Fast mppt system based on photovoltaic model fitting

Abstract

This invention discloses a fast MPPT system based on photovoltaic model fitting, comprising: a photovoltaic model parameter identification and fitting module, which constructs the objective function and error function required for model parameter identification and solves them using a cloud drift optimization algorithm; a maximum power point voltage search module, which, before each MPPT start, inputs the real-time ambient temperature T and solar irradiance G into an offline-trained BP neural network to obtain the current model parameters, and uses a goat optimization algorithm to search for the maximum power point voltage within a dynamic boundary determined by the open-circuit voltage; and a photovoltaic reference voltage control output module, which uses a dual closed-loop structure to receive the maximum power point voltage, and sets the PWM duty cycle of the buck and boost switches in segments according to the PID output value, so that the photovoltaic array operates at the maximum power point. The fast MPPT system based on photovoltaic model fitting provided by this invention intelligently and quickly locks the maximum power point through a herd of goats, and then uses dual closed-loop fine voltage control to keep the array in a state of high-efficiency power generation.

Description

Technical Field

[0001] This invention specifically relates to a fast MPPT system based on photovoltaic model fitting. Background Technology

[0002] The output power of photovoltaic arrays exhibits nonlinear multi-peak characteristics with variations in solar irradiance, temperature, and shading, making maximum power point tracking (MPPT) a core component for improving power generation efficiency. Existing technologies primarily employ perturbation-observation methods, incremental conductance methods, or intelligent methods such as particle swarm optimization and genetic algorithms. Conventional perturbation methods are simple to implement but prone to oscillations and misjudgments under rapid irradiance changes; incremental conductance methods, while offering slightly higher accuracy, rely on high-precision sampling and fail under low illumination. While particle swarm optimization and genetic algorithms, introduced in recent years, can escape local optima, they are computationally intensive and have slow convergence, making it difficult to balance real-time performance with hardware resource constraints. Furthermore, existing solutions often treat model parameters as fixed values, failing to reflect long-term changes such as component aging and temperature drift, leading to model mismatch and further reducing tracking accuracy. Therefore, there is an urgent need for an MPPT system capable of online model identification, rapid global optimization, and hardware-friendly design to address the shortcomings of traditional methods in terms of accuracy, speed, and adaptability. Summary of the Invention

[0003] This invention provides a fast MPPT system based on photovoltaic model fitting to solve the technical problem of insufficient accuracy of traditional early warning methods mentioned above. The specific technical solution is as follows:

[0004] A fast MPPT system based on photovoltaic model fitting includes:

[0005] The photovoltaic model parameter identification and fitting module is used to receive the photovoltaic model type and array row and column number selected by the user, automatically construct the objective function and error function required for model parameter identification based on the type, and use the cloud drift optimization algorithm to iteratively solve within the preset constraint space to obtain the optimal model parameters.

[0006] The maximum power point voltage search module is used to input the real-time ambient temperature T and solar radiation irradiance G into the offline-trained BP neural network before each MPPT start to obtain the current model parameters in real time, and then use the goat optimization algorithm to search for the maximum power point voltage within the dynamic boundary determined by the open circuit voltage.

[0007] The photovoltaic reference voltage control output module is used to receive the maximum power point voltage using a four-switch buck-boost circuit as the power stage and a dual closed-loop structure composed of an outer voltage PID and an inner current PID. The module also sets the PWM duty cycle of the buck and boost switches in segments according to the PID output value, so that the photovoltaic array continues to operate at the maximum power point.

[0008] Furthermore, the objective function of the model is the root mean square error of the error function:

[0009]

[0010] Where N is the number of experimental samples, f(V) L ,I L Let X be the error function, and V be the error function. L For the output voltage, I L Let X be the output current, and let X be the photovoltaic model parameter to be solved.

[0011] Furthermore, the error function of the single-diode model is:

[0012]

[0013] Among them, I ph For photocurrent, I sd R is the reverse saturation current of the diode, a is the diode ideality factor, and R is the diode reverse saturation current. s R is the equivalent series resistance. sh The resistance is the equivalent parallel resistance, and k is the Boltzmann constant, which has a value of 1.3806503 × 10⁻⁶. -23 J / K, T is the Kelvin temperature, measured using a temperature sensor, and q is the electron charge, with a value of 1.60217646 × 10⁻⁶. -19 C,V L For the output voltage, I L This is the output current.

