Tóm tắt Luận án Research and development a solar power system

el PWM Fig 2.5. The MPPT control diagram of PV 2.3 Conclusion of chapter 2 In this chapter, the author presented an overview of the solar power system including the main components of the system and the characteristics of PV solar. The author also points out the central outlines to develop solar power systems. The solar energy system has a big size so we have many problems to solve and develop for the whole perfect system. In this thesis, the author focuses on researching and developing the algorithm to find out the maximum power point of the MPPT to help the whole system can work efficiently and stably. Finally, the author also designed and manufactured an experimental system of grid-connected solar systems following the direction of developing the DC/DC and DC/AC power transformations. 7 CHAPTER 3 RESEARCH AND DEVELOPMENT OF CONTROLLER MAXIMUM POWER POINT TRACKING OF SOLAR POWER SYSTEM 3.1 Factors affecting MPP 3.1.1 The effect of solar irradiation Current according to irradiation as the formula (3.1). 𝐼 = ( 𝐺 𝐺0 ) 𝐼𝑆𝐶 − 𝐼0 [𝑒 𝑞𝑉𝑑 𝑛𝐾𝑇 − 1] (3.1) Fig 3.1. I - V, P - V and P - I characteristic curve with different radiation levels 3.1.2 The influence of temperature The short-circuit current is given by a formula of temperature (T) as (3.2). 𝐼𝑆𝐶(𝑇) = 𝐼𝑆𝐶(𝑇𝑟)[1 − 𝛼(𝑇 − 𝑇𝑟)] (3.2) I - V, P - V and P - I characteristic curve with different temperature levels as shown in Fig 3.2. As the temperature increases, the I - V characteristic shifts to the left, meaning the voltage decreases. The maximum power point (MPP) of the photovoltaic battery system has also changed. Therefore, it is required to have an algorithm to track MPP so that the system can work at MPP to reduce losses and improve performance for the whole system [111]. Fig 3.2. I - V, P - V and P - I characteristic curve with different temperature levels 8 3.1.3 The influence of shading Shade phenomena are defined when photovoltaic (PV) is partially covered, which can severely affect the performance of PV. Shading not only reduces the capacity of photovoltaic cells but also changes the open-circuit voltage Voc, short-circuit current Isc and their performance. Fig 3.3. I - V, P - V and P - I characteristic curve under partial shading condition 3.2 Maximum power point tracking An important component of the PV system is the maximum power point tracking (MPPT), which helps the PV system to generate the maximum output power of the system, reduce power loss and solve economic problems for the PV system. It acts as a power device that links the photovoltaic cell to the load, controlling the operating point of the photovoltaic battery system to obtain maximum power from the photovoltaic battery system with changing environmental conditions such as temperature, radiation, shade ... so the system performance is improved. 3.3 Research and development of incremental conductance algorithm for maximum power point tracking of photovoltaic system 3.3.1 MPPT by using Conventional Incremental Conductance algorithm INC algorithm diagram as shown Fig 3.4 [5]: Fig 3.4. Flowchart of the INC algorithm 9 If D is considered as a fixed control variable, the traditional INC algorithm flowchart is rewritten as shown Fig 3.5. Start Read V(k), I(k) dV = V(k) - V(k-1); dI = I(k) - I(k-1) dP = V(k)*I(k) - V(k-1)*I(k-1) D D(k) = D Dmax dV = 0 YesNo dI = 0 Yes D(k) = D(k-1)D(k) = D(k-1) Yes No dI > 0 Yes D(k) = D(k-1) + D D(k) D(k) = D(k-1) - D D(k) No V(k-1) = V(k); I(k-1) = I(k) Return No D(k) = D(k-1) - D D(k) D(k) = D(k-1) + D D(k) No Yes dI/dV = - I/V dI/dV > - I/V Fig 3.5. Flowchart of the INC algorithm with a fixed D variable In the algorithm flowchart of Fig 3.5 the step size is fixed. The fixed size jump affects the efficiency of the MPPT control algorithm INC. As mentioned above, the conventional MPPT methods based on fixed step size has a good performance. However, they are characterized by slow convergence; oscillations in the PV power around the MPP, operation fail under rapidly changing atmospheric conditions and they can get lost and track the MPP in the wrong direction during rapidly changing atmospheric conditions. Speedy tracking can be achieved with larger step size but excessive steady state oscillations are unavoidable, smaller step size can reduces the oscillations with slower dynamics. This limitation is overcome by the development of the INC algorithm with self-adjusting hop sizes that will be presented in the next section. 