MPPT control method of offshore floating photovoltaic power generation system

Abstract

An MPPT control method of a marine floating photovoltaic power generation system comprises the following steps: a dynamic updated marine floating photovoltaic power generation system maximum power point history optimal database is established by using time-varying periodicity of a marine floating photovoltaic array maximum power point; an initial population distribution of a biological heuristic algorithm is optimized by using the maximum power point history optimal database and a support vector machine method, and an elimination mechanism of an undesirable individual in an iteration process of the biological heuristic algorithm is improved; a hybrid improved MPPT control method is proposed by combining the improved biological heuristic algorithm with a traditional MPPT technology; the hybrid improvement is carried out by taking an improved cuckoo and a variable step size perturbation and observation method and an improved grey wolf algorithm and a variable step size conductance increment method as examples; and the control technology method is further illustrated by combining actual engineering data of a certain sea area; and the marine floating photovoltaic array can meet the requirements of maximum power tracking speed and precision under various sea conditions.

Description

(I) TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic power generation, and particularly relates to an MPPT (Maximum Power Point Tracking) control method of a sea floating photovoltaic power generation system. (II) BACKGROUND

[0002] Compared with land photovoltaic, there are no tall buildings such as trees around the sea floating photovoltaic, the light resource is sufficient, the air is humid and the precipitation is more, which is beneficial to the heat dissipation and cleaning of the photovoltaic module, and after the high-weather-resistant double-sided glass photovoltaic module is used, the sea floating photovoltaic power generation capacity is theoretically much higher than that of inland photovoltaic power station. However, the electric energy production of the sea floating photovoltaic power generation system is not only affected by the light resource and the photovoltaic module power generation efficiency, but also related to the sea wind and wave environment and the MPPT efficiency and other factors.

[0003] Affected by the sea wind, ocean current and irregular wave, the sea floating photovoltaic module moves with the floating body structure in six degrees of freedom, including the translation movement and rotation movement along the axis direction. Therefore, the radiation energy of the sea floating double-sided glass photovoltaic module is not only affected by the weather environment (such as cloud movement) and the installation angle of the module, but also changes dramatically with the increase of the sea state level. The complex and changeable marine environment will lead to the rapid fluctuation of the P-V curve of the photovoltaic array connected to the same MPPT and the possible appearance of multiple peaks, which may reach dozens of peaks, increasing the technical difficulty of the photovoltaic array MPPT and hindering the efficient and economic operation of the sea floating photovoltaic power station.

[0004] Methods to mitigate the effects of rapid and uneven irradiance changes in photovoltaic arrays can currently be categorized into two types: adding hardware facilities and employing software algorithms. However, the harsh operating environment at sea necessitates that the electrical equipment within offshore photovoltaic power plants possess extremely high weather resistance and reliability. Adding circuits and reconfiguration methods increase the electrical control devices of the power plant, not only increasing system complexity and cost but also raising the likelihood of system failures during operation. Therefore, improving MPPT efficiency should be approached from a software algorithm perspective. Software methods can be broadly classified into two categories. The first category comprises traditional MPPT techniques, primarily including perturbation observation, conductance increment, and fractional open-circuit voltage methods. Traditional MPPT techniques are popular and operate effectively under stable and uniform irradiance, but they cannot track the rapidly changing solar radiation and MPPT under partially uneven irradiance conditions of floating photovoltaic modules at sea. The second category mainly utilizes algorithms based on fuzzy logic control, artificial neural networks, and swarm intelligence. However, fuzzy logic control and artificial neural network computation methods require designers to possess a large amount of data for training, making them unsuitable for early-stage project development and research. Therefore, in order to promote the large-scale development and utilization of offshore floating photovoltaic power plants and ensure their safe operation and economic production, it is urgent to propose maximum power tracking technology suitable for offshore floating photovoltaic power plants in the project area. This technology will not only be useful for the early site selection survey of the project, but also help improve the economic efficiency of the project's production and operation. (III) Summary of the Invention:

[0005] The purpose of this invention is to provide an MPPT control method for a floating photovoltaic power generation system at sea. This method can overcome the shortcomings of existing MPPT control technologies and is a simple and easy-to-implement MPPT control method. This method reduces the complexity and cost of the floating photovoltaic system at sea and can effectively improve the reliability of the system during operation and the efficiency of the MPPT process.

[0006] The technical solution of this invention: an MPPT control method for a floating photovoltaic power generation system at sea, characterized by comprising the following steps:

[0007] (1) Utilizing the movement of seawater within a defined sea area at multiple time scales and the time-varying periodicity of the maximum power point of the floating photovoltaic array, a dynamically updated historical optimal database of the maximum power point of the floating photovoltaic array is established (d GMPPi V GMPPi ,P GMPPi ,t i This reduces the initial population distribution range of traditional biologically inspired algorithms, where d GMPPi For a historical moment t i The corresponding historical best duty cycle, V GMPPi For a historical moment t i The corresponding historical best voltage, P GMPPi For a historical moment ti The corresponding historical maximum power;

[0008] In step (1), by establishing a dynamically updated historical optimal database of the maximum power point of the floating photovoltaic array at sea, and narrowing the initial distribution range of the population in the traditional bio-inspired algorithm, specifically:

[0009] (1-1) The variations in irradiance and temperature received by offshore floating photovoltaic arrays are related to factors such as the Earth's rotation, the Earth's revolution around the sun, and seasonal winds and waves. The temperature varies with different working hours (t) within the same quarter. i There may be similar maximum power points, denoted as (d GMPPi V GMPPi ,P GMPPi ); because the MPPT algorithm requires the historical best duty cycle d to output GMPPi Value and time t i The correlation is significant; therefore, the global maximum power point searched during the operation of the offshore floating photovoltaic array is recorded in order to establish a dynamically updated historical optimal database of the maximum power point (d). GMPPi V GMPPi ,P GMPPi ,t i );

[0010] (1-2) Considering that the maximum power point of the photovoltaic array has certain seasonal variation characteristics, in order to reduce the storage size of the historical best database of the maximum power point and reduce the initial distribution range of the population of the bio-inspired algorithm, when the data size reaches the upper limit set by the historical best database of the maximum power point, the number of optimal data points will not be increased, and the individual with the highest dispersion in the historical best database of the maximum power point will be replaced by dynamic updates.

[0011] Steps (1-2) specifically refer to:

[0012] (1-2-1) Assume that the number of maximum power points stored in the historical best database of maximum power points is m. c Based on the time sequence of data entry, these data are divided into two parts: an online database and an offline database according to the ratios k1 and 1-k1.

[0013] (1-2-2) After each MPPT optimization process, record the latest maximum power point (d). GMPPi V GMPPi ,P GMPPi ,t i ), and store it in an online database;

[0014] (1-2-3) When storing the latest maximum power point in step (1-2-2), if the number of items stored in the online database has reached the upper limit value k1m cThen, a historical maximum power point is randomly selected from the online database and migrated into the offline database;

[0015] (1-2-4) In step (1-2-3), when randomly selecting a historical maximum power point from the online database and migrating it to the offline database, if the offline database storage capacity has reached the upper limit (1-k1)m c Then calculate the data in the offline database (d) GMPPi V GMPPi ,P GMPPi The variance of the data is used to remove the maximum power point with the largest variance (i.e., the highest dispersion) from the offline database, thus providing space for data migration to the online database. Simultaneously, removing the data with the highest dispersion from the offline database can filter out erroneous data collected in actual data collection and a very small number of extreme condition data within the seasonal cycle, reducing the historical time t in the historical optimal database of the maximum power point. i The corresponding historical optimal duty cycle d GMPPi The distribution range;

[0016] (2) Using the dynamically updated historical optimal database of maximum power points provided in step (1), and based on probability statistics and support vector machine methods, optimize the initial distribution position of the n individuals in the population at time 0, i.e., the initial duty cycle d1. 0 ~d n 0 This reduces power fluctuations and tracking time in the early stages of maximum power tracking;

[0017] Step (2) specifically includes the following steps:

[0018] (2-1) Considering that the purpose of the MPPT algorithm is to provide a suitable duty cycle d, thereby helping the photovoltaic array output to operate at the maximum power point; therefore, the partitioning of multidimensional data in the historical optimal database of the maximum power point can be simplified by the support vector machine method, and the historical optimal database of the maximum power point (d) obtained in step (1) can be extracted. GMPPi V GMPPi ,P GMPPi ,t i (Historical moments t) i The corresponding historical optimal duty cycle d GMPPi The historical optimal duty cycle d at this moment is obtained using the support vector machine method. GMPPi The array is used to classify one-dimensional data, thus eliminating the need for pre-training the support vector machine (SVM) model. When performing partitioning, the SVM only needs to find the historical time t. i The corresponding historical optimal duty cycle d GMPPi The maximum interval between adjacent data points can be used as the basis for dividing the region, and the average value of adjacent data points can be used as the basis for dividing the region; therefore, the support vector machine method can be used to divide the historical time t.i corresponding historical optimal duty cycle d GMPPi According to the size of the numerical value, it is divided into z partitions, respectively recorded as: [d0, d1], [d1, d2]…[d z-1 ,d z ];