[0014] Furthermore, the error function of the dual-diode model is:

[0015]

[0016] Among them, I ph For photocurrent, I sd1 Let a1 be the reverse saturation current of the first diode, and I be the ideality factor of the first diode. sd2 R is the reverse saturation current of the second diode, a2 is the ideality factor of the second diode, and R s R is the equivalent series resistance. sh The resistance is the equivalent parallel resistance, and k is the Boltzmann constant, which has a value of 1.3806503 × 10⁻⁶. -23 J / K, T is the Kelvin temperature, measured using a temperature sensor, and q is the electron charge, with a value of 1.60217646 × 10⁻⁶. -19 C,V L For the output voltage, I L This is the output current.

[0017] Furthermore, the error function of the three-diode model is:

[0018]

[0019] Among them, I ph For photocurrent, I sd1 Let a1 be the reverse saturation current of the first diode, and I be the ideality factor of the first diode. sd2 a2 is the reverse saturation current of the second diode, and I is the ideality factor of the second diode. sd3 R is the reverse saturation current of the third diode, a3 is the ideality factor of the third diode, and R s R is the equivalent series resistance. sh The resistance is the equivalent parallel resistance, and k is the Boltzmann constant, which has a value of 1.3806503 × 10⁻⁶. -23 J / K, T is the Kelvin temperature, measured using a temperature sensor, and q is the electron charge, with a value of 1.60217646 × 10⁻⁶. -19 C,V L For the output voltage, I L This is the output current.

[0020] Furthermore, the number of photovoltaic modules connected in series and in parallel can be set through the user interface.

[0021] The error function of the photovoltaic module model is:

[0022]

[0023] Among them, I ph For photocurrent, I sd R is the reverse saturation current of the diode, a is the diode ideality factor, and R is the diode reverse saturation current. s R is the equivalent series resistance. sh For the equivalent parallel resistance, N s N represents the number of photovoltaic cells connected in series. p Let be the number of photovoltaic cells connected in parallel, and k be the Boltzmann constant, with a value of 1.3806503 × 10⁻⁶. -23 J / K, T is the Kelvin temperature, measured using a temperature sensor, and q is the electron charge, with a value of 1.60217646 × 10⁻⁶. -19 C,V L For the output voltage, I L This is the output current.

[0024] Furthermore, the cloud drift optimization algorithm includes:

[0025] With N c Each cloud particle represents a parameter vector to be identified and is randomly initialized within the global search space Ω;

[0026] The adaptive weights w of each cloud particle are dynamically updated based on the current fitness values. i [j];

[0027] Alternately execute the position update formula during the local development phase and the position update formula during the global exploration phase.

[0028] The position update formula for alternating local development phases is:

[0029]

[0030] Among them, X * [j] represents the j-th element of the optimal cloud particle in the current iteration round t, ν b [j] = U(﹣0.2m, 0.2m), which is a function of factor m = (1﹣t / T) c A small random number, T, is used to control the local development program of cloud particles. c To determine the maximum number of iterations, n1 and n2 are in the range [1, N]. c A random integer between ] represents two randomly selected cloud particles.

[0031] Position update formula during global exploration phase:

[0032]

[0033] Where, ν c [j] = U(﹣0.2b, 0.2b) is a function of the factor b = (1﹣t / T) c The small random number is adjusted to control the degree of randomness in the global search of cloud particles, simulating the random drift motion of clouds in the natural atmosphere.

[0034] Perform random perturbation reinitialization with probability z until the accuracy ε is met or the maximum number of iterations T is reached. c .

[0035] Furthermore, the training methods for BP neural networks include:

[0036] Every 30 minutes, ambient temperature T, solar irradiance G, and the corresponding IV curve sampling array are collected from a temperature sensor and a total solar radiation sensor. L V L ] 2×n n is the number of IV samples, V L For the output voltage, I L For output current;

[0037] Using [T,G] as input and the optimal model parameters X calculated by the cloud drift optimization algorithm as output, a training set and a test set are constructed.

[0038] The error backpropagation algorithm is used to train until convergence, and finally the mapping relationship is established.

[0039] Furthermore, the ratio of the training set to the test set is 7:3.

[0040] Furthermore, the goat optimization algorithm includes:

[0041] Within the voltage range determined by the open-circuit voltage, two representative sheep are placed at the empirically optimal position and the model-predicted position, respectively, and the remaining sheep are randomly scattered to form the initial flock.

[0042] With the PID closed-loop control, the instantaneous power generation of each sheep was measured point by point, and the sheep with the highest power was identified as the leader.

[0043] The sheep sequentially perform random foraging, follow the lead sheep, and jump randomly, continuously updating their respective positions.

