Photovoltaic power generation control method and system

Abstract

The present application relates to the field of photovoltaic power generation, more particularly, it relates to a kind of photovoltaic power generation control method and system.A kind of photovoltaic power generation control system, comprising: actual output power monitoring module, to be detected photovoltaic board image acquisition module, environmental feature vector establishment vector, environmental feature map establishment vector, analog switch tube duty cycle output module, photovoltaic power generation delay control strategy establishment module and maximum output power update module.The present application generates the environmental feature vector corresponding to each photovoltaic board, constructs environmental feature map based on the environmental feature vector corresponding to all photovoltaic boards, then predicts the maximum output power of photovoltaic system under current environmental conditions based on environmental feature map, and generates corresponding photovoltaic power generation delay control strategy, so as to realize the consideration of the illumination condition of each photovoltaic board, avoid the influence of local extreme point on maximum output power due to local shielding.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of photovoltaic power generation, more particularly, it relates to a photovoltaic power generation control method and system. BACKGROUND

[0002] Photovoltaic (PV) power generation is a technology that converts solar radiation into electrical energy using solar cells. The key to this technology lies in the efficiency of the photovoltaic cell, that is, its ability to convert sunlight into electrical energy under certain lighting conditions. The design of a photovoltaic power generation system aims to maximize the capture of sunlight and convert it into electrical energy, but in practical applications, various factors can affect the maximum output power of the system.

[0003] In the process of photovoltaic power generation, local shading is an important influencing factor. Local shading can be caused by surrounding buildings, trees or other obstacles, causing part or some of the photovoltaic panels to be unable to receive sufficient sunlight. The impact of this local shading is far more than just the efficiency reduction of a single shaded cell; since the cells in a photovoltaic panel are usually connected in series, the output drop of one cell will affect the power output of the entire photovoltaic panel, and the output drop of one photovoltaic panel will affect the power output of the entire photovoltaic system. SUMMARY

[0004] The present application obtains the corresponding lighting conditions by image analysis of each photovoltaic panel, and generates the corresponding environmental feature vector of each photovoltaic panel based on the lighting conditions, constructs an environmental feature map based on the environmental feature vectors of all photovoltaic panels, and predicts the maximum output power of the photovoltaic system under the current environmental conditions based on the environmental feature map, and generates the corresponding photovoltaic power generation delay control strategy, thereby considering the lighting conditions of each photovoltaic panel, avoiding the influence of local extreme points on the maximum output power caused by local shading, and reducing the dependence on manual intervention, reducing the operation and maintenance cost, improving the automation level of the system, and avoiding the potential impact on system stability caused by rapid or sudden changes through the photovoltaic power generation delay control strategy.

[0005] A photovoltaic power generation control method, comprising:

[0006] The actual output power q of the photovoltaic system is continuously monitored, and it is determined whether the photovoltaic control restart condition is met, the photovoltaic control restart condition being "(η-q) / q≥ε", wherein η is the maximum output power currently recorded, which is updated through the photovoltaic control restart operation subsequently, and ε is a power deviation threshold; if the photovoltaic control restart condition is met, the photovoltaic control restart operation is performed; if the photovoltaic control restart condition is not met, the output power q of the photovoltaic system is continuously monitored;

[0007] The photovoltaic control restart operation includes the following contents:

[0008] Obtaining a to-be-detected photovoltaic panel image of each photovoltaic panel, the photovoltaic panel image being an RGB image;

[0009] For each to-be-detected photovoltaic panel image, processing the to-be-detected photovoltaic panel image through the trained image repairing model to obtain a non-stain photovoltaic panel image, performing illumination condition analysis on the non-stain photovoltaic panel image to output an illumination intensity distribution matrix; processing the illumination intensity distribution matrix through a multilayer perceptron to obtain an illumination distribution feature vector; performing spectral analysis on the non-stain photovoltaic panel image to output a spectral feature vector; and splicing the illumination distribution feature vector and the spectral feature vector and an environmental temperature value to obtain an environmental feature vector;

[0010] Splicing all the environmental feature vectors to obtain an environmental feature map, each coordinate point in the environmental feature map corresponding to a photovoltaic panel, and each coordinate point storing an environmental feature vector of the corresponding photovoltaic panel;

[0011] Sending the environmental feature map into the trained maximum power output model for processing to output a simulated switch tube duty cycle, the simulated switch tube duty cycle being a ratio of a time when a switch tube in a photovoltaic inverter is turned on, and the simulated switch tube duty cycle can be used to adjust an output voltage and an output current in a photovoltaic system, thereby adjusting an output power, and simultaneously obtaining a maximum output power η based on the simulated switch tube duty cycle; outputting a photovoltaic power generation delay control strategy based on the simulated switch tube duty cycle, and updating the maximum output power η based on the simulated maximum output power;

[0012] Controlling the photovoltaic system based on the photovoltaic power generation delay control strategy.

[0013] Preferably, the image repairing model is established based on a generated U-net network;

[0014] The training of the image repairing model includes the following steps:

[0015] Obtaining a plurality of photovoltaic panel images, the photovoltaic panel images being images actually photographed by a user, for each photovoltaic panel image, combining the photovoltaic panel image and a standard photovoltaic panel image corresponding thereto to form a first training sample; combining all the first training samples to form a first training sample set, and then sending the first training sample set into a parameter-initialized image repairing model for training, during the training, adjusting weight parameters with the standard photovoltaic panel image as a target, determining whether a first training condition is met, if the first training condition is met, outputting the trained image repairing model; otherwise, continuing to iteratively train through the first training sample set.