[0019] (2-2) Statistics of the number (N1-N z ) of historical optimal duty cycles d GMPPi in z partitions [d0, d1], [d1, d2]…[d z-1 ,d z ] obtained in step (2-1), and based on the obtained historical optimal duty cycle number (N1-N z ) of each partition, the probability distribution rate F j of the historical optimal duty cycle d GMPPi in the j interval [d j-1 ,d j ] is calculated; finally, according to the probability distribution F j of the interval, the number D j of randomly distributed populations in the j interval at the initial moment of the biological heuristic algorithm is determined, as shown in formula (1):

[0020]

[0021] D j =nθ c F j ,j∈{1,2,3...z}

[0022] In formula (1), n is the population number of the algorithm; θ c is a constant in the interval [0, 1];

[0023] Considering that the corresponding duty cycle of the maximum power point may appear outside the range of the minimum and maximum values of the historical optimal duty cycle d GMPPi , i.e. min(d GMPPi )~max(d GMPPi ), not all individuals in the population are determined by the historical optimal database to determine the initial position, only the total number of nθ c individuals are determined by formula (1) to determine the position; the remaining number of individuals is n(1-θ c ), which is randomly distributed in the region [0, min(d GMPPi )] and [max(d GMPPi ), d uplimit ]; steps (2-1) and (2-2) can be used to complete the initial distribution of the biological heuristic algorithm population, i.e. to obtain the initial distribution position of the n population individuals at time 0, i.e. the initial duty cycle d1 0 ~dn 0 ;

[0024] The value range of the constant θ in the step (2-2) is generally 0.55-0.75, if θ>0.75, the population outside the range of min(d c )~max(d GMPPi ) is less, and the maximum power point of the photovoltaic array output may be missed; if θ<0.55, the role of the historical optimal database of the maximum power point in the algorithm will be weakened, and the convergence time of the algorithm will be prolonged. GMPPi

[0025] (3) improving the elimination mechanism of the bad individuals in the iteration process of the biological heuristic algorithm from the perspective of the elimination probability and the generation mechanism of the new population individuals after the elimination of the bad individuals, and improving the convergence speed in the middle term of the maximum power tracking;

[0026] The step (3) of improving the biological heuristic algorithm is based on the initial optimized distribution positions d i 0 ~d n 0 of the population individuals at 0 time provided in the step (2), and the improved biological heuristic algorithm is realized, and specifically includes:

[0027] ① setting the number of nests n, the elimination probability p a and the iteration termination criterion,

[0028] ② confirming the initial positions of the n nests in the improved biological heuristic algorithm, i.e. the initial duty cycles d1 0 ~d n 0 , by the initial distribution positions d1 0 ~d n 0 of the population provided in the step (2);

[0029] ③ applying the nest positions at t time, i.e. the duty cycles d1 t ~d n t (t≥0) to the DC-DC rectifier units connected to the offshore floating photovoltaic array in turn, and the DC-DC rectifier units output the corresponding offshore floating photovoltaic array output powers P t ~P n t (d1 pv )~P t (d pv n t ) corresponding to the duty cycles d1 t ~d n t at t time, comparing and recording the maximum power P maxAnd the location of the bird's nest corresponding to this power, i.e., the optimal duty cycle d. best ;

[0030] ④ Check the location of each bird's nest, i.e., the duty cycle d1 at time t. t ~d n t The corresponding photovoltaic array output power P at time t pv (d1 t )~P pv (d n t The point P with the maximum output power of the photovoltaic array during the iteration process max Does the difference exceed the set threshold ε, i.e.:

[0031]

[0032] d i t For the i-th bird's nest, the duty cycle is d. i At the position of the t-th iteration, if the result satisfies equation (3), then proceed to step ⑤, that is, use the Levi flight and the elimination mechanism of bad nests to update the positions of n nests, i.e., the duty cycle d1 at time t. t ~d n t ;

[0033] If the location of each bird's nest, i.e., the duty cycle d1 at time t. t ~d n t The corresponding photovoltaic array output power P at time t pv (d1 t )~P pv (d n t The point P with the maximum output power of the photovoltaic array during the iteration process max When the difference is less than or equal to the set threshold ε, that is:

[0034]

[0035] Then record the latest maximum power point (d) at the end of the iteration. GMPPi V GMPPi ,P GMPPi ,t i ), and store it in the online database of step (1). At this time, the iteration of the improved cuckoo algorithm ends, and it proceeds to step (4), that is, using the maximum power point (d) GMPPi V GMPPi ,P GMPPi The perturbation is performed as the initial position for variable step size perturbation observation or incremental conductance method;

[0036] ⑤Levy flight is used to update n bird nests, i.e., the duty cycle d t ~d n t Numerical value:

[0037]

[0038] In formula (5), t is the iteration number in the iteration process, is the i-th bird nest, i.e., the duty cycle d i At the position of the t-th iteration, represents point-to-point multiplication, a is a limiting coefficient of the flight step, and conforms to the standard normal distribution, L is a Levy search path, i.e., the step length during flight, γ c is the flight scale of Levy flight; β c = 3 / 2, d best represents the bird nest position corresponding to the maximum power point in the iteration process, i.e., the optimal duty cycle, u and v both conform to uniform distribution, i.e. and Γ is a standard gamma function;

[0039] ⑥In order to improve the convergence speed of the algorithm, the elimination mechanism of the cuckoo biological heuristic algorithm is used, and the elimination probability p a is discarded, and the bird nest corresponding to the minimum output power P pv (d t worst of the photovoltaic array at the t-th moment, i.e., the bad duty cycle d t worst at the t-th moment, is updated to the new duty cycle d t worst,new ;

[0040] The step ⑥ specifically refers to: in order to improve the convergence speed of the algorithm and reduce steady-state oscillation, it is not necessary to eliminate each bird nest, i.e., the duty cycle d i (i≤n) after each iteration, but only the worst bird nest, i.e., the bad duty cycle d t worst at the t-th moment, is eliminated after each iteration, and the adjacent bird nest of the bad bird nest in the traditional cuckoo algorithm, i.e., is replaced by the optimal bird nest, i.e., the optimal duty cycle d best , as shown in formulas (6) and (7):

[0041]

[0042] That is, after each iteration, the minimum value in the output power P of the offshore floating photovoltaic array is selected, i.e., min{P pv (d i t), i≤n}, corresponding d t worst As a bad individual; a random number r is generated for the bad individual. i ∈(0,1), when r i Greater than the elimination probability p a At that time, the new individual generation mechanism using formula (7) will reduce the undesirable duty cycle d at time t. t worst Replace with the new duty cycle d t worst,new We obtain the improved n bird nests, i.e., the duty cycle d1 at time t. t ~d n t Then proceed to step ③:

[0043] The improved bio-inspired algorithm in step (3) is based on the initial optimized distribution position of the population individuals at time 0 provided in step (2), i.e., the initial duty cycle d1. 0 ~d n 0 The implementation of the Grey Wolf algorithm, which improves the elimination rules for undesirable individuals, specifically includes:

[0044] (I) Set the number of individual gray wolves n and the elimination probability p. a And the criteria for terminating the iteration;

[0045] (II) The initial population distribution d1 determined by step (2) 0 ~d n 0 The initial positions of n individual gray wolves in the improved gray wolf algorithm, i.e., the initial duty cycle d1. 0 ~d n 0 ;

[0046] (III) The position of the gray wolf at time t, i.e., the duty cycle d1 at time t. t ~d n t (t≥0) are sequentially applied to the DC-DC rectifier unit connected to the floating photovoltaic array at sea, and the DC-DC rectifier unit outputs a duty cycle d1 at time t. t ~d n t The corresponding output power P of the floating photovoltaic array at time t pv (d1 t )~P pv (d n t Compare and record the maximum power P. max And the gray wolf position corresponding to this power, i.e., the duty cycle d. best ;

[0047] (IV) check the position of each wolf, i.e. the duty cycle d1 t ~ d n t , the corresponding output power P pv (d1 t ) ~ P pv (d n t of the photovoltaic array at time t is greater than the maximum output power point P max of the photovoltaic array in the iteration process, i.e. whether the following condition is met:

[0048]

[0049] is the position of the i-th wolf, i.e. the duty cycle at the t-th iteration, if the result satisfies equation (8), step (V) is entered, i.e. the position of the n wolves, i.e. the duty cycle variable, is updated;

[0050] If the position of each wolf, i.e. the duty cycle d1 t ~ d n t , the corresponding output power P pv (d1 t ) ~ P pv (d n t of the photovoltaic array at time t is less than or equal to the maximum output power point P max of the photovoltaic array in the iteration process, i.e. equation (9) is met, the latest maximum power point (d GMPPi , V GMPPi , P GMPPi , t i ) at the end of the iteration is recorded and stored in the online database of step (1), at this time the improved grey wolf optimization algorithm iteration is ended and step (4) is entered, i.e. the maximum power point (d GMPPi , V GMPPi , P GMPPi ) is used as the initial position of the variable step size perturbation observation or conductance increment method for perturbation;