[0044] Eliminate the worst-performing individuals and regenerate them randomly to maintain population diversity;

[0045] When the power change exceeds the set threshold or the iteration converges, the leader position is used as the maximum power point voltage output; if the environment changes abruptly, the aforementioned steps are immediately restarted.

[0046] Furthermore, the four-switch buck-boost circuit includes:

[0047] The voltage outer loop PID outputs a reference current, which serves as the reference value for the current inner loop PID.

[0048] The inner loop PID controller outputs the total output range u. pid ∈[0,2];

[0049] PWM drive unit, used to adjust according to u pid The duty cycles of the buck and boost switches are set in segments so that:

[0050] When u pid When <0.95, the Buck comparison value = 0.95·u pid ·T pwm Boost comparison value = 0, T pwm The switching cycle of the switching transistor in the buck-boost circuit;

[0051] When 0.95≤u pid When ≤1.05, the Buck comparison value = 0.95·u pid ·T pwm Boost comparison value = 0.95·(u pid -0.95)·T pwm ;

[0052] When u pid When the value is greater than 1.05, the Buck comparison value = T. pwmBoost comparison value = 0.95·(u pid -0.95)·T pwm .

[0053] The fast MPPT system based on photovoltaic model fitting provided by this invention automatically corrects the photovoltaic model according to changes in light and temperature, and uses the intelligence of a herd of goats to quickly lock the maximum power point. Then, through dual closed-loop fine voltage control, the array is always in a state of high-efficiency power generation. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a schematic diagram of a fast MPPT system based on photovoltaic model fitting according to the present invention. Detailed Implementation

[0056] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0057] like Figure 1 The image shows a fast MPPT system based on photovoltaic model fitting according to this application, including: a photovoltaic model parameter identification and fitting module, a maximum power point voltage search module, and a photovoltaic reference voltage control output module.

[0058] The photovoltaic model parameter identification and fitting module receives the photovoltaic model type and array row and column number selected by the user, automatically constructs the objective function and error function required for model parameter identification based on the type, and uses the cloud drift optimization algorithm to iteratively solve within the preset constraint space to obtain the optimal model parameters.

[0059] Understandably, the type of photovoltaic model can be set in the front end of the mini program. If it is a single photovoltaic panel, you can choose the single diode model (SDM), dual diode model (DDM), or triple diode model (TDM). If it is a photovoltaic array module, you can choose the photovoltaic module model (MM) and set the number of rows and columns of the array.

[0060] The maximum power point voltage search module inputs the real-time ambient temperature T and solar irradiance G into an offline-trained BP neural network before each MPPT startup to instantly obtain the current model parameters. Then, a goat optimization algorithm searches for the maximum power point voltage within the dynamic boundary determined by the open-circuit voltage. The photovoltaic reference voltage control output module uses a four-switch buck-boost circuit as the power stage, employing a dual closed-loop structure consisting of an outer voltage PID and an inner current PID to receive the maximum power point voltage. The PWM duty cycle of the buck and boost switches is set segmentally according to the PID output value, ensuring the photovoltaic array continuously operates at the maximum power point. This application's fast MPPT system based on photovoltaic model fitting automatically corrects the photovoltaic model according to changes in light and temperature, and uses the goat herd intelligence to quickly lock the maximum power point. Then, through dual closed-loop fine voltage control, the array is always in a high-efficiency power generation state.

[0061] Based on the type of photovoltaic model set, the objective function required for model parameter identification is automatically constructed. The objective function is the root mean square error of the error function.

[0062]

[0063] Where N is the number of experimental samples, f(V) L ,I L Let X be the error function, and V be the error function. L For the output voltage, I L Let X be the output current, and let X be the photovoltaic model parameter to be solved.

[0064] Different types of photovoltaic models have different error functions. The error function for the single-diode model is:

[0065]

[0066] Among them, I ph For photocurrent, I sd R is the reverse saturation current of the diode, a is the diode ideality factor, and R is the diode reverse saturation current. s R is the equivalent series resistance. sh The resistance is the equivalent parallel resistance, and k is the Boltzmann constant, which has a value of 1.3806503 × 10⁻⁶. -23 J / K, T is the Kelvin temperature, measured using a temperature sensor, and q is the electron charge, with a value of 1.60217646 × 10⁻⁶. -19 C,V L For the output voltage, I L This is the output current.