3.3.2 Proposed Variable Step Size INC algorithm to achieve faster response time The variable step size method proposed is given as follows: 10 𝐷(𝑘) = 𝐷(𝑘 − 1) ± 𝑁 ∗ | 𝑑𝑃 𝑑𝑉 − 𝑑𝐼 | (3.3) where N is the scaling factor adjusted at the sampling period to regulate the step size, which is manually adjusted in INC algorithm. The flowchart diagram of the MPPT INC algorithm has a change step shown in Fig 3.6. Start Read V(k), I(k) dV = V(k) - V(k-1); dI = I(k) - I(k-1) dP = V(k)*I(k) - V(k-1)*I(k-1) DD(k) = N*abs(dP/(dV-dI)) dV = 0 YesNo dI = 0 Yes D(k) = D(k-1)D(k) = D(k-1) Yes No dI > 0 Yes D(k) = D(k-1) + DD(k) D(k) = D(k-1) - DD(k) No V(k-1) = V(k); I(k-1) = I(k) Return No D(k) = D(k-1) - DD(k) D(k) = D(k-1) + DD(k) No Yes dI/dV = - I/V dI/dV > - I/V Fig 3.6. The flowchart of the INC algorithm with variable step size aims to converge quickly 3.3.3 Proposed Variable Step Size INC algorithm to achieve less oscillations When the system is in the steady state, the power variation values are not large, step size should make a satisfactory tradeoff to increased oscillations, will reduce the power loss for the system. In this proposed algorithm, the thesis also applies a jump value of decreasing size to zero, to reduce the fluctuation of capacity when reaching MPP. |𝑑𝑃| ∆𝐷𝑘 = ∆𝐷𝑘−1 𝛼 (3.4) 11 With the condition as shown in the expression (3.4), the next size step will be determined based on the variability of the previous step, and if the power varies slightly, then the step value will decrease with a division factor α, and this α value is usually chosen as 2. Start Read V(k), I(k) dV = V(k) - V(k-1); dI = I(k) - I(k-1) dP = V(k)*I(k) - V(k-1)*I(k-1) dV = 0 YesNo dI = 0 Yes D(k) = D(k-1)D(k) = D(k-1) Yes No dI > 0 Yes D(k) = D(k-1) + D D(k) D(k) = D(k-1) - D D(k) No V(k-1) = V(k); I(k-1) = I(k) Return No D(k) = D(k-1) - D D(k) D(k) = D(k-1) + D D(k) No Yes dI/dV = - I/V dI/dV > - I/V dP Yes No D D(k) = D D(k-1)/2 Yes D D(k) = N*abs(dP/(dV-dI)) Fig 3.7. The flowchart of the INC algorithm with variable step size in order purpose of reducing oscillation at MPP 3.4 Research and development Particle Swarm Optimization based MPPT algorithm 3.4.1 Particle Swarm Optimization (PSO) PSO is one of evolutionary computation techniques proposed by Eberhart and Kennedy in 1995 [48,49]. It is inspired by social and cooperative behaviour displayed by various species to fill their needs in a 12 multi-dimensional search space. In PSO, the updated positions of each particle in the search space is given by the following two equation as[65] : 1 , , 1 1 , , 2 2 ,( ) ( ) k k k k k k i j i j i j i j j i jV wV c r Pbest X c r Gbest X       (3.5) 1 1 , , , k k k i j i j i jX X V    (3.6) A detailed flowchart of PSO considering the above steps is shown in Fig.3.8. k = k + 1 If k ≤ maxite? No Yes Set PSO parameters Initialize population of particles with positions and velocity Evaluate initial fitness of each particle and select Pbest and Gbest Set iteration count k = 1 Update velocity and position of each particle Evaluate fitness of each particle and update Pbest and Gbest Print optimum values of generator output as Gbest Fig 3.8 Flowchart of PSO algorithm 13 3.4.2 Development Particle Swarm Optimization algorithm 3.4.2.1 Differential Particle Swarm Optimization (DPSO) The DPSO is a modified version of PSO, in which particles are capable of escaping from local minima in order to find a better