is the frequency of the system after the primary and secondary control, and fallowed is the restored frequency (59.7Hz for power grids with a rated frequency of 60Hz). Fig. 1. The relationship between frequency and output power deviation. When the primary adjustment process for the power source is initiated, the response of the load to the frequency change is also performed. In addition, the RESs that respond slowly to frequency changes are considered not capable of adjusting the primary and secondary process, so they can be considered as negative loads. Equation (1) can be rewritten as: ( ) _ _ ( ) P P Gi primary P P P P P Lj main grid G storage RES L freq +∆ = − − + −∆∑ (2) where: 1 1 n f P primary Ri i − −∆ ∑∆ = = is the amount of primary control power of the system when fault occurs. In this case it will be the islanded operation mode of the MG. ∆f is the frequency droop when the MG turns to islanded operation. Ri is the ratio between the frequency and generator’s power. It also represents the adjustment of the speed of slip. ∆PL(freq) is the load whose power changes with the frequency. If the frequency will not go back to the allowed range after the primary adjustment process, the secondary adjustment will proceed. The secondary adjustment process is the process of the primary adjustment made through the effect of the Automatic Generating Center (AGC) system on a number of specified units, or a dispatch command [20, 21]. The balance power equation will be rewritten as: sec ( ) _ _ ( ) P P P Gi primary ondary P P P P P Lj main grid G storage RES L freq +∆ +∆ = − − + −∆∑ (3) where sec , P P P ondary Gn i primary ∆ = −∆ is the maximum secondary power adjustment of the generator. B. Particle Swarm Optimization PSO was first studied in 1995 [22]. In PSO, the expression "population" alludes to the particles which are dependent upon the best method of conduct. These particles can run in two Engineering, Technology & Applied Science Research Vol. 11, No. 1, 2021, 6680-6686 6682 www.etasr.com Phung & Le: Load Shedding in a Microgrid with Consideration of Voltage Quality Improvement modes: stochastic mode and deterministic mode. Mathematically, the searching process can be presented by simple equations using the position vector [ , ,...., ]1 2X x x xi i i in= and the velocity vector [ , ,...., ]1 2V v v vi i i in= according to the specific dimension of the searching space. In addition, the optimum of the solution in the PSO algorithm depends on the particle position and velocity update by the following equations: 1 . [ ] [ ] 1.1 2 2 k k k k k k V wV c r X X c r X X i i pbest i gbest i + = + − + − (4) 1 1k k k X X V i i i + + = + (5) where Vi k , Xi k represent the velocity and position of each particle i at iteration k, w is the inertia constant, ranging usually between 0 and 1, c1 and c2 are cognitive coefficients, ranging from 0 to 2, r1 and r2 are random values generated for each velocity update, and k X pbest , k X gbest are the global best positions earned in the swarm’s experience and the best position of each particle respectively. Each part of (4) is defined as: . k w V i is the inertia component, which is responsible for keeping the particle search in the same direction, [ ] 1.1 k k c r X X pbest i − is the cognitive component representing the particle’s memory. The cognitive coefficient c1 affects the step size of the particle to move toward its local best position Xpbest. [ ]2 2 k k c r X X gbest i − is the social component. It represents the movement of the particle towards the best region found by the swarm so far. The c2 factor is known as the social coefficient, and it also controls the step size of the particle to find the global best positon Xgbest. Equation (5) presents the situation of every individual particle by utilizing the new speed and its past position. In other words, a looking through cycle is re-initiated in order to locate the global ideal arrangement. The cycle rehashes itself until it meets one completion criterion, e.g. maximum number of repetitions. III. THE PROPOSED LOAD SHEDDING METHOD C. Construction of the Formula of Optimal Amount of Shedding Power When the primary and the secondary adjustment do not recover the frequency back to nominal stage, then load shedding is required. In that case, the power balance equation will be: r sec ( ) _ _ ( ) _ min P P P Gi p imary ondary P P P P Lj main grid G storage RES P P L freq shed +∆ +∆ = − − +∑ −∆ −∆ (6) 1 sec 1 ( ) _ _ .( ) _min fn allowedP P Gi ondaryRi i P P P P Lj main grid G storage RES D P shed ω ⇔ −∆− + +∆∑ = = − − +∑ − −∆ −∆ (7) where f f f allowed n droop ∆ = − is the allowed drop of the frequency, ∆Pshed_min is the minimum amount of load shed that has the ability to bring the frequency close to the nominal value, D is the characteristic factor change of load in percentage frequency change, ranging between 1% and 2%. The factor D is experimentally determined, e.g. if D= 2%, then if the frequency changes by 1%, it will lead to the change of loads by 2%. From the above equation, the minimum amount of load shedding can be determined as: _min [ ( ) .