A real-time monitoring and prediction system for new energy photovoltaic power based on the Internet of Things

Abstract

The present invention relates to the technical field of photovoltaic monitoring, and in particular to a real-time monitoring and prediction system for new energy photovoltaic power based on the Internet of Things. The system includes a data acquisition module for collecting spectral incident data and electrical parameters in the photovoltaic group area; an energy density conversion module for performing energy density conversion on the spectral incident data to obtain optical segment energy density distribution data; an attenuation identification module for obtaining the effective spectral energy value of the optical segment based on the optical segment energy density distribution data and the installation angle of the photovoltaic panel and loading it into a photovoltaic attenuation identification model to obtain a spectral attenuation factor vector; a global fitting module for performing hierarchical recursive analysis according to the photovoltaic group topology structure to obtain a global contribution matrix and combining it with the spectral attenuation factor vector to obtain a global attenuation degree; and a power prediction module for performing power prediction according to the global attenuation degree and the target optical segment energy density distribution data to obtain the overall predicted power of the photovoltaic group, realizing more accurate real-time monitoring and prediction of photovoltaic power.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic monitoring, and particularly to a real-time monitoring and prediction system for new energy photovoltaic power based on the Internet of Things. Background Art

[0002] The nano-coating of photovoltaic panels will age due to ultraviolet irradiation during long-term operation, resulting in a continuous decline in the absorption rate of photovoltaic panels for specific wavelength bands, and further causing hidden power attenuation of the generated power. Traditional photovoltaic power prediction methods usually adopt current-voltage curve analysis, which is difficult to decouple the attenuation of spectral absorption rate, resulting in significant errors in the prediction of photovoltaic power. Some existing technical teams use the global irradiance and component temperature to construct a linear prediction model for photovoltaic power prediction, which cannot distinguish the differential attenuation of ultraviolet light, visible light, and infrared light, and the prediction of photovoltaic power is not accurate enough under the condition of drastic changes in spectral distribution. For example, the error is large during the dawn and dusk periods due to large spectral changes. Therefore, how to achieve more accurate real-time monitoring and prediction of photovoltaic power when the absorption rate of photovoltaic panels for specific wavelength bands decreases due to coating aging has become an urgent problem to be solved. Summary of the Invention

[0003] The present invention provides a real-time monitoring and prediction system for new energy photovoltaic power based on the Internet of Things to solve the problem of inaccurate prediction of photovoltaic power caused by differential attenuation of the absorption rate of different wavelength bands of photovoltaic panels due to coating aging.

[0004] To achieve the above object, the present invention provides a real-time monitoring and prediction system for new energy photovoltaic power based on the Internet of Things. The system includes: a data acquisition module, an energy density conversion module, an attenuation identification module, a global fitting module, and a power prediction module, and the modules are connected in sequence.

[0005] The data acquisition module is used to collect spectral incident data of the photovoltaic panel group area in real time through multi-spectral sensors arranged in a preset area of the photovoltaic panel group, and collect current parameters and voltage parameters of each photovoltaic panel in real time through current and voltage sensors to obtain the photovoltaic power of each photovoltaic panel.

[0006] The energy density conversion module is used to divide the spectral incident data according to the light band type to obtain a light band data set and perform energy density conversion to obtain light band energy density distribution data; the light band types include ultraviolet light band, visible light band, and near-infrared light band.

[0007] The attenuation identification module is used to calculate the effective spectral energy value of the light band according to the light band energy density distribution data and the installation angle of the photovoltaic panel, and load the effective spectral energy value of the light band and the power data of each photovoltaic panel into the photovoltaic attenuation identification model to obtain the spectral attenuation factor vector of each photovoltaic panel.

[0008] The global fitting module is used to perform hierarchical recursive analysis based on the photovoltaic group topology structure to obtain the global contribution degree matrix, and calculate the global attenuation degree according to the global contribution degree matrix and the spectral attenuation factor vector of each photovoltaic panel.

[0009] The power prediction module is used to obtain the predicted data of weather indicators at the target prediction time point and determine the similarity of historical weather indicators to obtain the reference time point, use the light segment energy density distribution data corresponding to the reference time point as the target light segment energy density distribution data, and perform power prediction according to the global attenuation degree and the target light segment energy density distribution data to obtain the overall predicted power of the photovoltaic group.

[0010] Furthermore, the energy density conversion module further includes a data preprocessing module, an energy density calculation module, a discrete integration module, and a data integration module that are connected in sequence.

[0011] The data preprocessing module is used to divide the spectral incident data into irradiance data of discrete wavelength points contained in each light segment according to the light segment type to obtain the light segment data set.

[0012] The energy density calculation module is used to calculate the corresponding energy density value according to the irradiance data of discrete wavelength points as:

[0013] .

[0014] where is the spectral irradiance of the discrete wavelength point is the Planck constant, is the speed of light, is the wavelength of the discrete wavelength point is the discrete wavelength point is the wavelength, is the discrete wavelength point is the energy density value.