optimization solution in the search space. The proposed DPSO considers an additional feature in the classical PSO. The additional feature is the opinion of one of the particles selected randomly from the swarm. The randomly-scaled difference of the particle and its opinion-giver particle is included in the velocity equation of the particle necessary to escape from local minima. Mathematically, the concepts of DPSO can be expressed as follows: 1 ( ) ( ) ( ), , , , , ,11 2 2 3 3 , k k k k k k k kV wV c r Pbest X c r Gbest X c r X Xp q p q p q p q q p q p ql q         (3.11) 1 1 , , , k k kX X Vp q p q p q    (3.12) In eqn. (3.11), c3 is the scaling factor and r3 is a randomly-generated random number between 0 and 1, whereas l represents the expert particle corresponding to target particle p. In this equation, l varies from 1 to N but l ≠ p. Fig.3.9 shows the search mechanism of the proposed DPSO in a multidimensional search space. Gbestk Xp k Xp k+1 Vp Gbest Pbestp k Vp k Vp Pbest Vp k+1 Vp Diff Fig 3.9 Proposed DPSO search mechanism of pth particle at kth iteration in a multi-dimensional search space In Fig. 3.9, Pbestkp,q represents personal best qth component of pth individual, whereas Gbestkq represents qth component of the best individual of population up to iteration k. It is found from Fig. 3.9 that the proposed DPSO is performed by adding one more term (VpDiff) in the velocity equation, thus the MPP can be obtained much sooner than that using the classical PSO. Furthermore, this additional feature allows the particles to escape from a local optimum in order to search for a better solution in the other reasons in the search space. 14 Application of DPSO to MPPT As mentioned in chapter 2, photovoltaic cell characteristics are non- linear, The output of power system is varied with the variation of light radiation and temperature, therefor reducing the efficiency of PV solar. In addition, the effectiveness of these photovoltaic modules is not satisfactory, thus requiring the assistance of intelligent algorithms to overcome this problem. In this thesis, a model using the DPSO-based MPPT algorithm is proposed to improve the performance of PV. The complete flowchart for the proposed method is shown in Fig. 3.10: Yes Set w, c1 , c2,c3 , n Sense V(i), I(i). Calculate initial fitness of each particle: P(i)=V(i)*I(i) Evaluate initial fitness of each particle and select Pbest and Gbest Set iteration count k =1 Update velocity and position of each particle Evaluate fitness of each particle and update Pbest and Gbest If k <= Maxite ? Print optimum values of the duty cycle as Gbest k = k+1 No Initialize positions and velocity of each particle d=d0; v=0.1*d0 Fig 3.10. Flowchart of the proposed MPPT - DPSO 15 3.4.2.2 Perturbed Particle Swarm Optimization Algorithm (PPSO) The PPSO is a modified version of PSO, perturbation in the velocity vector of each particle needs to be performed whenever the particles get struck into a local optimum. Normally, this situation occurs when the solution (Gbest of the swarm) does not be improved for a pre-specified number of iterations, such as the tolerance or a stopping criterion (other than maximum iteration) is met. For this, the velocity vector of each particle needs to be reset, so that particles can get a big thrust to push them to escape from the local optimum. Mathematically, the perturbation concepts of velocity for each particle can be expressed as follow: 1 , , k k p q p qV pbest   (3.13) Besides, the tolerance limit needs to be relaxed, so that the swarm can get another chance to resume the search for a certain minimum number of iterations. Also, such perturbation may be allowed for a certain number of times. Consequently, the proposed perturbation in the velocity vector of each particle allows the particles to keep exploring the search space in order to escape from the local optimum. A typical distribution pattern of each particle with larger length of arrows in PPs