[0016] Preferably, the non-stain photovoltaic panel image is subjected to illumination condition analysis to output the illumination intensity distribution matrix, and the illumination condition analysis specifically includes the following contents:

[0017] The non-stain photovoltaic panel image is sent into the trained illumination distribution detection model for processing, and an illumination intensity distribution matrix is output.

[0018] The illumination distribution detection model is established based on a U-net network, and includes a U-net network layer, a self-attention mechanism layer, and an illumination intensity distribution matrix output layer.

[0019] Preferably, the non-stain photovoltaic panel image is sent into the trained illumination distribution detection model for processing, and the processing specifically includes the following steps:

[0020] The non-stain photovoltaic panel image is sent into the U-net network layer for processing to obtain illumination distribution features T.

[0021] In the self-attention mechanism layer, the illumination distribution features T are multiplied by a key weight matrix Wk to obtain a key matrix K, and the illumination distribution features T are multiplied by a value weight matrix Wp to obtain a value matrix P; the non-stain photovoltaic panel image is converted to HSV space through color conversion, and a feature map corresponding to the V channel is selected as a brightness feature map; the brightness feature map is multiplied by a query weight matrix WQ to obtain a query matrix Q, and an attention weight matrix ATT = softmax ((Q·KT) / (μ)0.5) is calculated, where μ is the number of columns of the key matrix K; the attention weight matrix ATT is multiplied by the value matrix P to obtain reinforced illumination distribution features T.

[0022] In the illumination intensity distribution matrix output layer, the reinforced illumination distribution features T are subjected to convolution operation to output an illumination intensity distribution matrix.

[0023] Preferably, the training of the illumination distribution detection model includes the following steps:

[0024] All photovoltaic panel images in a first training sample set are obtained, and all photovoltaic panel images in the first training sample set are processed through a trained image inpainting model to obtain a plurality of non-stain photovoltaic panel images; each non-stain photovoltaic panel image and its corresponding standard illumination intensity distribution matrix form a second training sample, all second training samples form a second training sample set, and the second training sample set is sent into a parameter-initialized illumination distribution detection model for training; during the training, the standard illumination intensity distribution matrix is used as a target for weight parameter adjustment, and it is determined whether a second training condition is met; if the second training condition is met, a trained illumination distribution detection model is output; otherwise, iterative training is continued through the second training sample set.

[0025] Preferably, the maximum power output model is established based on a convolutional neural network;

[0026] The training of the maximum power output model comprises the following steps:

[0027] All photovoltaic panel images obtained in a plurality of time periods are obtained, and the environmental feature maps corresponding to all photovoltaic panel images in any time period are taken as third training samples, so as to mark the third training samples by the simulated switch tube duty cycle output by the improved firefly optimization algorithm; all marked third training samples are combined to form a third training sample set, and the third training sample set is sent to the maximum power output model with initialized parameters for training, during which the simulated switch tube duty cycle corresponding to the third training sample is taken as a target to calculate a loss value, if the loss value is within a preset range, the preset range is artificially set to constrain the accuracy of the maximum power output model, and a trained maximum power output model is output; otherwise, the iteration training is continued through the third training sample set.

[0028] Preferably, the simulated switch tube duty cycle is output by the improved firefly optimization algorithm, and the method comprises the following steps:

[0029] S1: N firefly individuals Fn are established, n=1, 2, 3…N, and Fn stores the to-be-simulated switch tube duty cycle;

[0030] S2: a maximum iteration number M is set, and m=1, h is used to record the iteration number;

[0031] S3: the fitness δn of the firefly individual Fn is calculated, all firefly individuals Fn are arranged in descending order according to the corresponding fitness δn, all arranged firefly individuals Fn form a population set, and the first ζ firefly individuals Fn form a flame individual set, in the initial state, ζ=N, and the firefly individual Fn in the flame individual set is recorded as the flame individual Ri, i=1, 2, 3…ζ;

[0032] S4: the firefly individual Fn is selected in order from the population set, and its order position j in the population is recorded, and it is judged whether “j>ζ” is established, if “j>ζ” is not established, the flame individual Ri located at the j position is selected from the flame individual set, and recorded as the target flame individual Fn-tar corresponding to the selected firefly individual Fn, otherwise, if “j>ζ” is established, a flame individual Ri is selected as the target flame individual Fn-tar corresponding to the selected firefly individual Fn in the flame individual set by using the roulette selection algorithm according to the fitness δn of all firefly individuals Fn, and the selected firefly individual Fn is updated based on the target flame individual Fn-tar according to the following formula:

[0033] ;

[0034] wherein dis(Fn, Fn-tar) is the distance between the moth individual Fn and the target flame individual Fn-tar; c is a logarithmic spiral shape constant; t is a random number between the interval [-1, 1];

[0035] S5: updating the number ζ of flame individuals Ri in the flame individual set through the formula: ζ = round[(N-m)(N-1) / M];

[0036] S6: judging whether "m < M" is true, if "m < M" is true, going to S7; if "m < M" is not true, going to S8;

[0037] S7: calculating the fitness δn corresponding to the moth individual Fn, and arranging all the moth individuals Fn in descending order according to the corresponding fitness δn, reconstituting the population set from the arranged moth individuals Fn, and arranging all the moth individuals Fn in the population set and all the moth individuals Fn in the flame individual set in descending order according to the corresponding fitness δn, selecting the first ζ moth individuals Fn to form the flame individual set, and returning to S4;

[0038] S8: selecting the moth individual Fn with the maximum fitness δn from all the moth individuals Fn in the population set and all the moth individuals Fn in the flame individual set as the output of the simulated switch tube duty cycle.