[0067] The error function of the dual-diode model is:

[0068]

[0069] Among them, I ph For photocurrent, I sd1 Let a1 be the reverse saturation current of the first diode, and I be the ideality factor of the first diode. sd2 R is the reverse saturation current of the second diode, a2 is the ideality factor of the second diode, and R s R is the equivalent series resistance. sh The resistance is the equivalent parallel resistance, and k is the Boltzmann constant, which has a value of 1.3806503 × 10⁻⁶. -23 J / K, T is the Kelvin temperature, measured using a temperature sensor, and q is the electron charge, with a value of 1.60217646 × 10⁻⁶. -19 C,V L For the output voltage, I L This is the output current.

[0070] Furthermore, the error function of the three-diode model is:

[0071]

[0072] Among them, I ph For photocurrent, I sd1 Let a1 be the reverse saturation current of the first diode, and I be the ideality factor of the first diode. sd2 a2 is the reverse saturation current of the second diode, and I is the ideality factor of the second diode. sd3 R is the reverse saturation current of the third diode, a3 is the ideality factor of the third diode, and R s R is the equivalent series resistance. sh The resistance is the equivalent parallel resistance, and k is the Boltzmann constant, which has a value of 1.3806503 × 10⁻⁶. -23 J / K, T is the Kelvin temperature, measured using a temperature sensor, and q is the electron charge, with a value of 1.60217646 × 10⁻⁶. -19 C,V L For the output voltage, I L This is the output current.

[0073] The number of photovoltaic modules connected in series and in parallel can be set through the user interface. The error function of the photovoltaic module model is:

[0074]

[0075] Among them, I ph For photocurrent, I sd R is the reverse saturation current of the diode, a is the diode ideality factor, and R is the diode reverse saturation current. s R is the equivalent series resistance. sh For the equivalent parallel resistance, N s N represents the number of photovoltaic cells connected in series. pLet be the number of photovoltaic cells connected in parallel, and k be the Boltzmann constant, with a value of 1.3806503 × 10⁻⁶. -23 J / K, T is the Kelvin temperature, measured using a temperature sensor, and q is the electron charge, with a value of 1.60217646 × 10⁻⁶. -19 C,V L For the output voltage, I L This is the output current.

[0076] The optimal parameter values ​​for the photovoltaic model were solved using the Cloud Drift Optimization (CDO) algorithm.

[0077] The constraints for model parameters are set according to the type of photovoltaic model, as shown in Table 1.

[0078] Table 1. Constraints on Photovoltaic Model Parameters

[0079]

[0080] In this application, the cloud drift optimization algorithm includes:

[0081] With N c Each cloud particle represents a parameter vector to be identified and is randomly initialized within the global search space Ω. Specifically, photovoltaic model parameters X are used as cloud particles. Based on the constraints of the photovoltaic model, the global search space Ω for the multi-dimensional vector cloud particles is determined, and N are randomly generated within the search space. c An initialized cloud particle X i 0 The subscript of X represents the sequence number of the cloud particle, the superscript of X represents the iteration round, j represents the element index of the cloud particle, and N c B represents the number of cloud particle swarms. l B represents the lower bound of the search space. u U represents the upper limit of the search space, and U(0,1) represents a random number uniformly distributed within the range [0,1].

[0082]

[0083] The adaptive weights w of each cloud particle are dynamically updated based on the current fitness values. i [j]. Adaptive weights are based on the cloud particle f(X) in the current iteration round. i The fitness value of the cloud particle and the fitness value f of the globally optimal cloud particle. * The fitness value f of the worst cloud particle max The fitness value of a cloud particle is obtained through dynamic updates. It is the objective function value calculated using the cloud particle vector as parameters of the photovoltaic model. ε represents the precision requirement and is a very small value to prevent S from being divided by zero.

[0084]

[0085] S = f * -f max +ε

[0086] Alternately execute the position update formula during the local development phase and the position update formula during the global exploration phase.

[0087] The position update formula for alternating local development phases is:

[0088]

[0089] Among them, X * [j] represents the j-th element of the optimal cloud particle in the current iteration round t, ν b [j] = U(﹣0.2m, 0.2m), which is a function of factor m = (1﹣t / T) c A small random number, T, is used to control the local development program of cloud particles. c To determine the maximum number of iterations, n1 and n2 are in the range [1, N]. c A random integer between ] represents two randomly selected cloud particles.

[0090] Position update formula during global exploration phase:

[0091]

[0092] Where, ν c [j] = U(﹣0.2b, 0.2b) is a function of the factor b = (1﹣t / T) c A small random number is used to control the randomness of the global search of cloud particles, simulating the random drift motion of clouds in the natural atmosphere.