[0051]

[0052] (V) the position of the wolf, i.e. the duty cycle d1 t ~ d n t , the corresponding output power P pv (d1 t ) ~ P pv (d n t of the photovoltaic array is sorted from large to small, and P pv (d1 t) ~ P pv (d n t ) maximum value P max corresponding to the position of the gray wolf, i.e. named alpha wolf, P pv (d1 t ) ~ P pv (d n t ) the second largest value corresponding to the position of the gray wolf, i.e. named beta wolf, P pv (d1 t ) ~ P pv (d n t ) the third largest value corresponding to the position of the gray wolf, i.e. named delta wolf, and the remaining wolf positions, i.e. are divided into omega wolves; for each omega wolf, i.e. its new position is calculated according to the updating formula of the position of the wolf pack in the search process shown by formula (10) involving the positions of alpha wolf, beta wolf and delta wolf, aiming to imitate the hunting behavior of gray wolves and help explore to obtain better solutions;

[0053]

[0054] In formula (10), is the position of the i-th wolf in omega at the t-th iteration, is the position of the i-th wolf in omega at the t+1-th iteration, is the prey, i.e. the maximum output power point P max of the photovoltaic array in the iteration process, A and C are coefficient vectors, and D is the distance between the gray wolf individual and the prey, i.e. and , a is a convergence factor that linearly decreases in the interval [0, 2] with random iteration number, and r1 and r2 are random vectors taking values in the interval [0, 1];

[0055] (VI) when it is judged that the position of the prey, i.e. the optimal duty cycle, is obtained, the omega wolves are surrounded by the alpha wolf, the beta wolf and the delta wolf, and the position updating formula is shown by formula (11):

[0056]

[0057] In formula (9), D a , D β , D δ are the distances between the positions of the gray wolf individuals and the positions of the alpha wolf, the beta wolf and the delta wolf, i.e. and the distance and direction of the advance of the wolf individual to alpha wolf, beta wolf and delta wolf, i.e. to the distance and direction of the advance of the wolf individual to alpha wolf, beta wolf and delta wolf, i.e.

[0058] (Ⅶ) In order to improve the convergence speed of the algorithm, an improved mechanism for eliminating the worst individual is used, i.e. after each iteration, according to the elimination probability p a the gray wolf position corresponding to the minimum output power of the photovoltaic array is discarded, i.e. the corresponding is updated to d t worst,new In the process of algorithm iteration, r i ∈(0,1) is randomly taken, when r i is greater than the elimination probability p a , the worst gray wolf individual position is improved, i.e. the improved n gray wolf positions are obtained, i.e. the duty ratio d1 t ~d n t , and step (Ⅲ) is entered.

[0059]

[0060] (4) On the basis of steps (1), (2) and (3), a variable step MPPT algorithm is used to improve the tracking speed and accuracy of maximum power tracking, and after detecting the power fluctuation caused by environmental changes, the improved biological heuristic algorithm is restarted, so as to ensure that the offshore floating photovoltaic array is always quickly and accurately operated near the theoretical maximum power point.

[0061] The step (4) specifically refers to: using the improved biological heuristic algorithm in step (3) to iterate the population, and after each iteration of the population, the difference between the power value P corresponding to the ith individual in the population and the power value P best corresponding to the optimal individual d max in the iteration process is compared; when are all not greater than the threshold value ε, the latest maximum power point (d GMPPi , V GMPPi , P GMPPi , t i ) at the end of iteration is recorded, and is stored in the online database of step (1), and the latest recorded (d GMPPi , V GMPPi , P GMPPi) as the initial position of variable step size perturbation and observation or conductance increment method to perform maximum power tracking; when the variable step size perturbation and observation or conductance increment method is used to track, if the difference between the output power value P pv (V t ) at the t moment and the newly recorded maximum power P GMPPi is greater than a set proportion θ, it indicates that the external environment changes and causes the maximum power point to change, at this moment, the improved biological heuristic algorithm needs to be restarted in step (2), which specifically includes the following contents:

[0062] (4-1) The perturbation voltage of the variable step size perturbation and observation or variable step size conductance increment method can be calculated by formula (13):

[0063]

[0064] Where ΔV is the variable step size perturbation voltage, dP pv is the output power fluctuation of the photovoltaic array and dV is the voltage fluctuation at the last perturbation moment, λ is the step size scaling factor, and ΔV1 and ΔV2 are the upper and lower limits of the step size, respectively.

[0065] (4-2) When the variable step size MPPT technology is used, if the irradiance of the photovoltaic panel is not uniform or the lighting condition changes due to the marine environment and weather conditions, the output power of the offshore floating photovoltaic system also changes accordingly. In order to reduce the power loss and ensure that the offshore floating photovoltaic system can quickly and accurately work near the theoretical maximum power point, the improved biological heuristic algorithm needs to be restarted in step (2) to perform maximum power tracking, that is:

[0066] If the difference between the output power value P pv (V t ) at the t moment and the latest maximum power P GMPP,i at the end of the iteration in step (3) is greater than a set proportion θ, that is, formula (14) is satisfied, then step (2) is returned to restart the improved biological heuristic algorithm to perform maximum power tracking;

[0067]

[0068] If the difference between the output power value P pv (V t ) at the t moment and the latest maximum power P GMPP,i at the end of the iteration in step (3) is less than or equal to a set proportion θ, that is, formula (15) is satisfied, step (4-1) is returned to continue perturbation operation, which ensures that the offshore floating photovoltaic array always quickly and accurately operates near the theoretical maximum power point.

[0069]

[0070] The working principle of the present application: the improved biological heuristic algorithm reduces the power oscillation in the early and middle optimization process, but still has the problem of slow convergence speed in the later period. Therefore, in order to further improve the convergence speed and accuracy of MPPT, the improved biological heuristic algorithm can be combined with the traditional MPPT technology, i.e. the perturbation and observation method and the conductance increment method, to optimize the speed and accuracy of MPPT tracking in the later period by using the good local search ability and small convergence oscillation of the perturbation and observation method and the conductance increment method. The principle of the perturbation and observation method and the conductance increment method is basically the same, that is, the position of the maximum power point is judged by the power fluctuation caused by voltage perturbation, and the next perturbation is carried out. According to the size of voltage perturbation, the fixed step perturbation and observation method and the conductance increment method can be divided into two types. The large step method uses a large duty cycle perturbation to quickly find the maximum power point, but the oscillation at the maximum power point is too large, which reduces the tracking accuracy. The small step method has small oscillation at the maximum power point, which improves the maximum power point tracking accuracy, but reduces the tracking speed. In order to improve the convergence speed in the later period and reduce the power fluctuation after convergence, and improve the MPPT tracking efficiency, the improved biological heuristic algorithm can be combined with the variable step traditional algorithm such as variable step perturbation and observation method and variable step conductance increment method.

[0071] Advantages of the present application:

[0072] (1) The MPPT control method proposed in the present application can not need to use redundant external sensors including irradiance and temperature sensors in the marine environment, and the self-adaptive update of the maximum power point history optimal database (d GMPPi ,V GMPPi ,P GMPPi ,t i ) is carried out through the internal sensors in the electrical system during the operation of the photovoltaic power station, without the need for additional electrical devices, which reduces the complexity and cost of the system and increases the reliability of the system during operation.

[0073] (2) When establishing the maximum power point history optimal database, the periodic wind and wave environment of the target engineering sea area is taken into account in the update principle of the maximum power point history optimal database, so as to realize the exclusion of bad data and the protection of effective data, reduce the distribution range of the historical optimal duty cycle d GMPPi in the database, and improve the performance of the optimal population distribution of the maximum power point history optimal database.

[0074] (3) In order to simplify the partition of data by the support vector machine method, the historical optimal duty cycle d GMPPi in the maximum power point history optimal database is extracted for 1-dimensional data classification, so that a large amount of historical electrical data is not needed for pre-training of the support vector machine method, and the maximum power tracking performance of the newly built offshore floating photovoltaic power station lacking historical electrical data is guaranteed.

[0075] (4) The application utilizes the establishment of a dynamically updated maximum power point history optimal database, probability statistics and support vector machine processing of low-dimensional array auxiliary biological heuristic algorithm optimization population initial distribution, proposes a bad individual elimination replacement mechanism, optimizes the population distribution and convergence step of the biological heuristic algorithm, and reduces the power oscillation and tracking time in the early stage of tracking.

[0076] (5) Considering the problem of slow convergence speed of the biological heuristic algorithm in the later stage, therefore, in order to further improve the MPPT convergence speed and accuracy, the application combines the improved biological heuristic algorithm with the traditional MPPT technology, i.e. the perturbation and observation method and the conductance increment method, utilizes the good local search ability and small convergence oscillation of the traditional MPPT technology, and realizes the effect of optimizing the speed and accuracy in the later stage of MPPT tracking. (Four) DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 It is a flowchart of the MPPT control method of the offshore floating type photovoltaic power generation system.