[0039] Preferably, N moth individuals Fn are established, specifically including the following steps:

[0040] S1.1: for each moth individual Fn, the following formula is used for assignment operation:

[0041] ;

[0042] S1.2: repeating step S1.1 N times to establish N moth individuals Fn.

[0043] Preferably, the fitness δn corresponding to the moth individual Fn is calculated, specifically including the following steps: on the basis of the actually used photovoltaic system, a corresponding photovoltaic system simulation model is established, the environmental conditions corresponding to each third training sample are set, the simulated switch tube duty cycle corresponding to the moth individual Fn is used as a parameter to set the photovoltaic system simulation model, photovoltaic power generation simulation is performed through the photovoltaic system simulation model, and the output power is used as the fitness δn; and the fitness corresponding to the simulated switch tube duty cycle is the maximum output power.

[0044] A photovoltaic power generation control system, comprising:

[0045] An actual output power monitoring module is configured to continuously monitor the actual output power of the photovoltaic system and determine whether the actual output power meets a photovoltaic control restart condition.

[0046] A to-be-detected photovoltaic panel image acquisition module is configured to acquire to-be-detected photovoltaic panel images of each photovoltaic panel.

[0047] An environmental feature vector establishing vector is configured to establish an environmental feature vector corresponding to each photovoltaic panel for each to-be-detected photovoltaic panel image.

[0048] An environmental feature map establishing vector is configured to splice all the environmental feature vectors to obtain an environmental feature map, in which each coordinate point corresponds to a photovoltaic panel, and each coordinate point stores an environmental feature vector of the corresponding photovoltaic panel.

[0049] An analog switch tube duty cycle output module is configured to input the environmental feature map into a trained maximum power output model for processing and output an analog switch tube duty cycle.

[0050] A photovoltaic power generation delay control strategy establishing module is configured to output a photovoltaic power generation delay control strategy based on the analog switch tube duty cycle.

[0051] A maximum output power updating module is configured to obtain a maximum output power based on the analog switch tube duty cycle and update the maximum output power based on the analog maximum output power.

[0052] The present application has the following advantages:

[0053] 1、The present application analyzes each photovoltaic panel through image analysis, acquires the corresponding illumination condition, generates an environmental feature vector corresponding to each photovoltaic panel based on the illumination condition, constructs an environmental feature map based on the environmental feature vectors corresponding to all photovoltaic panels, predicts the maximum output power of the photovoltaic system under the current environmental condition based on the environmental feature map, and generates a corresponding photovoltaic power generation delay control strategy, thereby considering the illumination condition of each photovoltaic panel, avoiding the influence of local extreme points on the maximum output power caused by local shielding, reducing the dependence on manual intervention, reducing the operation and maintenance cost, improving the automation level of the system, and avoiding the potential influence on system stability caused by rapid or sudden changes through the photovoltaic power generation delay control strategy.

[0054] 2、In the application, the U-net network layer first extracts features from the non-stripe photovoltaic panel image, and the U-net network is selected due to its excellent performance in image segmentation field, which can effectively extract complex features from the image, which is particularly important for the detection of light distribution, then the features are strengthened through the self-attention mechanism, further improving the understanding and analysis ability of the model to the light distribution features, the self-attention mechanism can highlight important features and suppress unimportant information by calculating the correlation between the weight matrix and the features, so that the light intensity distribution detection is more accurate, especially when dealing with the problem of uneven light caused by sunlight reflection, this mechanism shows its strong effect, in the self-attention mechanism layer, the use of brightness feature map is based on the characteristics of HSV color space, effectively utilizing the brightness information of the image, further enhancing the recognition ability of the light distribution features, through a series of steps, the final light intensity distribution matrix not only accurately reflects the light conditions of the photovoltaic panel, but also provides key data for subsequent photovoltaic system optimization.

[0055] 3、The application initializes the individuals through Fuch chaotic mapping, ensures the wide distribution of the initial population in the search space, thereby providing a good foundation for global search, and the fitness δn of each moth individual is calculated through a photovoltaic system simulation model constructed by MATLAB / SIMULINK software, the simulation model simulates the photovoltaic power generation process according to the changes of environmental conditions such as light and temperature, and takes the output power as the measurement standard of fitness; meanwhile, the position of the moth individual, i.e. the value of the duty cycle, is constantly optimized through the iteration process, in each iteration, the position of the moth individual is updated according to the distance between the moth individual and the flame individual and other parameters, this updating process not only considers the current fitness, but also adjusts the size of the flame individual set through self-adaption, ensuring that the algorithm can effectively explore the solution space and accurately utilize the current optimal solution, so as to effectively find the switch duty cycle that maximizes the output power, thereby optimizing the performance of the photovoltaic system. In summary, the application improves the power generation efficiency of the photovoltaic system through simulation and iteration, and improves the application flexibility and stability of the algorithm, which has important significance for improving the overall performance and economy of the photovoltaic power generation technology. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 The structure diagram of the photovoltaic power generation control system used in the embodiment of the application. DETAILED DESCRIPTION

[0057] In order to enable the personnel in the technical field to better understand the technical solutions in the application, the technical solutions in the embodiments of the application will be described clearly and completely below in combination with the drawings in the embodiments of the application.