[0093] Perform random perturbation reinitialization with probability z until the accuracy ε is met or the maximum number of iterations T is reached. c Specifically, cloud particles are initialized with a small probability z to randomly represent their positions, simulating the unpredictable changes of clouds in the natural atmosphere and preventing the algorithm from prematurely getting trapped in local optima. The method for updating the positions of randomly perturbed cloud particles is as follows:

[0094] X i [j] = B l [j]+(B u [j]-B l [j])·U(0,1), with probability z

[0095]

[0096] If the fitness value of the optimal cloud particle is less than the accuracy requirement ε or the number of algorithm iterations reaches the maximum value T.c Stop the optimization process and use the position of the optimal cloud particle as the solution X for the photovoltaic model parameters.

[0097] Furthermore, the training methods for BP neural networks include:

[0098] Every 30 minutes, ambient temperature T, solar irradiance G, and the corresponding IV curve sampling array are collected from a temperature sensor and a total solar radiation sensor. L V L ] 2×n n is the number of IV samples, V L For the output voltage, I L The output current is defined as [T, G]. Training and testing sets are constructed using [T, G] as input and the optimal model parameters X, calculated by the cloud drift optimization algorithm, as output. The model is trained until convergence using the error backpropagation algorithm, and a mapping relationship is finally established.

[0099] Specifically, every 30 minutes, the ambient temperature T and solar irradiance G measured by the system temperature sensor and solar total radiation sensor are recorded, the photovoltaic maximum power point tracking algorithm is restarted, and the photovoltaic output voltage V during the execution of the maximum power point tracking algorithm is saved. L and output current value I L As the actual sampling array of the photovoltaic IV characteristic curve [I] L V L ] 2×n The output voltage and output current are measured by the voltage divider measurement simulation detection circuit of the DC-DC converter module and the differential high-precision operational amplifier detection circuit of the shunt. Then, based on the photovoltaic IV sampling array [I L V L ] 2×n Given the ambient temperature T and solar irradiance G, the cloud drift optimization algorithm is used to resolve the parameters X of the photovoltaic model. This set of real-time photovoltaic data is then saved as PV. i =[T i G i ,X i [i] represents the current number of times the photovoltaic (PV) data has been saved. A backpropagation (BP) neural network is used to fit the relationship between ambient temperature T, solar irradiance G, and PV model parameters X. The saved PV data... i Divide into input parameters X i BP =[T i G i Output parameter Y i BP =X i30% of the input and output parameters were randomly selected as the test dataset, and the remaining 70% were used as the training dataset. The BP neural network was trained until the convergence accuracy or the required number of iterations was reached. The fitting relationship of the neural network is recorded as follows:

[0100] X = f ANN (T,G)

[0101] The maximum power point voltage search module uses the Goat Optimization Algorithm (GOA) to track the photovoltaic maximum power point and outputs the photovoltaic reference output voltage V. ref The reference voltage of the DC-DC converter module is used for PID closed-loop control, enabling the photovoltaic system to adaptively operate near its highest power point to generate electricity.

[0102] In this application, the goat optimization algorithm includes: within a voltage range determined by the open-circuit voltage, two representative goats are placed at the empirically optimal position and the model-predicted position, respectively, while the remaining goats are randomly scattered to form an initial flock. Under PID closed-loop control, the instantaneous power generation at each goat's position is measured point-by-point, and the goat with the highest power is designated as the leader. The flock sequentially performs random foraging, follows the leader, and jumps randomly, continuously updating their positions. The worst-performing individual is eliminated and randomly regenerated to maintain population diversity. When the power change exceeds a set threshold or the iteration converges, the voltage output at the leader's position is taken as the maximum power point. If the environment changes abruptly, the aforementioned steps are immediately restarted.

[0103] Specifically, before activating the MPPT control algorithm, the open-circuit voltage of the photovoltaic cell is measured and recorded. Ambient temperature T g Solar radiation illuminance G g The open-circuit voltage of the photovoltaic cells was measured by the voltage divider measurement simulation detection circuit of the DC-DC converter module, the ambient temperature was measured by the system temperature sensor, and the solar radiation illuminance was measured by the system total solar radiation sensor. The environmental variables [T,G] were then input into the BP neural network f. ANN (T g G g Obtain the photovoltaic model parameters X under the current environment. Substitute X into the photovoltaic model to obtain the photovoltaic maximum power point voltage.