is shown in Fig. 3.11. Fig 3.11 Dynamics of particle in the multi-dimensional search space in PSO algorithm Application of PPSO to MPPT The overall flowchart of the proposed PPSO-based MPPT approach is shown in Fig. 3.12. Firstly, the duty cycle value d of the dc–dc converter is selected as the particle position in the search space. Meanwhile, the PV module output power is defined as the fitness value evaluation function. It can be noted that the number of particles should be selected as the number of the series LCP1 LCP2 LCP3 LCP4 LCP5 IP2 IP1 IP3 IP5 Local optimum IP4 PP5 IP4 PP1 PP3 Global optimum PP3 PP4 IP: Initialized Particle PP: Perturbed Particle LCP: Locally Converged Particle Arrow: Represents direction of move Length of arrow: Values of velocity 16 connected cells in PV array. Next, the PPSO algorithm will start the optimization process from a random initial value that is chosen in the following range: Dik = [D1, D2, D3, DN] Yes Set w, c1 , c2,c3 , n Sense V(i), I(i). Calculate initial fitness of each particle: P(i)=V(i)*I(i) Evaluate initial fitness of each particle and select Pbest and Gbest Set iteration count k =1 Update velocity and position of each particle Evaluate fitness of each particle and update Pbest and Gbest If k <= Maxite ? Print optimum values of the duty cycle as Gbest k = k+1 No Initialize positions and velocity of each particle d=d0; v=0.1*d0 If tol<tol_max Yes tol=1 V=Pbest tol = Fb k - Fb1 k+1 No Hình 3.12 Flowchart of the proposed PPSO-MPPT 17 Where N is the number of particles and k is the number of iterations. It means the particles are initialized on random positions, which cover the search space [Dmin, Dmax]. Dmax and Dmin are the maximum and minimum duty cycle of the utilized dc-dc converter, respectively. Then, the digital controller sends the PWM command according to the duty cycle, which also represents the position of particle i, the PV voltage VPV and current IPV are defined. These values are the basis of calculating the fitness evaluation of particle i (the PV module output power). The fitness of each particle is evaluated   k kp pF f X , ∀p and then finding the best particle index b. Finally, the PPSO algorithm updates Pbest, Gbest until the optimum value of the duty cycle, known as Gbest is obtained. It can be noted that once tol < tol_max then V = Pbest and tol =1, and else tol = Fbk - Fb1k+1. 3.5 Conclusion of chapter 3 In the PV system, it is always desirable whether in any weather condition, the power output from the PV array to the load is always the maximum, which is the goal of MPPT control problem. Chapter 3 the author studies the common maximum power control algorithms and makes development improvements for the algorithm to apply in MPPT control of PV system to make the system operate better performance. In this chapter, the author introduces some algorithms for MPPT controller and research INC algorithm and makes development improvements for INC algorithm to make the system reach the maximum power point faster, fluctuate around the point, maximum power is narrower and less, minimizing the loss of transmitted power due to fluctuations around the maximum power point. In this chapter 3, the author also focused on the PSO algorithm applied in MPPT control and proposed new algorithms by developing based on the classic PSO algorithm that is DPSO algorithm and PPSO algorithm to improve the operational efficiency of the PV system. 