[0058] Embodiment 1

[0059] A photovoltaic power generation control method, comprising:

[0060] The actual output power q of the photovoltaic system is continuously monitored, and it is determined whether the photovoltaic control restart condition is met, the photovoltaic control restart condition is "(η-q) / q≥ε", wherein η is the maximum output power currently recorded and is updated through the photovoltaic control restart operation subsequently, and ε is a power deviation threshold value, which is set by a configuration file and is generally 0.5; if the photovoltaic control restart condition is met, the photovoltaic control restart operation is performed; if the photovoltaic control restart condition is not met, the output power q of the photovoltaic system is continuously monitored; through the judgment of the photovoltaic control restart condition, the preset power deviation threshold value is used to ensure that the system always operates at a near-optimal performance;

[0061] The photovoltaic control restart operation comprises the following contents:

[0062] The image to be detected of each photovoltaic panel is obtained, and the image to be detected is an RGB image. It should be noted that the entire photovoltaic power generation array comprises a plurality of photovoltaic panels, each of which can convert light energy into electrical energy. By arranging a camera on each photovoltaic panel, the surface of the photovoltaic panel that receives light energy can be imaged to obtain the image to be detected.

[0063] For each image to be detected, the image to be detected is processed by a trained image repair model to obtain a non-spot photovoltaic panel image. When the photovoltaic panel is photographed, due to the reflection of sunlight, glare or light spots may appear in the image to be detected, which may affect the analysis of the light conditions of the photovoltaic panel. Therefore, the image to be detected can be processed by the image repair model, the pixels around the position of the glare or light spot in the image to be detected are repaired, the non-spot photovoltaic panel image is analyzed under the light condition, and a light intensity distribution matrix is output. The light intensity distribution matrix is processed by a multilayer perceptron to obtain a light distribution feature vector. The non-spot photovoltaic panel image is analyzed by spectrum to output a spectrum feature vector. The light distribution feature vector and the spectrum feature vector are spliced with the ambient temperature value to obtain an environmental feature vector. The environmental feature vector can represent the current environmental data of the photovoltaic panel, which is used to output the maximum power point subsequently. Through light intensity analysis and spectrum analysis technology, the environmental feature vector reflecting the light and spectrum conditions of each photovoltaic panel is generated. These environmental feature vectors not only contain light intensity information, but also combine environmental temperature and other factors, thereby comprehensively representing the working environment of the photovoltaic panel.

[0064] The environment feature vectors are spliced to obtain an environment feature map, each coordinate point in the environment feature map corresponds to a photovoltaic panel, and each coordinate point stores the environment feature vector of the corresponding photovoltaic panel; all feature vectors are spliced into an environment feature map, which provides an intuitive and detailed environment data representation for the system. This representation not only helps to monitor the performance of each photovoltaic panel, but also provides data support for the management and optimization of the entire photovoltaic system.

[0065] The environment feature map is input into the trained maximum power output model for processing to output a simulated switch tube duty cycle. The simulated switch tube duty cycle refers to the time ratio of the switch tube conduction in the photovoltaic inverter, which can realize the adjustment of the output voltage and output current in the photovoltaic system, and further adjust the output power. The maximum output power η is obtained based on the simulated switch tube duty cycle. The photovoltaic power generation delay control strategy is output based on the simulated switch tube duty cycle, and the maximum output power η is updated based on the simulated maximum output power. The maximum power output model can accurately predict the ideal output power and the best switch tube duty cycle under the current conditions. This prediction is crucial for dynamically adjusting the operation of the inverter to ensure that the system always operates at the highest efficiency.

[0066] The photovoltaic system is controlled based on the photovoltaic power generation delay control strategy, specifically by adjusting the conduction time ratio of the switch tube in the photovoltaic inverter in the photovoltaic system. The photovoltaic power generation delay control strategy allows the system to make gradual adjustments, avoiding potential impacts on system stability due to rapid or sudden changes. This gradual adjustment mechanism not only improves the response capability of the system, but also reduces the wear and tear on the inverter and other key components, prolonging the overall life of the system.

[0067] The present application obtains the corresponding illumination conditions by image analysis of each photovoltaic panel, generates the corresponding environment feature vector of each photovoltaic panel based on the illumination conditions, constructs an environment feature map based on the environment feature vectors of all photovoltaic panels, predicts the maximum output power of the photovoltaic system under the current environmental conditions based on the environment feature map, and generates the corresponding photovoltaic power generation delay control strategy, thereby considering the illumination conditions of each photovoltaic panel, avoiding the influence of local extreme points on the maximum output power due to local shading, reducing the dependence on manual intervention, reducing the operation and maintenance cost, improving the automation level of the system, and avoiding potential impacts on system stability due to rapid or sudden changes through the photovoltaic power generation delay control strategy.

[0068] The image restoration model is established based on a generated U-net network. The specific settings can refer to the existing U-net network model. In this embodiment, an attention mechanism is added to the U-net network.

[0069] The training of the image restoration model includes the following steps:

[0070] A plurality of photovoltaic panel images are acquired, the photovoltaic panel images being images actually photographed by the user, for each photovoltaic panel image, the photovoltaic panel image and its corresponding standard photovoltaic panel image are used to form a first training sample, the standard photovoltaic panel image being an image obtained by repairing the photovoltaic panel image by software or manually by the user; all the first training samples are used to form a first training sample set, and the first training sample set is sent to the image repairing model with initialized parameters for training, during the training, the standard photovoltaic panel image is used as a target to adjust the weight parameters, and it is determined whether the first training condition is met, the first training condition can be that a certain number of training times is reached or the similarity between the image output by the model and the standard photovoltaic panel image reaches a certain threshold, which is not limited herein, if the first training condition is met, the trained image repairing model is output; otherwise, the first training sample set is iteratively trained.