[0015] The discrete integration module is used to perform discrete integration on the energy density values of discrete wavelength points in the ultraviolet light segment, visible light segment, and near-infrared light segment to obtain the light segment energy density value as:

[0016] .

[0017] where is the spectral resolution of the multispectral sensor, is the light segment energy density value.

[0018] The data integration module is used to summarize the light segment energy density values of the ultraviolet light segment, visible light segment, and near-infrared light segment to obtain the light segment energy density distribution data.

[0019] Further, the attenuation recognition module further includes an angle conversion module, an energy analysis module, and a generative adversarial network module that are connected in sequence.

[0020] The angle conversion module is configured to obtain solar position data from a regional weather station through an API interface, and obtain the light incident angle based on the solar position data and the installation angle of the photovoltaic panel as:

[0021] .

[0022] Where is the solar altitude angle, is the solar azimuth angle, is the installation inclination of the photovoltaic panel, is the azimuth angle of the photovoltaic panel, is the light incident angle.

[0023] The energy analysis module is configured to obtain the effective spectral energy value of the optical segment based on the optical segment energy density value of each optical segment in the optical segment energy density distribution data and the light incident angle as:

[0024] .

[0025] Where is the area of a single photovoltaic panel, is the effective spectral energy value of the optical segment.

[0026] The generative adversarial network module is configured to load the effective spectral energy value of the optical segment and the power data of each photovoltaic panel into a photovoltaic attenuation recognition model established by the generative adversarial network algorithm to obtain the spectral attenuation factor vector of each photovoltaic panel.

[0027] Further, the generative adversarial network module further includes an input unit, a spectral segment response unit, a power reconstruction unit, and a discriminant unit that are connected in sequence.

[0028] The input unit is configured to generate a fusion feature through a fully connected layer by taking the effective spectral energy value of the optical segment and the power data of each photovoltaic panel as inputs as:

[0029] .

[0030] Where is the power of the photovoltaic panel, and are the weight matrix and bias term of the fully connected layer, is the fusion feature.

[0031] The spectral response unit is used to set up independent sub-networks for the ultraviolet light band, visible light band, and near-infrared light band respectively to perform spectral response on the fusion features, and obtain the spectral attenuation factor vectors of each light band through LeakyReLU activation. That is:

[0032] .

[0033] Among them is the spectral attenuation factor vector, and are the weight matrix and bias term of each light band sub-network, and LeakyReLU is the activation function.

[0034] The power reconstruction unit is used to reconstruct the power of the photovoltaic panel through a fully connected regression network according to the spectral attenuation factor vector of the photovoltaic panel to obtain the reconstructed power of the photovoltaic panel. The expression is:

[0035] .

[0036] and are the weight matrix and bias term of the fully connected regression network, is the reconstructed power of the photovoltaic panel.

[0037] The discrimination unit is used to input the absolute error between the reconstructed power of the photovoltaic panel and the photovoltaic panel power data into a multi-layer perceptron to obtain the attenuation discrimination probability. When the attenuation discrimination probability is less than the preset discrimination threshold, it is determined that the spectral attenuation factor vector of the photovoltaic panel is valid and stored. Otherwise, parameter update is performed through backpropagation of the photovoltaic attenuation generator loss function to regenerate the spectral attenuation factor vector.

[0038] The multi-layer perceptron uses the Sigmoid function, and the expression is:

[0039] ;

[0040] Among them is the power of the photovoltaic panel, , are the weight matrix and bias term of the photovoltaic attenuation discriminator, is the attenuation discrimination probability;

[0041] Furthermore, the training steps of the photovoltaic attenuation recognition model are as follows:

[0042] Use the effective spectral energy values of each light band and the corresponding photovoltaic panel power data within the historical operation period as training data.

[0043] Initialize the parameters of the PV attenuation generator and the PV attenuation discriminator by the He initialization method, set the initial learning rates of the PV attenuation generator and the PV attenuation discriminator, and update the learning rates through the cosine annealing learning rate scheduling strategy.

[0044] Iteratively train the PV attenuation generator and the PV attenuation discriminator through the alternating training strategy. Each training iteration includes:

[0045] Fix the parameters of the PV attenuation generator, continuously train the PV attenuation discriminator, update the discriminator parameters according to the loss function of the PV attenuation discriminator, and update the parameters with the discriminator learning rate through the optimizer.

[0046] 。

[0047] where is the attenuation discrimination probability, is the value of the loss function of the PV attenuation discriminator.

[0048] Fix the parameters of the PV attenuation discriminator, train the PV attenuation generator, update the generator parameters according to the loss function of the PV attenuation generator, and update the parameters with the generator learning rate through the Adam optimizer. The loss function of the PV attenuation generator is:

[0049] 。

[0050] where is the reconstructed power of the PV panel, is the reconstruction error weight coefficient, is the power of the PV panel, is the value of the loss function of the PV attenuation generator.