[0078] Figure 2 It is a schematic diagram of the establishment method of the dynamically updated maximum power point history optimal database in the MPPT control method of the offshore floating type photovoltaic power generation system.

[0079] Figure 3 It is a schematic diagram of the optimization of the biological heuristic algorithm population initial distribution by utilizing the maximum power point history optimal database and the support vector machine method in the MPPT control method of the offshore floating type photovoltaic power generation system.

[0080] Figure 4 It is a schematic diagram of the variable step perturbation and observation method in the MPPT control method of the offshore floating type photovoltaic power generation system.

[0081] Figure 5 It is a schematic diagram of the MPPT control circuit of the offshore floating type photovoltaic array in the embodiment.

[0082] Figure 6 It is a schematic diagram of the offshore floating type photovoltaic array uneven irradiance scene in the embodiment.

[0083] Figure 7 It is a schematic diagram of the P-V curve under the offshore floating type photovoltaic array uneven irradiance scene in the embodiment.

[0084] Figure 8 It is a schematic diagram of the photovoltaic array output power curve in the embodiment 1.

[0085] Figure 9This is a schematic diagram of the output power curve of the photovoltaic array in Embodiment 2 of the present invention. (V) Specific Implementation Methods:

[0086] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below. It should be noted that the embodiments described below are only some embodiments of the present invention, and not all embodiments. Where there is no conflict, the embodiments and features described in the present invention can be combined with each other.

[0087] Example 1: This embodiment of the invention provides an MPPT control method suitable for offshore floating photovoltaic power generation systems, as shown in the attached figure. Figure 1 As shown, the photovoltaic array adjusts its output voltage through a DC-DC rectifier unit to ensure it operates at its maximum power point. This method includes the following steps: initial population distribution optimization driven by a historical optimal database of the maximum power point, iterative improvement of the cuckoo algorithm, variable step-size perturbation observation method, and algorithm restart.

[0088] 101: Establish a dynamically updated historical optimal database of maximum power points, and use the historical optimal database of maximum power points and support vector machine to optimize the initial distribution and convergence step size of the population in the bio-inspired algorithm, thereby reducing power oscillations and tracking time in the early stage of tracking;

[0089] Record the global maximum power point searched during the operation, (d GMPPi V GMPPi ,P GMPPi ,t i Set the number of maximum power points stored in the historical best database of maximum power points to m. c Based on the data entry sequence, these data are divided into online and offline databases according to the ratios k1 and 1-k1. After each MPPT process, the maximum power point with the highest dispersion (i.e., variance) in the offline database is removed, and a maximum power point is randomly selected from the online database and migrated to the offline database. Finally, the new maximum power point obtained from the MPPT is migrated to the online database. See [link to MPPT process]. Figure 2 .

[0090] To simplify the partitioning of multidimensional data in the historical optimal database of maximum power points using the support vector machine method, the historical optimal database of maximum power points (d) is extracted. GMPPi V GMPPi ,P GMPPi ,t i The historical best duty cycle d GMPPi Using the support vector machine method to analyze d GMPPi The array is used to classify one-dimensional data. This eliminates the need for pre-training the Support Vector Machine (SVM) model. When performing partitioning, the SVM only needs to find d...GMPPi The maximum interval of adjacent data and the average value of adjacent data are used as the basis for dividing the region. The data d GMPPi is divided into three partitions, [d0, d1], [d1, d2], and [d2, d3], as shown in Figure 3 .

[0091] The number of historical optimal duty cycles (N1, N2, N3) is counted, and d GMPPi is calculated. GMPPi The probability distribution rates (F1, F2, F3) of the three intervals are calculated. Finally, the number of randomly distributed populations (D1, D2, D3) in each interval at the initial moment of the bio-inspired algorithm is determined according to the probability distribution of the interval, as shown in equation (1).

[0092]

[0093] D j = nθ c F j , j e {1, 2, 3}

[0094] In equation (1), n is the population size of the algorithm; θ c is a constant in the interval [0, 1]. Considering that the maximum power point may appear outside of min(d GMPPi ) to max(d GMPPi ), not all individuals in the population are determined by the initial position of the historical optimal database, only the total number of nθ c individuals are determined by equation (1). In addition to the 1-z region, the remaining number of individuals is n(1-θ c ), which is randomly distributed in the region [0, min(d GMPPi )] and [max(d GMPPi ), d uplimit ]. The above steps complete the initial distribution of the bio-inspired algorithm population, that is, d1 0 ~ d n 0 .

[0095] The value range of θ c is generally 0.55-0.75. If θ > 0.75, d min ~ d maxThe smaller number of off-grid populations may cause them to miss the range containing the maximum power output of the photovoltaic array. If θ < 0.55, the role of the historical best database of maximum power points in the algorithm will be weakened, and the convergence time will be prolonged. After each run of the bio-inspired algorithm, the historical best database of maximum power points needs to be updated promptly, with the latest maximum power point replacing the historical data. Furthermore, to improve the convergence speed of traditional bio-inspired algorithms and reduce power oscillations during convergence, the iteration step size and the replacement of defective individuals can be improved based on the characteristics of bio-inspired algorithms.

[0096] 102: The following uses the improved cuckoo algorithm in bio-inspired algorithms and the variable step size perturbation observation method in traditional MPPT techniques as examples to further illustrate the present invention, but this is not intended to limit the present invention.

[0097] The cuckoo algorithm is based on the parasitic reproduction of some cuckoo species and the Levy flight mechanism of birds. When applied to MPPT control in a photovoltaic system, the nest position corresponds to the duty cycle d, and the nest mass represents the output power P of the photovoltaic array at the corresponding duty cycle d. pv Because the cuckoo method abandons the conventional isotropic random walk and relies on Lévy flight to enhance the search effect, theoretically, the cuckoo method has a better search path and faster optimization speed than methods such as particle swarm optimization based on particle random walks. In photovoltaic MPPT applications, it can track the global maximum power point faster and more accurately. Furthermore, according to the discovery probability p... a The improved nest location also allows the cuckoo algorithm to more effectively escape local maximum power points, thereby reducing power mismatch losses.

[0098] An improved cuckoo algorithm is used iteratively. First, the number of nests (i.e., the number of duty cycles, n) and the elimination probability p are set. a And the iteration termination criterion. The initial positions d1 of the n bird nests in the improved cuckoo algorithm, determined by 101. 0 ~d n 0 And then apply them sequentially to the DC-DC rectifier unit, according to the output power P of the corresponding floating photovoltaic array. pv (d1 0 )~P pv (d n 0 The maximum output power P of the photovoltaic array during the iteration process is recorded. max And the corresponding location of the Bird's Nest, i.e., the optimal duty cycle d. best Then, using Levi's flight, the positions of the n bird nests at time t are updated, i.e., the duty cycle d1 at time t. t ~d n t :

[0099]

[0100] In equation (2), t is the current iteration number. For the i-th bird's nest, the duty cycle is d. i At the position of the t-th iteration, This represents point-to-point multiplication, where α is the limiting coefficient for the flight step size and follows a standard normal distribution, L is the Lévy search path (i.e., the step size during flight), and γ... c For Lévy flight, the flight scale; β c =3 / 2u and v both follow a uniform distribution, that is and Γ is the standard gamma function.

[0101] To improve the convergence speed of the algorithm, the traditional cuckoo algorithm utilizes a nest elimination mechanism based on probability p. a Discarding nests and updating them means that during the algorithm iteration process, each nest, i.e., the duty cycle, is randomly selected as r. i ∈(0,1), when r i Greater than the elimination probability p a At that time, the location of the bird's nest was improved.

[0102]

[0103] In equation (3), To and Adjacent duty cycles. To improve the convergence speed of the algorithm, the elimination mechanism of undesirable individuals in the cuckoo-inspired algorithm is utilized, based on the elimination probability p. a Discarding the minimum output power P of the photovoltaic array pv (d t worst The corresponding Bird's Nest, i.e., the defective duty cycle d at time t. t worst And update to the new duty cycle d t worst,new And it will replace the adjacent duty cycles of bad duty cycles in the traditional cuckoo algorithm. Replace with the optimal duty cycle d best That is, after each iteration, select the power value. The minimum value in is min{P} pv (d i t ), i≤n} corresponds to d t worst For d t worst Generate a random number r i ∈(0,1), when r i Greater than the elimination probability p aAt that time, the mechanism for generating new individuals will be used to generate d t worst Replace with d t worst,new This improves the convergence speed of biological populations.