[0071] The non-stain photovoltaic panel image is subjected to light condition analysis, and a light intensity distribution matrix is output, which specifically includes the following contents:

[0072] The non-stain photovoltaic panel image is sent to the trained light distribution detection model for processing, and a light intensity distribution matrix is output;

[0073] The light distribution detection model is established based on a U-net network, including a U-net network layer, a self-attention mechanism layer and a light intensity distribution matrix output layer, wherein the U-net network layer is used to extract light distribution features of the non-stain photovoltaic panel image, and the U-net network layer refers to an existing U-net network for parameter setting; the self-attention mechanism layer is used to strengthen the light distribution features and improve the accuracy of light intensity distribution detection; the light intensity distribution matrix output layer is used to output the light intensity distribution matrix based on the strengthened light distribution features;

[0074] The non-stain photovoltaic panel image is sent to the trained light distribution detection model for processing, specifically including the following steps:

[0075] The non-stain photovoltaic panel image is sent to the U-net network layer for processing to obtain light distribution features T;

[0076] In the self-attention mechanism layer, the illumination distribution feature T is multiplied by the key-value weight matrix Wk to obtain the key-value matrix K; the illumination distribution feature T is multiplied by the value weight matrix Wp to obtain the value matrix P; the non-streak photovoltaic panel image is converted to HSV space through color conversion, HSV space is a commonly used color model, HSV represents hue (Hue), saturation (Saturation) and brightness (Value), the feature map corresponding to the V channel in the HSV space is selected as the brightness feature map, the brightness feature map is multiplied by the query weight matrix WQ to obtain the query matrix Q, the attention weight matrix ATT=softmax ((Q·KT) / (μ)0.5) is calculated, wherein μ is the number of columns of the key-value matrix K; the attention weight matrix ATT is multiplied by the value matrix P to obtain the reinforced illumination distribution feature T;

[0077] In the illumination intensity distribution matrix output layer, the reinforced illumination distribution feature T is subjected to convolution operation, the convolution kernel size is 1×1, and the illumination intensity distribution matrix is output;

[0078] In the present application, the U-net network layer first extracts features from the non-streak photovoltaic panel image, the U-net network is selected because of its excellent performance in the field of image segmentation, it can effectively extract complex features from the image, which is particularly important for illumination distribution detection, then the features are reinforced through the self-attention mechanism, which further improves the understanding and analysis ability of the model to the illumination distribution feature, the self-attention mechanism can highlight important features and suppress unimportant information by calculating the correlation between the weight matrix and the feature, so that the illumination intensity distribution detection is more accurate, especially when dealing with the problem of uneven illumination caused by sunlight reflection, this mechanism shows its strong effect, in the self-attention mechanism layer, the use of the brightness feature map is based on the characteristics of the HSV color space, effectively utilizing the brightness information of the image, further enhancing the recognition ability of the illumination distribution feature, through a series of steps, the final obtained illumination intensity distribution matrix not only accurately reflects the illumination condition of the photovoltaic panel, but also provides key data for subsequent photovoltaic system optimization;

[0079] The training of the illumination distribution detection model includes the following steps:

[0080] All the photovoltaic panel images in the first training sample set are acquired, and all the photovoltaic panel images in the first training sample set are processed through the trained image repair model to obtain a plurality of non-stain photovoltaic panel images. Each non-stain photovoltaic panel image and its corresponding standard light intensity distribution matrix form a second training sample. The standard light intensity distribution matrix stores the light intensity values obtained by the user through software or actual sensors and standardized. All the second training samples form a second training sample set, and the second training sample set is sent to the parameter initialized light distribution detection model for training. During the training, the standard light intensity distribution matrix is used as the target for weight parameter adjustment. It is judged whether the second training condition is met. The second training condition can be that a certain number of training times are reached or the similarity between the matrix output by the model and the standard light intensity distribution matrix reaches a certain threshold. This is not limited. If the second training condition is met, the trained light distribution detection model is output. Otherwise, the iterative training is continued through the second training sample set.

[0081] The maximum power output model is established based on a convolutional neural network. For details, refer to the existing convolutional neural network based on a regression task;

[0082] The training of the maximum power output model includes the following steps:

[0083] All the photovoltaic panel images acquired in a plurality of time periods are acquired, and the environmental feature maps corresponding to all the photovoltaic panel images in any time period are taken as third training samples. The simulated switch tube duty cycle output by the improved firefly optimization algorithm is used to label the third training samples. All the labeled third training samples form a third training sample set, and the third training sample set is sent to the parameter initialized maximum power output model for training. During the training, the simulated switch tube duty cycle corresponding to the third training sample is used as the target to calculate the loss value. If the loss value is within a preset range, the preset range is artificially set to constrain the accuracy of the maximum power output model, and the trained maximum power output model is output. Otherwise, the iterative training is continued through the third training sample set.