[0104] Initialize the constraints and goat population for the goat optimization algorithm. The goat population is set to the output voltage X of the photovoltaic cell. i t =(V Li t ), where subscript i = 1, 2, ..., N g The population number is less than the population size N. g The superscripts t = 1, 2, ..., T gThe number of iterations in the goat optimization algorithm is less than the total number of iterations T. g The population boundary conditions are:

[0105]

[0106] Where δ = 0.1 is the boundary margin coefficient, because open-circuit and short-circuit outputs of photovoltaic systems are unlikely to be used under actual power generation conditions. The initial goat population is:

[0107]

[0108] Where rand(1) is a random number in the range [0,1]. The initial position X1 of goat number 1. 0 Based on the constant voltage method, the maximum power point voltage is approximately the photovoltaic open-circuit voltage. 78%. Goat #2's initial position x2 0 The IV curve was obtained based on the theoretical curve after parameter identification of the photovoltaic model and set as follows: The initial positions of goats 1 and 2 are close to the voltage at the maximum power point, which can improve the search efficiency. The initial positions of other goat populations are randomly generated according to the search space to ensure that the algorithm can achieve global search and avoid getting stuck in local optima under the light spot condition.

[0109] The adaptive value is set to the photovoltaic output power P. i t =X i t ·I i t I i t The output voltage of the photovoltaic cell is equal to X. i t The photovoltaic output current at that time is obtained by sampling from the DC-DC converter. pid =20kHz, MPPT calculation frequency is f mppt =100Hz, dividing one MPPT calculation cycle evenly into population size N g There are N time slices, and the number of PID closed-loop control cycles in each time slice is N. pid =f pid \(f mppt ·N g Sampling traverses all X i t Corresponding I i t , will X i t The photovoltaic output reference voltage V serves as the PID closed-loop controller. ref PID period counter T is reset to zeropid =0, when the photovoltaic output voltage V L Satisfy | V L —V ref When <ζ|, the current value I read by the DC-DC current measurement circuit L Record as I i t Until T pid >N pid Then, proceed with the next sampling. Using the above method, the location X of all goats in the population can be obtained. i t The corresponding fitness value P i t and for P i t Sort them.

[0110] The algorithm mimics the adaptive foraging behavior of goats to explore the globally optimal location. The goat randomly updates its own position.

[0111]

[0112] Where α = 0.05 is the goat exploration coefficient, and R is a random variable that follows a Gaussian distribution N(0,1), making the direction and magnitude of the goat's exploration behavior random.

[0113] By mimicking the following behavior of the lead goat, the group can develop locally optimal positions. Goats gravitating towards the lead goat brings the group closer to the optimal position.

[0114]

[0115] Where X t best The maximum fitness value P in the t-th iteration is t best The corresponding individual goat position, β = 0.5 is the goat development coefficient.

[0116] Mimicking the random jumping behavior of goats to avoid getting trapped in local optima. Goats jump randomly with a certain probability to prevent individuals from getting stuck in unfavorable local positions.

[0117]

[0118] Where ζ = 0.1 is the jump coefficient, X r t It is a randomly selected location of an individual goat.

[0119] The search speed of the algorithm is improved by mimicking the avoidance behavior of goat parasites. The updated goat positions are sequentially set as the photovoltaic output reference voltage of the DC-DC converter, and the fitness value P of each individual in the population is obtained by sampling. i t+1Then sort them, and regenerate random positions for the 20% of goats with the smallest fitness values:

[0120]

[0121] Convergence condition judgment and search algorithm restart judgment, when the maximum number of computation iterations T is reached. g Or when the maximum fitness value of goats remains almost unchanged End the goat optimization algorithm and set the maximum fitness value P. t best The corresponding goat position X t best The photovoltaic (PV) maximum power point voltage is output to the DC-DC converter. If the rate of change of PV power generation is greater than ρ, i.e., |P|... real —P best | / P best When ρ > 0, the goat optimization algorithm is restarted to calculate the new maximum power point of the photovoltaic system after the environmental changes, so that the photovoltaic system always operates at the highest power generation efficiency.

[0122] In the embodiments of this application, a four-switch buck-boost circuit is used as the topology of the DC-DC converter to convert photovoltaic power generation to DC output. The PID closed-loop control circuit uses the photovoltaic output voltage as the voltage outer loop PID and the buck-boost output current as the current inner loop PID for dual PID closed-loop control.

[0123] Specifically, the four-switch buck-boost circuit of the photovoltaic reference voltage control output module includes:

[0124] The voltage outer loop PID controller outputs a reference current, which serves as the reference value for the current inner loop PID controller.

[0125] The inner current loop PID receives the output of the outer voltage loop PID and outputs the total output range u. pid ∈[0,2].