[0104]

[0105] The determined iteration positions of n nests in the improved cuckoo algorithm, i.e., the duty cycle d1 at time t. t ~d n t They are then applied sequentially to the DC-DC rectifier unit, based on the output power P of the corresponding floating photovoltaic array. pv (d1 t )~P pv (d n t ), compare and record the maximum power P max And the corresponding location of the Bird's Nest, i.e., the optimal duty cycle d. best If the duty cycle of each bird's nest at time t is d1 t ~d n t The corresponding photovoltaic array output power P at time t pv (d1 t )~P pv (d n t The point of maximum output power of the photovoltaic array P during the iteration process max If the difference still exceeds the set threshold ε, continue to use equations (2) and (4), i.e., the Levi flight and the elimination mechanism of bad nests, to update the position of n nests, i.e., the duty cycle d1 at time t. t ~d n t :

[0106]

[0107] When the duty cycle of each bird's nest is d1 at time t t ~d n t The corresponding photovoltaic array output power P at time t pv (d1 t )~P pv (d n t The point P with the maximum output power of the photovoltaic array during the iteration process max When the difference is less than the set threshold ε:

[0108]

[0109] Record the latest maximum power point (d) at the end of the iteration.GMPPi ,V GMPPi ,P GMPPi ,t i ), and store it into the online database in step 101, at which time the improved cuckoo algorithm iteration ends, and turn to step 103, i.e. to perform perturbation using the latest maximum power point (d GMPPi ,V GMPPi ,P GMPPi ) at the end of the iteration as the initial position of the variable step size perturbation observation;

[0110] 103: variable step size perturbation observation method with good local search and small convergence oscillation;

[0111] At the time when the variable step size perturbation observation method is enabled, the latest maximum power point (d GMPPi ,V GMPPi ,P GMPPi ) at the end of the improved cuckoo algorithm iteration is used for voltage perturbation, and the difference ΔP pv , ΔV t between the t time output power P pv (V t ), voltage V t and the t-1 time output power P pv (V t-1 ), voltage V t-1 is used to determine the perturbation direction to perform voltage perturbation, as shown in equation (7).

[0112] Specific perturbation principle Figure 4 As shown in equation (7), the closer to the maximum power point, the closer to zero the power perturbation caused by voltage perturbation.

[0113]

[0114] The voltage perturbation step size ΔV of the variable step size perturbation observation method each time is:

[0115]

[0116] In equation (8), ΔV is the variable step size, λ is the step size scaling factor, and ΔV1 and ΔV2 are the upper and lower limits of the voltage perturbation step size. In the early stage of optimization, ΔP pv is large, so the tracking step size is large and the tracking speed is fast; in the later stage of optimization, ΔP pv is gradually reduced, so the step size is reduced, which will not produce large fluctuations around the MPP, the tracking accuracy is improved, and then the power loss is reduced.

[0117] 104: determine whether to restart the algorithm;

[0118] When the irradiance of photovoltaic panels is not uniform or the light condition changes due to marine environment and weather conditions, the output power of the offshore floating photovoltaic system also changes accordingly, and the output power of the photovoltaic power generation system also changes accordingly. In order to reduce power loss, the variable step size perturbation and observation method needs to be terminated in time, and the improved cuckoo algorithm is restarted to restart the maximum power tracking in step 101.

[0119] When the output power value P pv (V t ) at time t is detected, and the difference between the latest maximum power P GMPPi at the end of step 102 iteration is greater than the set proportion θ, that is, formula (9) is satisfied, return to step 102 to restart the improved cuckoo algorithm.

[0120]

[0121] When the output power value P pv (V t ) at time t is detected, and the difference between the latest maximum power P GMPPi at the end of step 102 iteration is less than or equal to the set proportion θ, that is, formula (10) is satisfied, continue to use formula (7), (8) perturbation operation.

[0122]

[0123] A certain offshore floating photovoltaic power generation system is taken as an example for maximum power tracking test. The DC-DC rectifier unit is selected as a Boost rectifier circuit, and the MPPT control circuit is as shown in Figure 5 . The capacity of the offshore floating photovoltaic array is about 30kW, the photovoltaic array contains 26 photovoltaic components per group string, and there are 2 photovoltaic group strings connected to the same 1-way MPPT control circuit. The unevenly distributed irradiance of the photovoltaic array under the action of sea waves is as shown in Figure 6 , and the photovoltaic component parameters are as shown in Table 1.

[0124] Table 1: Parameters of double-sided photovoltaic components

[0125]

[0126] The P-V curve corresponding to the unevenly distributed irradiance of the photovoltaic array under the action of sea waves is as shown in Figure 7 , at this time the local maximum power point of the photovoltaic array has 4, and the global maximum power point (1286V, 13.820kW) is located at the right side of the curve. From Figure 8It can be seen that the hybrid improved algorithm based on improved cuckoo algorithm and variable step size perturbation and observation method under the non-uniform irradiance field scenario tracks the global maximum power point (1286V, 13.820kW), the algorithm convergence time is about 19ms, and the average tracking efficiency of the algorithm is about 99.88%, which can meet the requirements of the maximum power tracking of the offshore floating photovoltaic power generation system.

[0127] Although specific bio-inspired algorithms, conventional MPPT techniques and embodiments are employed in the present application to describe the present application, it should be made clear that these algorithms and embodiments are merely examples of the principles and applications of the present application. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed, without departing from the spirit and scope of the present application as defined in the appended claims.

[0128] Embodiment 2: The present application provides a MPPT control method suitable for offshore floating photovoltaic power generation system, as shown in the accompanying Figure 1 photovoltaic array adjusts the output voltage of the photovoltaic array through the DC-DC rectifier unit, and ensures that the photovoltaic array operates at the maximum power point. The method comprises the following steps: population initial distribution optimization based on maximum power point historical optimal database driving, improved grey wolf algorithm iteration, variable step size conductance increment method, algorithm restart.

[0129] 101: Establish a dynamically updated maximum power point historical optimal database, use the maximum power point historical optimal database and support vector machine to optimize the initial distribution and convergence step of the population of the bio-inspired algorithm, reduce the power oscillation and tracking time in the early stage of tracking;

[0130] record the global maximum power point searched during operation, (d GMPPi ,V GMPPi ,P GMPPi ,t i ). Set the number of maximum power points stored in the maximum power point historical optimal database as m c , according to the time sequence of data warehousing, divide these data into online database and offline database according to the proportion k1 and 1-k1. After each MPPT process is completed, the maximum power point with the highest dispersion, i.e. the maximum variance, in the offline database is removed, and a maximum power point is randomly selected from the online database and migrated to the offline database, and finally the new maximum power point obtained by MPPT is migrated to the online database, as shown in Figure 2 .

[0131] To simplify the partitioning of multi-dimensional data in the maximum power point historical optimal database by support vector machine method, the historical optimal database (d GMPPi ,V GMPPi ,P GMPPi ,t i) in d GMPPi Using the support vector machine method to analyze d GMPPi The array is used to classify one-dimensional data. This eliminates the need for pre-training the Support Vector Machine (SVM) model. When performing partitioning, the SVM only needs to find d... GMPPi The maximum interval between adjacent data points is used, and the average value of adjacent data points is used as the basis for dividing the region. The data d is set to... GMPPi Divide into 3 partitions, the intervals are [d0,d1], [d1,d2], and [d2,d3], as shown in the figure.

[0132] Calculate the values ​​of d in the three partitions [d0,d1], [d1,d2], and [d2,d3]. GMPPi The number of historical best duty cycles (N1, N2, N3) is statistically calculated as d. GMPPi Finally, based on the probability distribution of the three intervals (F1, F2, F3), the number of randomly distributed populations in each interval at the initial moment of the bio-inspired algorithm (D1, D2, D3) is determined, as shown in Equation (1).

[0133]

[0134] D j =nθ c F j j∈{1,2,3}

[0135] In formula (1), n ​​is the number of the algorithm population; θ c Let be a constant over the interval [0, 1]. Considering that the maximum power point may occur at d... min ~d max Furthermore, not all individuals in the population have their initial positions determined by the historical best database; only those with a total number of nθ are included. c The position of an individual is determined by equation (1). The number of remaining individuals, excluding the region 1 to z, is n(1-θ). c ), randomly distributed in the region [0, d min ] and [d max ,d uplimit The above steps complete the initial distribution of the population for the biologically inspired algorithm, thus obtaining d1. 0 ~d n 0 .

[0136] θ c The value range of θ is generally 0.55 to 0.75. If θ > 0.75, then d min ~d maxThe population of the outer distribution is less, and the maximum power point of the photovoltaic array output interval can be missed. If θ < 0.55, the role of the historical optimal database in the algorithm will be weakened, and the convergence time of the algorithm will be prolonged. After each biological heuristic algorithm is run, the maximum power point historical optimal database needs to be updated in time, and the latest maximum power point is added to replace the historical data. In addition, in order to improve the convergence speed of the traditional biological heuristic algorithm and reduce the power oscillation in the convergence process, the iteration step and the replacement of the poor individual can be improved according to the characteristics of the biological heuristic algorithm.

[0137] 102: The improved grey wolf algorithm in the biological heuristic algorithm and the variable step size conductance increment method in the traditional MPPT technology are taken as examples to further illustrate the present application, but not as a limitation of the present application.