18 CHAPTER 4 SIMULATION VERIFICATION MPPT ALGORITHMS EVELOPMENT SYSTEM FOR SOLAR ENERGY POWER 4.1. Simulation results of improve INC algorithm 4.1.1. Simulation diagram Fig 4.1. Simulation of INC algorithm 4.1.2 Simulation results 4.1.2.2 INC algorithm to achieve faster response time The author conducted the simulation according to the proposed theory, resulting in the response of the PV system compared to the conventional INC as shown in Fig 4.2. a) Comparison of current response b) Comparison of voltage response 0 2 4 6 8 10 12 14 0 5 10 15 20 25 Time(s) I( A ) CURRENT Traditional I Variable I 0 2 4 6 8 10 12 14 0 50 100 150 200 250 300 350 400 Time(s) V (V ) VOLTAGE Traditional V Variable V 19 c) Comparison of power response Fig 4.2. Compare the responses of the conventional INC algorithm and variable size step Fig 4.3 D response of the conventional INC algorithm and variable size step at start Fig 4.4 P response of the conventional INC algorithm and variable size step at start Fig 4.5 P response of the conventional INC algorithm and variable size step when the radiation changes from G = 700W / m2 to G = 900W / m2 Through simulation results, it is easy to observe the speed of MPP reaching at the beginning of INC operation. The variable step case much faster than the conventional INC shown in Fig 4.4. Even in the case of a 0 2 4 6 8 10 12 14 0 1000 2000 3000 4000 5000 6000 7000 8000 Time(s) P (W ) POWER Theory P Traditional P Variable P 0 0.5 1 1.5 2 2.5 3 3.5 4 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 Time(s) D u ty DUTY CYCLE Traditional D Variable D 0 0.5 1 1.5 2 2.5 3 3.5 4 0 1000 2000 3000 4000 5000 6000 7000 Time(s) P (W ) POWER Theory P Traditional P Variable P 4 4.05 4.1 4.15 4.2 4.25 4.3 4.35 4.4 4.45 4.5 5800 6000 6200 6400 6600 6800 7000 7200 7400 7600 7800 Time(s) P (W ) POWER Theory P Traditional P Variable P 20 radiation change from G = 700W / m2 to G = 900W / m2, the improve INC algorithm is MPPT better, as shown in Fig 4.5, we can see the variable step INC fluctuating changes around MPP are lighter than conventional INC. 4.1.2.3 INC algorithm to reduce oscillations The results obtained from Matlab software simulation as shown in Fig 4.6 show that this proposed plan has a fast convergence rate and a significant reduction in oscillation at MPP. a) Comparison of current response b) Comparison of voltage response c) Comparison of power response Fig 4.6. Compare the responses of the conventional INC algorithm and variable size step to reduce oscillations 0 2 4 6 8 10 12 14 0 5 10 15 20 25 Time(s) I( A ) CURRENT Traditional I Modify Variable I 0 2 4 6 8 10 12 14 0 50 100 150 200 250 300 350 400 Time(s) V (V ) VOLTAGE Traditional V Modify Variable V 0 2 4 6 8 10 12 14 16 0 1000 2000 3000 4000 5000 6000 7000 8000 Time(s) P (W ) POWER Theory P Modify Variable P 21 a) D response at start b) D response when radiation change Fig 4.7. Graph D when transient and increasing radiation From Fig 4.7 a) we see that, with a variable step value, the system will disturb longer steps, limited to 10%, and as close to the maximum power value, the small step value is gradually and approaching zero, where the descending value α is chosen as 2, which means that the algorithm will continuously reduce the hop value for 2 until it reaches a value near zero. It rudece significantly the oscillation at MPP In order to compare the efficiency among the methods, the author redraws the power lines at the start-up time as shown in Fig 4.8 a), radiation change as shown in Fig 4.8 b) and the stabilization system in Fig 4.8 c). a) P response at start 0 0.5 1 1.5 2 2.5 3 3.5 4 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 Time(s) D u ty DUTY CYCLE Traditional D Modify Variable D 3.9 3.95 4 4.05 4.1 4.15 4.2 4.25 4.3 4.35 4.4 0.46 0.47 0.48 0.49 0.5 0.51 0.52 0.53 0.54 0.55 Time(s) D u ty DUTY CYCLE Traditional D Modify Variable D 0 0.5 1 1.5 2 2.5 3 3.5 4 0 1000 2000 3000 4000 5000 6000 7000 Time(s) P (W ) POWER Theory P Traditional P Variable P Modify Variable P 22 b) P response when radiation change c) P response when stabilization state Fig 4.8 Compare power response of three algorithms and theory From the comparision of simulation results, it has been proved that the two proposed algorithms have the ability to improve the performance of the conventional INC algorithm in terms of response time, oscillation at MPP and power loss. This proves the suitability of the proposed algorithm compared with conventional algorithm. 