[0084] The simulated switch tube duty cycle is output by the improved firefly optimization algorithm, which includes the following steps:

[0085] S1: N firefly individuals Fn are established, n=1, 2, 3…N, and Fn stores the to-be-simulated switch tube duty cycle;

[0086] S2: The maximum number of iterations M is set, and m=1 is set. h is used to record the number of iterations;

[0087] S3: Calculate the fitness δn corresponding to each moth individual Fn, and arrange all moth individuals Fn in descending order of the corresponding fitness δn, form a population set with all arranged moth individuals Fn, and select the first ζ moth individuals Fn to form a flame individual set, in the initial state, ζ = N, record the moth individuals Fn in the flame individual set as flame individuals Ri, i = 1, 2, 3…ζ;

[0088] S4: Select a moth individual Fn from the population set in order, and record its order position in the population as j, judge whether "j>ζ" is true, if "j>ζ" is not true, directly select the flame individual Ri located at the jth position from the flame individual set, and record it as the target flame individual Fn-tar corresponding to the selected moth individual Fn, otherwise, if "j>ζ" is true, then select a flame individual Ri from the flame individual set as the target flame individual Fn-tar corresponding to the selected moth individual Fn according to the fitness δn of all moth individuals Fn using the roulette selection algorithm, and update the selected moth individual Fn based on the target flame individual Fn-tar through the following formula:

[0089] ;

[0090] Where dis(Fn, Fn-tar) is the distance between the selected moth individual Fn and the target flame individual Fn-tar; c is the logarithmic spiral shape constant; t is a random number in the interval [-1, 1];

[0091] S5: Update the number of flame individuals Ri in the flame individual set through the formula: ζ = round[(N-m)(N-1) / M];

[0092] S6: Judge whether "m

[0093] S7: Calculate the fitness δn corresponding to each moth individual Fn, and arrange all moth individuals Fn in descending order of the corresponding fitness δn, form a population set with all arranged moth individuals Fn, and arrange all moth individuals Fn in the population set and all moth individuals Fn in the flame individual set in descending order of the corresponding fitness δn, select the first ζ moth individuals Fn to form a flame individual set, and return to S4;

[0094] S8: Select the moth individual Fn with the maximum fitness δn from all moth individuals Fn in the population set and all moth individuals Fn in the flame individual set as the output of the simulated switching tube duty cycle.

[0095] Preferably, the N firefly individuals Fn are established, specifically including the following steps:

[0096] S1.1: For each firefly individual Fn, the assignment operation is performed by the following formula:

[0097] ;

[0098] The initialization of each firefly individual Fn by the Fuch chaotic mapping (the above formula) has the advantages of insensitivity to initial values, strong ergodicity, and fast convergence, so that the initial firefly individuals are evenly distributed in the search space, laying a foundation for the wide-area search of the initial iteration of the population;

[0099] S1.2: Repeat step S1.1 N times to establish N firefly individuals Fn.

[0100] Preferably, the fitness δn corresponding to the firefly individual Fn is calculated, specifically including the following steps: on the basis of the actually used photovoltaic system, a corresponding photovoltaic system simulation model is established by using MATLAB / SIMULINK software, then the environmental conditions corresponding to each third training sample are set, the environmental conditions include light and temperature, etc., the duty cycle of the simulated switching tube corresponding to the firefly individual Fn is set as a parameter to set the photovoltaic system simulation model, the photovoltaic power generation simulation is performed through the photovoltaic system simulation model, and the output power is taken as the fitness δn; and the fitness corresponding to the simulated switching tube duty cycle is the maximum output power. It needs to be noted that the simulation of the simulated switching tube duty cycle is in the training stage of the maximum power output model, not in the actual application stage, so there is sufficient freedom in the setting of the environmental conditions.

[0101] The application initializes these individuals by the Fuch chaotic mapping, ensures the extensive distribution of the initial population in the search space, and thus provides a good foundation for global search. The fitness δn of each firefly individual is calculated by the photovoltaic system simulation model constructed by MATLAB / SIMULINK software. This simulation model simulates the photovoltaic power generation process according to the changes of environmental conditions such as light and temperature, and takes the output power as the measurement standard of the fitness. Meanwhile, the position of the firefly individual, i.e. the value of the duty cycle, is constantly optimized through the iteration process. In each iteration, the position of the firefly individual is updated according to the distance between the firefly individual and the flame individual and other parameters. This updating process not only considers the current fitness, but also adjusts the size of the flame individual set adaptively, so as to ensure that the algorithm can effectively explore the solution space and accurately utilize the current optimal solution. The switching tube duty cycle maximizing the output power can be effectively found, so as to optimize the performance of the photovoltaic system. In summary, the application uses the simulation and iteration methods, not only improves the power generation efficiency of the photovoltaic system, but also improves the application flexibility and stability of the algorithm, which has important significance for improving the overall performance and economy of the photovoltaic power generation technology.

[0102] Embodiment 2

[0103] A photovoltaic power generation control system, see Figure 1 , comprising:

[0104] An actual output power monitoring module is configured to continuously monitor the actual output power of the photovoltaic system and determine whether the actual output power meets a photovoltaic control restart condition.

[0105] A to-be-detected photovoltaic panel image acquisition module is configured to acquire to-be-detected photovoltaic panel images of each photovoltaic panel.

[0106] An environmental feature vector establishing vector is configured to establish an environmental feature vector corresponding to each photovoltaic panel for each to-be-detected photovoltaic panel image.

[0107] An environmental feature map establishing vector is configured to splice all the environmental feature vectors to obtain an environmental feature map, wherein each coordinate point in the environmental feature map corresponds to a photovoltaic panel, and each coordinate point stores an environmental feature vector of the corresponding photovoltaic panel.

[0108] An analog switch tube duty cycle output module is configured to input the environmental feature map into a trained maximum power output model for processing to output an analog switch tube duty cycle.

[0109] A photovoltaic power generation delay control strategy establishing module is configured to output a photovoltaic power generation delay control strategy based on the analog switch tube duty cycle.