[0138] The grey wolf algorithm is based on the hierarchical system of wolf packs and the hunting characteristics of wolf packs. The grey wolf algorithm is applied to the MPPT control of the photovoltaic system. The position of each wolf in the wolf pack corresponds to the duty cycle d, and the position of the wolf pack corresponds to the output power P of the photovoltaic array pv . The grey wolf algorithm has few parameters and is easy to implement, has an information sharing mechanism, and has strong convergence ability.

[0139] The improved grey wolf iteration is adopted. Firstly, the number of grey wolf individuals n, the elimination probability p a and the iteration termination criterion are set. The initial positions d1 0 ~ d n 0 of the n grey wolves in the improved grey wolf algorithm determined by 101 are applied to the DC-DC rectifier unit in turn, and the initial output power P pv (d1 0 ) ~ P pv (d n 0 of the corresponding offshore floating photovoltaic array is compared. max The maximum output power P best of the photovoltaic array in the iteration process is recorded, and the corresponding grey wolf position, i.e. the optimal duty cycle d t is recorded.

[0140] The grey wolf position, i.e. the duty cycle d1 n ~ d t (t≥0) at time t corresponds to the output power P pv (d1 t ) ~ P pv (d n t of the photovoltaic array at time t is sorted from large to small, and P pv (d1 t ) ~ P pv (d n t) maximum value P max corresponding to the position of the gray wolf named alpha wolf, P pv (d1 t ) ~ P pv (d n t ) the second largest value corresponding to the position of the gray wolf named beta wolf, P pv (d1 t ) ~ P pv (d n t ) the third largest value corresponding to the position of the gray wolf named delta wolf, and the remaining wolves are divided into omega wolves; for each omega wolf, the new position is calculated according to an equation involving the positions of the alpha wolf, the beta wolf and the delta wolf, and the updating formula for the position of the wolf pack during the search process is:

[0141]

[0142] In equation (2), P is the position of the i-th wolf in omega at the t-th iteration, is the position of the i-th wolf in omega at the t+1-th iteration, is the prey, i.e. the maximum output power point P max of the photovoltaic array, A and C are coefficient vectors, and D is the distance between the gray wolf individual and the prey, i.e. and a is a convergence factor that linearly decreases in the interval [0, 2] with the number of random iterations, r1 and r2 are random vectors that take values in the interval [0, 1]. When the position of the prey (optimal duty cycle) is determined, the omega wolves are led by the alpha, beta and delta wolves to surround the prey, and the position updating formula is:

[0143]

[0144] In equation (3), D a , D β , D δ are the distances between the positions of the gray wolf individuals and the positions of the alpha wolf, the beta wolf and the delta wolf, i.e.

[0145] and a is a convergence factor that linearly decreases in the interval [0, 2] with the number of random iterations, r1 and r2 are random vectors that take values in the interval [0, 1]. A1, A2, A3, C1, C2, C3 are the coefficient vectors described in equation (2), and d1, d2 and d3 are the distances and directions of the omega wolves to the alpha wolf, the beta wolf and the delta wolf, i.e. to .

[0146] ​​To improve the convergence speed of the algorithm, an improved elimination mechanism for undesirable individuals is used: after each iteration, based on the elimination probability p... a Discard the gray wolf position corresponding to the minimum output power of the photovoltaic array, i.e., the defective duty cycle at time t. Right now corresponding Updated to d t worst,new During the algorithm iteration process, r is randomly selected. i ∈(0,1), when r i Greater than the elimination probability p a At time t, the worst position of the gray wolf individual is improved, i.e., the poor duty cycle. The improved n gray wolf positions are obtained, i.e., the duty cycle d1 at time t. t ~d n t :

[0147]

[0148] The positions of n individual gray wolves in the determined improved gray wolf algorithm, i.e., the duty cycle d1 at time t. t ~d n t This is then applied sequentially to the DC-DC rectifier unit, based on the output power P of the corresponding floating photovoltaic array at time t. pv (d1 t )~P pv (d n t ), compare and record the maximum power P max And the corresponding wolf pack position, i.e., the optimal duty cycle d. best If the duty cycle of each individual gray wolf at time t is d1 t ~d n t The corresponding photovoltaic array output power P at time t pv (d1 t )~P pv (d n t The point P with the maximum output power of the photovoltaic array during the iteration process max If the difference still exceeds the set threshold ε, continue to update the n gray wolf positions, i.e., the duty cycle positions, using formulas (2), (3), and (4) in the gray wolf algorithm:

[0149]

[0150] When the duty cycle d1 is at each gray wolf position at time t t ~d n t The corresponding photovoltaic array output power P at time tpv (d1 t )~P pv (d n t ) with the difference between the maximum power P max of the photovoltaic array in the iteration process and the maximum power P GMPPi of the photovoltaic array in the iteration process is less than the set threshold ε:

[0151]

[0152] Record the latest maximum power point (d GMPPi ,V GMPPi ,P i ) at the end of the iteration, store it into the online database in step 101, at this time the improved grey wolf algorithm iteration ends, and turn to step 103, that is, the latest maximum power point (d GMPPi ,V GMPPi ,P GMPPi ) at the end of the iteration is taken as the initial position of the variable step size conductance increment method for disturbance;

[0153] 103: variable step size conductance increment method with good local search and small convergence oscillation;

[0154] At the moment when the variable step size conductance increment method is started, the voltage disturbance is carried out on the newly recorded maximum power point (d GMPPi ,V GMPPi ,P GMPPi ) at the end of the improved grey wolf algorithm iteration, and the disturbance direction is determined according to the difference dP pv , dV t between the output power P t (V pv ), voltage V t-1 of the offshore floating photovoltaic array at time t and the output power P t-1 (V pv ), voltage V t at time t-1 to carry out voltage disturbance, as shown in formula (8).

[0155] Specific disturbance principle Figure 4 As shown in formula (7), the closer to the maximum power point, the closer to zero the power disturbance caused by voltage disturbance.

[0156]

[0157] The voltage disturbance step size ΔV of the variable step size conductance increment method each time is:

[0158]

[0159] In formula (9), ΔV is the variable step size, λ is the step size scaling factor, and ΔV1 and ΔV2 are the upper and lower limits of the voltage disturbance step size. In the initial stage of optimization, dPpv Larger, so the tracking step is larger, the tracking speed is fast; in the later stage of optimization dP pv Gradually decreases, so the step size decreases, and large fluctuations around the MPP will not occur, the tracking accuracy is improved, and then the power loss is reduced.

[0160] 104: Determine whether to restart the algorithm;

[0161] When the irradiance of the photovoltaic panel is uneven or the lighting condition changes due to the marine environment and weather conditions, the output power of the photovoltaic power generation system also changes accordingly. In order to reduce the power loss, the variable step size perturbation and observation method needs to be terminated in time, and the improved grey wolf algorithm is restarted to return to step 101 for maximum power tracking.

[0162] When the output power value P pv (V t ) at time t is detected, and the difference between the latest maximum power P GMPPi at the end of step 102 iteration is greater than the set proportion θ, that is, formula (10) is satisfied, return to step 102 to restart the improved grey wolf algorithm.

[0163]

[0164] When the output power value P pv (V t ) at time t is detected, and the difference between the latest maximum power P GMPPi at the end of step 102 iteration is less than or equal to the set proportion θ, that is, formula (11) is satisfied, return to step 103 to continue using formulas (8), (9) for perturbation operation.

[0165]

[0166] In order to verify the effectiveness of the method, a certain offshore floating photovoltaic power generation system is taken as an example for maximum power tracking test. The DC-DC rectifier unit is selected as a Boost rectifier circuit, and the MPPT control circuit is as shown in Figure 5 . The capacity of the offshore floating photovoltaic array is about 30kW, each group string of the photovoltaic array contains 26 photovoltaic components, and there are 2 photovoltaic group strings connected to the same 1-way MPPT control circuit. The unevenly distributed irradiance of the photovoltaic array under the action of sea waves is as shown in Figure 6 , and the photovoltaic component parameters are as shown in Table 1.

[0167] Table 1: Parameters of double-sided photovoltaic components

[0168]

[0169] The P-V curve corresponding to the unevenly distributed irradiance of the photovoltaic array under the action of sea waves is as shown in Figure 7As shown, there are four local maximum power points of the photovoltaic array at this time, and the global maximum power point (1286V, 13.820kW) is located at the rightmost side of the curve. From Figure 9 It can be seen that the global maximum power point (1286V, 13.820kW) is tracked by the hybrid improved algorithm based on the improved grey wolf algorithm and the variable step size perturbation and observation method under the non-uniform irradiance field, the algorithm convergence time is about 12ms, and the average tracking efficiency of the algorithm is about 99.89%, which can meet the maximum power tracking requirements of the offshore floating photovoltaic power generation system.