[0126] PWM drive unit, used to adjust according to u pid The duty cycles of the buck and boost switches are set in segments so that:

[0127] When u pid When <0.95, set the complementary PWM timer comparison value of the buck switch transistor to 0.95·u. pid ·T pwm The complementary PWM timer compare value of the boost switch transistor is 0, T pwm This represents the switching cycle of the switching transistor in the buck-boost circuit.

[0128] When 0.95≤u pidWhen ≤1.05, set the complementary PWM timer comparison value of the buck switch transistor to 0.95·u. pid ·T pwm The complementary PWM timer comparison value of the boost switch transistor is 0.95·(u pid -0.95)·T pwm .

[0129] When u pid When the value is greater than 1.05, set the complementary PWM timer comparator value of the buck switch transistor to T. pwm The complementary PWM timer comparison value of the boost switch transistor is 0.95·(u pid -0.95)·T pwm .

[0130] A buck-boost topology DC-DC converter, combined with a photovoltaic parameter identification and simulation module, records and saves the photovoltaic output voltage V. L Current I L The external temperature sensor value T and the total solar radiation sensor value G, together with the maximum power point voltage exploration module, record and save the photovoltaic output voltage and current, and receive the photovoltaic output reference voltage V. ref Implement dual closed-loop PID control.

[0131] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the above embodiments do not limit the present invention in any way, and all technical solutions obtained by equivalent substitution or equivalent transformation fall within the protection scope of the present invention.

Claims

1. A fast MPPT system based on photovoltaic model fitting, characterized in that, include: The photovoltaic model parameter identification and fitting module is used to receive the photovoltaic model type and array row and column number selected by the user, automatically construct the objective function and error function required for model parameter identification based on the type, and use the cloud drift optimization algorithm to iteratively solve within the preset constraint space to obtain the optimal model parameters. The maximum power point voltage search module is used to input the real-time ambient temperature T and solar radiation irradiance G into the offline-trained BP neural network before each MPPT start to obtain the current model parameters in real time, and then use the goat optimization algorithm to search for the maximum power point voltage within the dynamic boundary determined by the open circuit voltage. The photovoltaic reference voltage control output module is used to receive the maximum power point voltage using a four-switch buck-boost circuit as the power stage and a dual closed-loop structure composed of an outer voltage PID and an inner current PID. The module also sets the PWM duty cycle of the buck and boost switches in segments according to the PID output value, so that the photovoltaic array continues to operate at the maximum power point. BP neural network training methods include: Every 30 minutes, ambient temperature T, solar irradiance G, and the corresponding IV curve sampling array are collected from a temperature sensor and a total solar radiation sensor. L V L ] 2×n n is the number of IV samples, V L For the output voltage, I L For output current; Using [T,G] as input and the optimal model parameters X calculated by the cloud drift optimization algorithm as output, a training set and a test set are constructed. The error backpropagation algorithm is used to train until convergence, and finally the mapping relationship is established; The mapping relationship is the relationship between ambient temperature T, solar irradiance G, and photovoltaic model parameters X, fitted using a BP neural network, and is recorded as: X=f ANN (T,G); The four-switch buck-boost circuit includes: The voltage outer loop PID outputs a reference current, which serves as the reference value for the current inner loop PID. The inner loop PID controller outputs the total output range u. pid ∈[0,2]; PWM drive unit, used to adjust according to u pid The duty cycles of the buck and boost switches are set in segments so that: When u pid When <0.95, the Buck comparison value = 0.95·u pid ·T pwm Boost comparison value = 0, T pwm The switching cycle of the switching transistor in the buck-boost circuit; When 0.95≤u pid When ≤1.05, the Buck comparison value = 0.95·u pid ·T pwm Boost comparison value = 0.95·(u pid -0.95)·T pwm ; When u pid When the value is greater than 1.05, the Buck comparison value = T. pwm Boost comparison value = 0.95·(u pid -0.95)·T pwm .

2. The fast MPPT system based on photovoltaic model fitting according to claim 1, characterized in that, The objective function of the model is the root mean square error of the error function: Where N is the number of experimental samples, f(V) L ,I L Let X be the error function, and V be the error function. L For the output voltage, I L Let X be the output current, and let X be the photovoltaic model parameter to be solved.

3. The fast MPPT system based on photovoltaic model fitting according to claim 2, characterized in that, The error function for the single-diode model is: Among them, I ph For photocurrent, I sd R is the reverse saturation current of the diode, a is the diode ideality factor, and R is the diode reverse saturation current. s R is the equivalent series resistance. sh The resistance is the equivalent parallel resistance, and k is the Boltzmann constant, which has a value of 1.3806503 × 10⁻⁶. -23 J / K, T is the Kelvin temperature, measured using a temperature sensor, and q is the electron charge, with a value of 1.60217646 × 10⁻⁶. -19 C,V L For the output voltage, I L This is the output current.