( )] _ _ 1 ( ) sec 1 P shed P P P P D Lj main grid G storage RES fn allowedP P Gi ondaryR i i ω ∆ = − − + − −∆∑ −∆− − + +∆∑ = (8) Fig. 2. Proposed load shedding method. The load shedding method is illustrated in Figure 2. The first step will be to determine all the necessary parameters of the system, including the parameters of loads and generators. Then the operation mode of the system will be considered as well. The next step to calculate the frequency droop, then primary and secondary power adjustment if necessary. After that, the minimum amount of load shedding and the amount of load shed at each bus will be determined by the PSO algorithm. After the amount of load shed is calculated, the load shedding process will be operated. And finally, the overall check must be processed to ensure the system is safe and secure. D. Construction of the Objective Function for PSO The objective function in the PSO algorithm is based on the voltage quality on each load bus. The main purpose of this function is to find the optimal amount of load shedding on each load bus if the MG turns into islanded mode in order to get the best result of voltage droop. This optimized load shedding scheme not only prevents the system from blacking out when the disconnection occurs, it also provides an improved way to shed load that avoids unnecessary loss. To obtain the objective function or voltage droop control function, the Jacobian matrix will be applied because of its resolution which is finding the amplitude of voltage and phase angle based on the real and Engineering, Technology & Applied Science Research Vol. 11, No. 1, 2021, 6680-6686 6683 www.etasr.com Phung & Le: Load Shedding in a Microgrid with Consideration of Voltage Quality Improvement reactive power in an electrical system. The basic formula of the Jacobian Matrix is shown below: 1 1 1 1 1 1 1 1 1 1 1 1 11 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 11 1 P P P P V V n n m P P P P P n n n n P V V n mn n Q Q Q Q Q V V n n m Q n Q Q Q Q n n n n V V n mn δ δ δ δ δ δ δ δ  ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ − −  ∆     ∂ ∂ ∂ ∂  − − − −   ∆ ∂ ∂ ∂ ∂ −− −  =  ∆ ∂ ∂ ∂ ∂      ∂ ∂ ∂ ∂  − −   ∆ −  ∂ ∂ ∂ ∂ − − − − ∂ ∂ ∂ ∂ −− … … � � � � � � � � � … … � � � � � � � � � 1 1 1 n V V n m δ δ           ∆                ∆   −  ×                       −             � � (9) where n is the number of nodes in the grid, m is the number of generators in the system, and n-m represents the loads in the electrical network. By converting (9) and ignoring the 1st generator from the bulk grid in the system because it will be removed after the islanded operation occurs, (10) can be claimed. Besides, in order to achieve the voltage droop equation, the J2 and J4 elements of the Jacobian matrix will be measured: 2 2 2 2 2 2 2 2 2 22 2 2 2 2 2 2 2 2 2 22 P P P P V V n n P P P P n n n n V V nnn V Q Q Q Q V V n n V n Q Q Q Q n n n n V V nn δ δ δ δ δδ δ δ δ δ  ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∆     ∂ ∂ ∂ ∂   − −   ∂ ∂ ∂ ∂ ∆   =   ∂ ∂  ∂ ∂     ∂ ∂ ∂ ∂         ∂ ∂ ∂ ∂ − − ∂ ∂ ∂ ∂  … … � � � � � � � � � … … � � � � � � � � � 1 2 2 P P n Q Q n −  ∆          ∆   ×   ∆           ∆                � � (10) The final function will be the voltage droop control on each bus: 1 1 2 4 V P J Q J i i i − −∆ =∆ × +∆ × (11) According to the power electrical triangle, tanQ P i i ϕ∆ =∆ × , so (12) can be written : 1 1 tan 2 4 V P J P J i i i ϕ − −∆ =∆ × +∆ × × (12) where |Vi| is the voltage droop on each bus, ∆Pi is the real power on the bus, 1 2 J − and 1 4 J − are the inverse J2, J4 elements of the Jacobian matrix, and tanφ the phase angle (tanφ=0.8). From (12), the detailed voltage droop equation can be obtained: 1 1 ....+ 2( ,2) 2 2( , ) 1 1 ......+ 2( , 1) 1 2( , ) 1 tan ...... 4( , 1) 2 1 + tan 4( ,2 2) V J P J P i i i m m J P J P i m m i n n J P i n J P i n n ϕ ϕ − −∆ = ×∆ + ×∆ − −+ ×∆ + ×∆ + + − + ×∆ × + + − ×∆ × − (13) In load shedding, the voltage droop will be enormous, and that is one of the most critical dangers of such schemes. This will make the system unstable and cause the whole system to collapse. There are still limitations in the control of that droop, therefore, the first thing that the objective function must do is to find the maximum value that can occur when the voltage falls, then from all the maximum values it must call out the least in the aforesaid index by using the PSO algorithm to make sure that the system is safe and secure after the disconnection. So, the objective function for PSO will be expressed as: 1 1 min(max( )) min(max( tan ) 2 4 V P J P J i i i ϕ − − ∆ = ∆ × +∆ × × (14) This algorithm has two main constrictions: • The amount of shedding power in each load must not be larger than its base power: 0<∆P