4.2. Simulation results of PSO algorithm and improve PSO algorithm 4.2.1 Simulation results of MPPT-PSO Following the flowchart of PSO algorithm in Fig 3.8, the author simulates the output power responses of the PV system and compares with the case without MPPT controller as shown in Fig 4.9. From the output response we see with the PSO algorithm that the output power response is not fluctuated as the INC or P&O algorithm . Hình 4.9 The response time of output power using the PSO 4 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 5 5800 6000 6200 6400 6600 6800 7000 7200 7400 7600 7800 Time(s) P (W ) POWER Theory P Traditional P Variable P Modify Variable P 11 11.2 11.4 11.6 11.8 12 12.2 12.4 12.6 12.8 13 6660 6680 6700 6720 6740 6760 6780 6800 6820 6840 6860 Time(s) P (W ) POWER Theory P Traditional P Variable P Modify Variable P 23 4.2.2 Simulation results of MPPT-DPSO The proposed DPSO algorithm has been tested on five benchmark problems to prove the effectiveness of the developed algorithms in solving complex optimization problems. Fig. 4.10 shows the response time of output power under constant environmental condition (the solar insolation G = 900 W/m2 and the temperature T = 25°C) in case of without MPPT controller, using DPSO and the conventional methods. Hình 4.10. The response time of output power using the DPSO and the conventional methods To demonstrate the superiority of the DPSO, a comparison of tracking the optimization point using the PSO is performed for the test PV system, as shown in Fig. 4.11. It can be clearly seen that both of the AI algorithms can find the optimization of centralized arrays without fluctuation around the MPP, but the improved tracked faster. The optimization time of the DPSO and PSO is 0.57 s and 2.12 s, respectively. Fig 4.11. The response time of output power using the DPSO and PSO Fig 4.12 shows the dynamic characteristic of the output power under solar radiation varied between 700 W/m2 and 800 W/m2 at a fixed temperature of 25°C. 24 Fig 4.12 The dynamic characteristic of output power during rapidly increasing radiation values Fig 4.13 The dynamic characteristic of output power during rapidly decreasing radiation values The results of the different MPPT controllers under fluctuations of irradiation are summarized in Table 4. It is observed that the power produced by using the proposed algorithm was greater than 99% under all test conditions. Table 4.1 The obtained results of output power without/with the MPPT controller G (W/m2) Without MPPT P&O InCond PSO DPSO The theoretical value of PV 600 4567.0 5137.0 5137.4 5157.2 5157.5 5157.7 700 5913.0 5994.8 5995.0 6009.0 6009.2 6009.7 800 6820.0 6812.0 6812.3 6849.1 6849.5 6850.0 900 7360.0 7655.0 7656.0 7677.2 7678.0 7678.3 4.2.3 Simulation results of MPPT-PPSO The proposed PPSO algorithm has been tested on five benchmark problems to prove the effectiveness of the developed algorithms in solving complex optimization problems, and then it is used to detect the global 25 maximum power point considering the phenomenon of shade (PSC). In addition, these results are compared with the traditional PSO algorithm to prove the effectiveness of the proposed method. For the purpose of tracking MPP, the proposed PPSO-MPPT algorithm is performed on a model that consists of a PV solar panel connected to a resistive load through a boost converter with MPPT controller. The PV array consists of the four PV modules which were connected in series, as shown Fig. 4.14. Fig 4.14. Simulation model of four PV models connected in series performed in Simulink Firstly, the simulation of a PV array under PSC, in which the PV array is not connected to a boost converter and MPPT algorithms is performed. The corresponding maximum powers generated by the PV array under different combinations of solar irradiance patterns are shown in Table 4.2. Table 4.2. The solar irradiance patterns for G1 to G4 Case The solar irradiance G1