[0170] Although specific bio-inspired algorithms, conventional MPPT techniques and embodiments are used in the present application to describe the present application, it should be made clear that these algorithms and embodiments are only examples of the principles and applications of the present application. It should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed, as long as they do not deviate from the spirit and scope of the present application defined in the appended claims.

Claims

1. A method for MPPT control of a marine floating photovoltaic power generation system, characterized in that It comprises the following steps: (1) By utilizing the movement of seawater in the determined sea area at multiple time scales and the time-varying periodicity of the maximum power point of the floating photovoltaic array, a dynamically updated historical optimal database of the maximum power point of the floating photovoltaic array is established (d GMPPi V GMPPi ,P GMPPi ,t i This reduces the initial population distribution range of traditional biologically inspired algorithms, where d GMPPi For a historical moment t i The corresponding historical best duty cycle, V GMPPi For a historical moment t i The corresponding historical best voltage, P GMPPi For a historical moment t i The corresponding historical maximum power; (2) using the dynamic updated maximum power point history optimal database provided by step (1), and based on probability statistics and support vector machine method, optimizing the initial distribution position of n population individuals of traditional biological heuristic algorithm at 0 time, that is, the initial duty cycle d1 0 ~d n 0 , reducing the power fluctuation and tracking time in the early stage of maximum power tracking, specifically comprising the following steps: (2-1) The historical optimal database of the maximum power point obtained by the extraction step (1) is divided into z partitions according to the value of the historical optimal duty cycle d GMPPi at the historical time t GMPPi , and the historical optimal duty cycle d GMPPi at the historical time t i is classified by using the support vector machine method, without pre-training the support vector machine model. i The support vector machine only needs to find the maximum interval of the historical optimal duty cycle d GMPPi at the historical time t GMPPi and take the average value of the adjacent data as the basis for dividing the region. i The historical optimal duty cycle d GMPPi at the historical time t i is divided into z partitions according to the value, and is respectively recorded as: [d0, d1], [d1, d2]…[d GMPPi , z-1 d z ]. (2-2) Statistically analyze the z partitions [d0,d1], [d1,d2]…[d2] obtained in step (2-1). z-1 ,d z [The best historical duty cycle d] GMPPi Quantity (N1~N) z Based on the number of historical best duty cycles obtained for each partition (N1 to N), z Calculate the historical best duty cycle d GMPPi In the interval [d] j-1 ,d j The probability distribution law F within ] j Based on the probability distribution F of the interval j The size D of the population is determined by randomly distributing it within the interval j at the initial time of the bio-inspired algorithm. j As shown in equation (1): In formula (1), n is the population size of the algorithm; θ c is a constant on the interval [0, 1]; Considering that the corresponding duty ratio of the maximum power point in the actual operation process can appear in the historical optimal duty ratio d GMPPi The minimum value is outside the maximum value range, that is, min(d GMPPi )~max(d GMPPi ), so not all individuals in the population determine the initial position by the historical optimal database, only the total number of nθ c individuals are determined by formula (1); the remaining number of individuals is n(1-θ c ), which is randomly distributed in the area [0, min(d GMPPi )] and [max(d GMPPi ), d uplimit ], to complete the initial distribution of the population of the biological heuristic algorithm, that is, to obtain the initial distribution position of the n population individuals at time 0, that is, the initial duty ratio d1 0 ~d n 0 ; (3) From the perspective of the elimination probability and the generation mechanism of new population individuals after the elimination of undesirable individuals, the elimination mechanism of undesirable individuals in the iterative process of the biological heuristic algorithm is improved, and the convergence speed in the medium term of the maximum power tracking is improved; (4) Based on steps (1), (2) and (3), the variable step size MPPT algorithm is used to improve the tracking speed and accuracy of maximum power point tracking. After detecting power fluctuations caused by environmental changes, the improved bio-inspired algorithm is restarted to ensure that the floating photovoltaic array at sea always operates quickly and accurately near the theoretical maximum power point. That is, the improved bio-inspired algorithm in step (3) is used to iterate the population, and after each iteration of the population, the i-th individual d in the population is compared. i t The corresponding power value P pv (d i t The optimal individual d during the iteration process best The corresponding power value P max The difference; when |P pv (d i t )-P max When all values ​​are not greater than the threshold ε, record the latest maximum power point (d) at the end of the iteration. GMPPnew V GMPPnew ,P GMPPnew ,t new ), and store it in the online database of step (1), and use the new record at this time (d GMPPnew V GMPPnew ,P GMPPnew ,t new The initial position for maximum power tracking is used as the observation position for variable step size perturbation or incremental conductance method; when using variable step size perturbation or incremental conductance method for tracking, the output power value P is detected at time t. pv (V t ) and the newly recorded maximum power P GMPPi If the difference is greater than the set ratio θ, it means that the external environment has changed and caused the maximum power point to change. At this time, return to step (2) to restart the improved bio-inspired algorithm.

2. The MPPT control method of a marine floating photovoltaic power generation system according to claim 1, characterized in that In the step (1), the dynamic updating of the maximum power point historical optimal database of the offshore floating photovoltaic array is established, and the initial distribution range of the population of the traditional biological heuristic algorithm is narrowed, specifically referring to: (1-1) Same quarter, different working hours t i A similar maximum power point will occur, denoted as (d GMPPi ,V GMPPi ,P GMPPi ); record the global maximum power point found during the operation of the offshore floating photovoltaic array, so as to establish a dynamically updated maximum power point history optimal database (d GMPPi ,V GMPPi ,P GMPPi ,t i ); (1-2) The maximum power point of the photovoltaic array has seasonal variation characteristics, in order to reduce the storage size of the maximum power point historical optimal database and reduce the initial distribution range of the population of the biological heuristic algorithm, when the data size reaches the upper limit of the maximum power point historical optimal database, the number of optimal data points is no longer increased, and the individual with the highest dispersion in the maximum power point historical optimal database is replaced through dynamic updating.

3. The MPPT control method of a marine floating photovoltaic power generation system according to claim 2, characterized in that The step (1-2) specifically refers to: (1-2-1) Assuming that the number of maximum power points stored in the maximum power point history optimal database is m c According to the time sequence of data warehousing, the data is divided into online database and offline database according to the proportion k1 and 1-k1. (1-2-2) After each MPPT optimization process, record the current latest maximum power point (d GMPPnew ,V GMPPnew ,P GMPPnew ,t new ) and store it in the online database; (1-2-3) When storing the current latest maximum power point, if the number of storage in the online database has reached the upper limit value k1m c , then randomly select one historical maximum power point from the online database and migrate it into the offline database. (1-2-4) When randomly selecting one historical maximum power point from the online database to migrate into the offline database, if the storage quantity of the offline database has reached the upper limit value (1-k1)m c , then calculate the variance of the data (d GMPPiL , V GMPPiL , P GMPPiL ) in the offline database, remove the maximum power point with the highest variance, i.e., the highest dispersion, to provide space for the data migration in the online database, and reduce the distribution range of the historical optimal duty cycle d i at the historical time t GMPPi corresponding to the maximum power point in the historical optimal database.

4. The MPPT control method of a marine floating photovoltaic power generation system according to claim 1, characterized in that In step (2-2), the constant θ c The value range is 0.55 to 0.

75.

5. The method of claim 1, wherein the method further comprises: The step (3) improves the biological heuristic algorithm based on the initial optimization distribution position d of the population individuals provided by the step (2) at 0 time i 0 ~d n 0 The cuckoo algorithm is realized by improving the elimination rule of the bad individuals, and specifically includes: ① Set the number of nests n, elimination probability p a and iteration termination criterion, (2) determining the initial distribution position d1 of the population from step (1) 0 ~d n 0 , confirming the initial position of the n nests in the cuckoo algorithm, i.e. the initial duty cycle d1 0 ~d n 0 ; ③ the bird nest position of the jth iteration, i.e. the duty cycle d1 j ~d j is applied to the DC-DC rectifier unit connected to the offshore floating photovoltaic array in sequence, and the DC-DC rectifier unit outputs the corresponding offshore floating photovoltaic array output power P j ~d t (d1 pv )~P j (d pv n j , the maximum power P max is compared and recorded, and the bird nest position corresponding to the maximum power P max , i.e. the optimal duty cycle d best ; wherein t≥0;​ iv. Check the position of each bird's nest, i.e. the duty cycle d1 of the tth iteration t ~ d n t The corresponding photovoltaic array tth iteration output power P pv (d1 t ) ~ P pv (d n t The difference between the photovoltaic array maximum output power point P max in the iteration process is greater than the set threshold value ε, that is, whether formula (3) is satisfied: d i t For the i-th bird nest, i.e. duty cycle d i At the position of the t-th iteration, if the result satisfies equation (3), go to step 5, i.e. update the position of the n bird nests by using the Levy flight and the elimination mechanism of bad bird nests, i.e. the duty cycle d1 of the t-th iteration t ~ d n t ; If each bird's nest position, i.e. the duty cycle d1 of the tth iteration t ~ d n t , the corresponding photovoltaic array tth iteration output power P pv (d1 t ) ~ P pv (d n t The difference between the tth iteration output power P max and the maximum output power point P of the photovoltaic array in the iteration process is less than or equal to the set threshold ε, i.e. formula (4) is satisfied: the latest maximum power point (d GMPPnew ,V GMPPnew ,P GMPPnew ,t new ) at the end of iteration is recorded and stored into the online database of step (1), at this time the improved cuckoo algorithm iteration ends and goes to step (4) to perturb at the maximum power point (d GMPPnew ,V GMPPnew ,P GMPPnew ,t new ) as the initial position of variable step size perturbation observation or conductance increment method; (5) Update the n-nest locations using Levy flight, i.e., the duty cycle di of the tth iteration t ~ d n t Numerical: In formula (5), t is the iteration number in the iteration process, d i t is the i th bird nest, i.e. the duty cycle d i is the position in the t th iteration, represents point-to-point multiplication, a is a limiting coefficient of the flight step, and conforms to the standard normal distribution, L is a Lévy search path, i.e. the step length in flight, γ c is the flight scale of Lévy flight; β c = 3 / 2, d best represents the bird nest position corresponding to the maximum power point in the iteration process, i.e. the optimal duty cycle, u and v both conform to uniform distribution, i.e. u ~ N(0, σ u 2 ) and v ~ N(0, σ v 2 ), Γ is a standard gamma function; ⑥To improve the convergence speed of the algorithm, the elimination mechanism of the cuckoo biological heuristic algorithm is used, and the elimination probability p a Discard the minimum output power P of the tth iteration of the photovoltaic array pv (d t worst ) corresponding to the bird nest, that is, the tth iteration of the bad duty cycle d t worst , and update it to the new duty cycle d t worst,new .