[0110] A maximum output power updating module is configured to acquire a maximum output power based on the analog switch tube duty cycle and update the maximum output power based on the analog maximum output power.

[0111] It should be understood that those skilled in the art can make improvements or changes according to the above description, and all these improvements and changes shall belong to the protection scope of the appended claims of the present application. The parts not described in detail in the specification belong to the prior art known to those skilled in the art.

Claims

1. A photovoltaic power generation control method characterized by, The method comprises the following steps: Continuously monitoring the actual output power q of the photovoltaic system, and determining whether the photovoltaic control restart condition is met, wherein the photovoltaic control restart condition is "(η-q) / q≥ε", wherein η is the maximum output power recorded at present and is updated through the photovoltaic control restart operation subsequently, and ε is a power deviation threshold; if the photovoltaic control restart condition is met, the photovoltaic control restart operation is performed; if the photovoltaic control restart condition is not met, the output power q of the photovoltaic system is continuously monitored; The photovoltaic control restart operation comprises the following steps: Obtaining a to-be-detected photovoltaic panel image of each photovoltaic panel, wherein the photovoltaic panel image is an RGB image; For each to-be-detected photovoltaic panel image, the to-be-detected photovoltaic panel image is processed through the trained image repairing model to obtain a non-stain photovoltaic panel image, the non-stain photovoltaic panel image is subjected to illumination condition analysis to output an illumination intensity distribution matrix, the illumination intensity distribution matrix is processed through a multilayer perceptron to obtain an illumination distribution feature vector, the non-stain photovoltaic panel image is subjected to spectral analysis to output a spectral feature vector, and the illumination distribution feature vector and the spectral feature vector are spliced with an environmental temperature value to obtain an environmental feature vector; All the environmental feature vectors are spliced to obtain an environmental feature map, each coordinate point in the environmental feature map corresponds to a photovoltaic panel, and each coordinate point stores the environmental feature vector of the corresponding photovoltaic panel; The environmental feature map is input into the trained maximum power output model for processing to output a simulated switch tube duty cycle, the simulated switch tube duty cycle refers to the time ratio of the conduction of the switch tube in the photovoltaic inverter, the output voltage and the output current of the photovoltaic system can be adjusted to further adjust the output power, and the maximum output power η is obtained based on the simulated switch tube duty cycle; the photovoltaic power generation delay control strategy is output based on the simulated switch tube duty cycle, and the maximum output power η is updated based on the simulated maximum output power; The photovoltaic system is controlled based on the photovoltaic power generation delay control strategy.

2. The photovoltaic power generation control method according to claim 1, characterized by, The image repairing model is established based on a generated U-net network; The training of the image repairing model comprises the following steps: A plurality of photovoltaic panel images are obtained, the photovoltaic panel images are images actually photographed by a user, and for each photovoltaic panel image, the photovoltaic panel image and a standard photovoltaic panel image corresponding to the photovoltaic panel image are used to form a first training sample; All the first training samples are used to form a first training sample set, and the first training sample set is input into the image repairing model with initialized parameters for training, the standard photovoltaic panel image is used as a target for weight parameter adjustment during the training, and it is determined whether a first training condition is met; if the first training condition is met, the trained image repairing model is output; otherwise, the first training sample set is continuously used for iterative training.

3. The photovoltaic power generation control method according to claim 2, wherein The non-stain photovoltaic panel image is subjected to illumination condition analysis to output an illumination intensity distribution matrix, and the illumination intensity distribution matrix is output through the trained illumination distribution detection model. ​ The illumination distribution detection model is established based on a U-net network, and includes a U-net network layer, a self-attention mechanism layer and an illumination intensity distribution matrix output layer, wherein the U-net network layer is used for extracting illumination distribution features of the non-stain photovoltaic panel image; the self-attention mechanism layer is used for strengthening the illumination distribution features; and the illumination intensity distribution matrix output layer is used for outputting the illumination intensity distribution matrix based on the strengthened illumination distribution features. The non-stain photovoltaic panel image is sent into the trained illumination distribution detection model for processing, and the processing specifically includes the following steps:

4. The photovoltaic power generation control method according to claim 3, wherein The non-stain photovoltaic panel image is sent into the U-net network layer for processing to obtain illumination distribution features T; In the illumination intensity distribution matrix output layer, the strengthened illumination distribution features T are subjected to convolution operation to output the illumination intensity distribution matrix. In the self-attention mechanism layer, the illumination distribution feature T is multiplied with the key weight matrix W k to obtain a key-value matrix K; the illumination distribution feature T is multiplied with the value weight matrix W p to obtain a value matrix P; the non-spotted photovoltaic panel image is converted to HSV space through color conversion, and a feature map corresponding to the V channel in the HSV space is selected as a brightness feature map; the brightness feature map is multiplied with the query weight matrix W Q to obtain a query matrix Q, and an attention weight matrix ATT=softmax((Q·K T ) / (μ) 0.5 ) is calculated, where μ is the number of columns of the key-value matrix K; the attention weight matrix ATT is multiplied with the value matrix P to obtain a reinforced illumination distribution feature T; The training of the illumination distribution detection model includes the following steps:

5. The photovoltaic power generation control method according to claim 4, wherein All photovoltaic panel images in the first training sample set are obtained, and all photovoltaic panel images in the first training sample set are processed by the trained image restoration model to obtain a plurality of non-stain photovoltaic panel images. Each non-stain photovoltaic panel image and its corresponding standard illumination intensity distribution matrix form a second training sample, all second training samples form a second training sample set, and the second training sample set is sent into the parameter initialized illumination distribution detection model for training. During the training, the standard illumination intensity distribution matrix is taken as the target for weight parameter adjustment, and it is judged whether the second training condition is met. If the second training condition is met, the trained illumination distribution detection model is output; otherwise, the iterative training is continued by using the second training sample set. The maximum power output model is established based on a convolutional neural network; 6. The photovoltaic power generation control method according to claim 5, wherein The training of the maximum power output model includes the following steps: All photovoltaic panel images obtained in a plurality of time periods are obtained, and the environmental feature map corresponding to all photovoltaic panel images in any time period is taken as a third training sample, and the simulated switching tube duty cycle output by the improved firefly optimization algorithm is used to mark the third training sample. All marked third training samples form a third training sample set, and the third training sample set is sent into the parameter initialized maximum power output model for training. During the training, the simulated switching tube duty cycle corresponding to the third training sample is taken as the target to calculate the loss value. If the loss value is within a preset range, the preset range is artificially set to constrain the accuracy of the maximum power output model, and the trained maximum power output model is output. Otherwise, the iterative training is continued by using the third training sample set. The simulated switching tube duty cycle is output by the improved firefly optimization algorithm, and the method specifically includes the following steps:

7. The photovoltaic power generation control method according to claim 6, wherein S2: set the maximum iteration number M, let m=1, and h is used to record the iteration number; S1: Establish N number of flying moth individuals F n , n = 1, 2, 3…N, and F n The storage in F is the duty cycle of the switch tube to be simulated. S6: judge whether "m S3: Calculate the moth individual F n The corresponding fitness δ n And all the moth individuals F n According to the corresponding fitness δ n Arrange from large to small, arrange all the moth individuals F n Form a population set, and select the first ζ moth individuals F n Form a flame individual set, initially ζ=N, and the moth individuals F n In the flame individual set are recorded as flame individuals R i , i=1, 2, 3…ζ; S4: Select a moth individual F from the population set in order n and record its order position as j in the population, determine whether "j > ζ" is true, if not, directly select the flame individual R at the jth position from the flame individual set i and record it as the selected moth individual F n corresponding to the target flame individual F n-tar , on the contrary, if "j > ζ" is true, then select a flame individual R from the flame individual set according to all moth individuals F n corresponding to the fitness δ n using the roulette wheel selection algorithm i as the selected moth individual F n corresponding to the target flame individual F n-tar based on the target flame individual F n-tar update the selected moth individual F n through the following formula: ； where dis(F n , F n-tar ) is the distance between the selected moth individual F n and the target flame individual F n-tar ; c is a logarithmic spiral shape constant; and t is a random number between the interval [-1, 1]. S5: Update the number ζ by the formula: ζ = round[(N - m)(N - 1) / M] for the flame individual R in the flame individual set i number ζ; The system applies the photovoltaic power generation control method in any one of the above claims 1-9, and includes: S7: Calculate the moth individual F n According to the corresponding fitness δ n And all the moth individuals F n According to the corresponding fitness δ n Arrange from large to small, arrange all the moth individuals F n Recompose the population set, and all the moth individuals F n And all the moth individuals F n According to the corresponding fitness δ n Arrange from large to small, select the first ζ moth individuals F n Composed of flame individual set, back to S4; S8: select all moth individuals F in the population set n and all moth individuals F in the flame individual set n fitness delta n the largest moth individual F n output as analog switch tube duty cycle.

8. The photovoltaic power generation control method according to claim 7, wherein establishing N firefly individuals F n comprising the following steps: S1.1: For each individual moth F n The assignment operation is performed by the following equation: ； S1.2: Repeat step S1.1 N times to establish N firefly individuals F n .

9. The photovoltaic power generation control method according to claim 8, wherein The moth individual F n The corresponding fitness δ n , specifically comprising the following steps: on the basis of the actually adopted photovoltaic system, a corresponding photovoltaic system simulation model is established, and the environment condition corresponding to each third training sample is set, and the moth individual F n The corresponding analog switch tube duty cycle is set as a parameter for the photovoltaic system simulation model, the photovoltaic power generation simulation is carried out through the photovoltaic system simulation model, and the output power is output as the fitness δ n ; and the fitness corresponding to the analog switch tube duty cycle is the maximum output power.

10. A photovoltaic power generation control system characterized by comprising: An actual output power monitoring module is configured to continuously monitor the actual output power of the photovoltaic system and judge whether the actual output power meets the photovoltaic control restart condition. ​ A to-be-detected photovoltaic panel image acquisition module is configured to acquire a to-be-detected photovoltaic panel image of each photovoltaic panel; An environmental feature vector establishment vector is configured to establish an environmental feature vector corresponding to a photovoltaic panel for each to-be-detected photovoltaic panel image; An environmental feature map establishment vector is configured to splice all the environmental feature vectors to obtain an environmental feature map, wherein each coordinate point in the environmental feature map corresponds to a photovoltaic panel, and each coordinate point stores an environmental feature vector of the corresponding photovoltaic panel; An analog switch tube duty cycle output module is configured to input the environmental feature map into a trained maximum power output model for processing and output an analog switch tube duty cycle; A photovoltaic power generation delay control strategy establishment module is configured to output a photovoltaic power generation delay control strategy based on the analog switch tube duty cycle; A maximum output power update module is configured to acquire a maximum output power based on the analog switch tube duty cycle and update the maximum output power based on the analog maximum output power.

Images