4. The fast MPPT system based on photovoltaic model fitting according to claim 2, characterized in that, The error function of the dual-diode model is: Among them, I ph For photocurrent, I sd1 Let a1 be the reverse saturation current of the first diode, and I be the ideality factor of the first diode. sd2 R is the reverse saturation current of the second diode, a2 is the ideality factor of the second diode, and R s R is the equivalent series resistance. sh The resistance is the equivalent parallel resistance, and k is the Boltzmann constant, which has a value of 1.3806503 × 10⁻⁶. -23 J / K, T is the Kelvin temperature, measured using a temperature sensor, and q is the electron charge, with a value of 1.60217646 × 10⁻⁶. -19 C,V L For the output voltage, I L This is the output current.

5. The fast MPPT system based on photovoltaic model fitting according to claim 2, characterized in that, The error function of the three-diode model is: Among them, I ph For photocurrent, I sd1 Let a1 be the reverse saturation current of the first diode, and I be the ideality factor of the first diode. sd2 a2 is the reverse saturation current of the second diode, and I is the ideality factor of the second diode. sd3 R is the reverse saturation current of the third diode, a3 is the ideality factor of the third diode, and R s R is the equivalent series resistance. sh The resistance is the equivalent parallel resistance, and k is the Boltzmann constant, which has a value of 1.3806503 × 10⁻⁶. -23 J / K, T is the Kelvin temperature, measured using a temperature sensor, and q is the electron charge, with a value of 1.60217646 × 10⁻⁶. -19 C,V L For the output voltage, I L This is the output current.

6. The fast MPPT system based on photovoltaic model fitting according to claim 2, characterized in that, The number of photovoltaic modules connected in series and in parallel can be set through the user interface. The error function of the photovoltaic module model is: Among them, I ph For photocurrent, I sd R is the reverse saturation current of the diode, a is the diode ideality factor, and R is the diode reverse saturation current. s R is the equivalent series resistance. sh For the equivalent parallel resistance, N s N represents the number of photovoltaic cells connected in series. p Let be the number of photovoltaic cells connected in parallel, and k be the Boltzmann constant, with a value of 1.3806503 × 10⁻⁶. -23 J / K, T is the Kelvin temperature, measured using a temperature sensor, and q is the electron charge, with a value of 1.60217646 × 10⁻⁶. -19 C,V L For the output voltage, I L This is the output current.

7. The fast MPPT system based on photovoltaic model fitting according to claim 1, characterized in that, The cloud drift optimization algorithm includes: With N c Each cloud particle represents a parameter vector to be identified and is randomly initialized within the global search space Ω; The adaptive weights w of each cloud particle are dynamically updated based on the current fitness values. i [j]; Alternately execute the position update formula during the local development phase and the position update formula during the global exploration phase. The position update formula for alternating local development phases is: Among them, X * [j] represents the j-th element of the optimal cloud particle in the current iteration round t, ν b [j]=U(﹣0.2m,0.2m), is a function of factor m=(1﹣t / T) c A small random number, T, is used to control the local development program of cloud particles. c To determine the maximum number of iterations, n1 and n2 are in the range [1, N]. c A random integer between ] represents two randomly selected cloud particles. Position update formula during global exploration phase: Where, ν c [j]=U(﹣0.2b,0.2b) is a function of the factor b=(1﹣t / T) c The small random number is adjusted to control the degree of randomness in the global search of cloud particles, simulating the random drift motion of clouds in the natural atmosphere. Perform random perturbation reinitialization with probability z until the accuracy ε is met or the maximum number of iterations T is reached. c .

8. The fast MPPT system based on photovoltaic model fitting according to claim 1, characterized in that, Goat optimization algorithms include: Within the voltage range determined by the open-circuit voltage, two representative sheep are placed at the empirically optimal position and the model-predicted position, respectively, and the remaining sheep are randomly scattered to form the initial flock. With the PID closed-loop control, the instantaneous power generation of each sheep was measured point by point, and the sheep with the highest power was identified as the leader. The sheep sequentially perform random foraging, follow the lead sheep, and jump randomly, continuously updating their respective positions. Eliminate the worst-performing individuals and regenerate them randomly to maintain population diversity; When the power change exceeds the set threshold or the iteration converges, the leader position is used as the maximum power point voltage output; if the environment changes abruptly, the aforementioned steps are immediately restarted.

Images