6. The MPPT control method of a marine floating photovoltaic power generation system according to claim 5, characterized in that The step ⑥ is specifically referring to: in order to improve the algorithm convergence speed and reduce the steady-state oscillation, it is not necessary to eliminate each bird nest, i.e. duty cycle d i after each iteration, but only the worst bird nest, i.e. the tth iteration of the worst duty cycle d t worst eliminate, and replace the adjacent bird nest of the worst bird nest in the traditional cuckoo algorithm, i.e. the optimal bird nest, i.e. the optimal duty cycle d best as shown in formulas (6) and (7): That is: after each iteration, the minimum value of the output power P pv (d i t ) of the selected offshore floating photovoltaic array, that is, min{P pv (d i t ), i≤n}, the corresponding d t worst is selected as the bad individual; a random number r i ∈(0, 1) is taken for the bad individual i When r a is greater than the elimination probability p a , the new individual generation mechanism of formula (7) is used to replace the bad duty cycle d t worst of the tth iteration with a new duty cycle d t worst,new , and the improved n bird nests, that is, the tth iteration duty cycle d1 t ~d n t are obtained, and step ③ is entered.

7. The method of claim 1, wherein the method further comprises: The step (3) improves the biological heuristic algorithm based on the initial optimized distribution position of the population individuals at 0 time provided in step (2), that is, the initial duty cycle d1 0 ~d n 0 The improved gray wolf algorithm is realized by improving the elimination rule of bad individuals, and specifically includes: (I) setting the number of gray wolf individuals n h , elimination probability p ah and iteration termination criterion; (II) the initial distribution d of the population determined by step (2) 1h 0 ~d nh 0 The initial position of the n wolves in the improved grey wolf algorithm is improved, that is, the initial duty cycle d h 1h 0 ~d nh 0 ;​ (III) the tth iteration of the gray wolf position, i.e. the duty cycle d 1h t ~d nh t In turn applied to the DC-DC rectifier unit connected to the offshore floating photovoltaic array, the DC-DC rectifier unit respectively outputs the tth iteration of the duty cycle d 1h t ~d nh t The corresponding offshore floating photovoltaic array tth iteration output power P pv (d 1h t )~P pv (d nh t ), compare and record the maximum power P max and the power corresponding to the gray wolf position, i.e. the duty cycle d besth ; wherein t≥0; (IV) check the position of each wolf individual, i.e. the duty cycle d of the tth iteration 1h t ~ d nh t the corresponding photovoltaic array tth iteration output power P pv (d 1h t ~ P pv (d nh t the difference between the photovoltaic array maximum output power point P max in the iteration process and the set threshold value ε, i.e. whether formula (8) is satisfied: d ih t For the i-th gray wolf, i.e. the duty cycle at the position of the t-th iteration, if the result satisfies equation (8), go to step (V), i.e. update n h the position of the i-th gray wolf individual, i.e. the duty cycle variable; If the position of each gray wolf, i.e., the position of the gray wolf in the t-th iteration, is equal to the duty cycle d... 1h t ~d nh t The corresponding photovoltaic array output power P in the t-th iteration pv (d 1h t )~P pv (d nh t The point P with the maximum output power of the photovoltaic array during the iteration process max If the difference is less than or equal to the set threshold ε, that is, if equation (9) is satisfied, then the latest maximum power point (d) at the end of the iteration is recorded. GMPPnew V GMPPnew ,P GMPPnew ,t new ), and store it in the online database of step (1). At this point, the iteration of the improved gray wolf algorithm ends, and the process proceeds to step (4) to use the maximum power point (d) GMPPnew V GMPPnew ,P GMPPnew ,t new The perturbation is performed as the initial position for variable step size perturbation observation or incremental conductance method; (V) the gray wolf position, i.e., the duty cycle d 1h t ~ d nh t the corresponding photovoltaic array output power P pv (d 1h t ) ~ P pv (d nh t ) is sorted from large to small, P pv (d 1h t ) ~ P pv (d nh t ) maximum value P max the corresponding gray wolf position named alpha wolf, P pv (d 1h t ) ~ P pv (d nh t ) the second largest value corresponding to the gray wolf position named beta wolf, P pv (d 1h t ) ~ P pv (d nh t ) the third largest value corresponding to the gray wolf position s, named delta wolf, and the remaining wolf positions are divided into omega wolves; for each omega wolf its new position is calculated according to the updating formula of the wolf pack position in the search process shown by formula (10) involving the positions of alpha wolf, beta wolf and delta wolf, aiming to imitate the hunting behavior of gray wolves and help explore to obtain better solutions; In formula (10), is the position of the i-th wolf in the ω at the t-th iteration, is the position of the i-th wolf in the ω at the t+1-th iteration, is the position of the prey, i.e. the maximum output power point P of the photovoltaic array in the iteration process max corresponding position, A, C are coefficient vectors, D is the distance between the gray wolf individual and the prey, i.e. and the distance, a is a convergence factor that linearly decreases in the interval [0, 2] for the number of random iterations, r1 and r2 are random vectors with values in the interval [0, 1]. (VI) When the position of the prey is determined, i.e., the optimal duty cycle is obtained, the position updating formula is shown as formula (11): In formula (11), D a , D β , D δ is the distance between the position of the gray wolf individual and the positions of the alpha wolf, the beta wolf and the delta wolf, i.e. and ; A1, A2, A3, C1, C2, C3 are the coefficient vectors described in formula (2), and d1, d2 and d3 are the distance and direction of the advance of the omega wolf individual to the alpha wolf, the beta wolf and the delta wolf, i.e. to ; (Ⅶ) To improve the convergence speed of the algorithm, the improved mechanism of eliminating bad individuals is used, that is, after each iteration, according to the elimination probability p ah Discard the gray wolf position corresponding to the minimum output power of the photovoltaic array, that is Corresponding Update d t worst,new As formula (12), in the process of algorithm iteration, randomly take r ih ∈(0,1), when r i Is greater than the elimination probability p ah , the duty cycle d besth , that is, the maximum power P max In the iteration process and the gray wolf position corresponding to the power, the position of the worst gray wolf individual is improved, that is Get the improved n gray wolf position, that is, the duty cycle d 1h t ~d nh t , and turn to step (Ⅲ); 8. The method of claim 1, wherein the method further comprises: The step (4) specifically refers to: (4-1) The disturbance voltage of the variable step size disturbance observation or variable step size conductance increment method is calculated by formula (13): where ΔV is the variable step perturbation voltage, dP pv where dV is the photovoltaic array output power fluctuation and voltage fluctuation at the previous perturbation time, λ is the step scaling factor, and ΔV1 and ΔV2 are the upper and lower limits of the step, respectively. (4-2) In the variable step MPPT technology, when the output power value P pv (V t ) at time t is detected, and the difference between the latest maximum power P GMPP,i at the end of the iteration of step (3) and P is greater than the set proportion θ, i.e. formula (14) is satisfied, then return to step (2) to restart the improved bio-inspired algorithm for maximum power tracking. When the output power value P is detected at time t... pv (V t The latest maximum power P at the end of step (3) iteration GMPP,i When the difference is less than or equal to the set ratio θ, that is, when equation (15) is satisfied, return to step (4-1) and continue the perturbation operation to ensure that the floating photovoltaic array at sea always operates quickly and accurately near the theoretical maximum power point;

Images