[0051] The reconstruction error weight coefficient is used to balance the adversarial loss and the power accuracy. The initial value of the reconstruction error weight coefficient is set to 0.5 and is gradually increased to 1 through linear update. The linear update is:

[0052] 。

[0053] where is the reconstruction error weight coefficient at the th iteration, is the initial reconstruction error weight coefficient, is the final reconstruction error weight coefficient, is the number of iteration rounds, is the maximum number of iterations.

[0054] Terminate the iteration when the training reaches the preset maximum number of iterations.

[0055] Furthermore, the global fitting module further includes a hierarchical parsing module, a contribution degree analysis module, and an attenuation degree analysis module that are connected in sequence.

[0056] The hierarchical parsing module is configured to determine the series level and parallel level of each photovoltaic panel according to the photovoltaic array topology, and obtain the voltage contribution coefficient of the photovoltaic panel based on the series level of the photovoltaic panel. It is:

[0057] 。

[0058] Where is the number of photovoltaic panels in the series level where the photovoltaic panel is located, is the photovoltaic panel 's voltage contribution coefficient.

[0059] Obtain the current contribution coefficient of the photovoltaic panel according to the parallel level of the photovoltaic panel It is:

[0060] 。

[0061] Where is the number of photovoltaic panels in the parallel level where the photovoltaic panel is located, is the photovoltaic panel 's current contribution coefficient.

[0062] The contribution degree analysis module is configured to obtain the contribution degree of the photovoltaic panel according to the voltage contribution coefficient and current contribution coefficient of the photovoltaic panel It is:

[0063] 。

[0064] Where is the contribution degree of the photovoltaic panel 。

[0065] The attenuation degree analysis module is configured to aggregate the contribution degrees of all photovoltaic panels to obtain a global contribution degree matrix, and sequentially retrieve the contribution degree of each photovoltaic panel in the global contribution degree matrix and multiply it by its corresponding spectral attenuation factor vector and then sum to obtain the global attenuation degree It is:

[0066] 。

[0067] Where is the total number of photovoltaic panels in the photovoltaic array, is the spectral attenuation factor corresponding to the th light segment in the spectral attenuation factor vector, is the th light segment's global attenuation degree.

[0068] Further, the power prediction module further includes a similarity determination module, a first prediction unit, a second prediction unit, and a third prediction unit that are connected in sequence.

[0069] The similarity determination module is configured to obtain weather index prediction data for a target prediction date, calculate a difference coefficient between the weather index prediction data for the target prediction date and historical weather index data, select the date with the smallest difference coefficient as a reference date, and retrieve the light segment energy density distribution data corresponding to the reference date as the target light segment energy density distribution data.

[0070] The first prediction unit is configured to obtain solar position prediction data for a target time point from a regional weather station, and obtain a predicted value of the light incident angle based on the solar position prediction data and the photovoltaic panel installation angle.

[0071] The obtaining the predicted value of the light incident angle based on the solar position prediction data and the photovoltaic panel installation angle is:

[0072] .

[0073] Where is the predicted value of the solar altitude angle, is the predicted value of the solar azimuth angle, is the installation inclination angle of the photovoltaic panel, is the azimuth angle of the photovoltaic panel, is the predicted value of the light incident angle.

[0074] The second prediction unit is configured to obtain a total predicted value of the effective spectral energy based on the light segment energy density values of each light segment in the target light segment energy density distribution data and the predicted value of the light incident angle is:

[0075] .

[0076] Where is the target light segment energy density of the th light segment, is the total predicted value of the effective spectral energy of the th light segment, is the area of a single photovoltaic panel, is the total number of photovoltaic panels in the photovoltaic group.

[0077] The third prediction unit is configured to calculate the overall predicted power of the photovoltaic group based on the global attenuation degree and the total predicted value of the effective spectral energy is:

[0078] .

[0079] Where is the The energy conversion coefficient of the segment of light is the overall predicted power of the photovoltaic group.

[0080] Furthermore, the weather indicators include: solar irradiance, cloud cover, temperature, visibility; the similarity determination module further includes an index normalization unit and a difference coefficient calculation unit connected in sequence.

[0081] The index normalization unit is used to normalize the predicted weather index data of each target prediction date and the historical weather index of the corresponding category by the maximum-minimum normalization method to obtain the normalized weather index data.

[0082] The difference coefficient calculation unit is used to calculate the Euclidean distance of the normalized weather index data according to the Euclidean distance calculation method based on the normalized weather index data to obtain the difference coefficient.

[0083] Compared with the prior art, the beneficial effects of the present invention are: by performing discrete integration on the energy density of discrete wavelength points in the ultraviolet light segment, visible light segment, and near-infrared light segment to obtain the light segment energy density distribution data, realizing the adversarial accurate extraction of the spectral attenuation factor vector through the adversarial network spectral response for each light segment, and further combining the hierarchical recursive analysis of the photovoltaic group topology structure to obtain the global attenuation degree, so as to more accurately monitor and predict the photovoltaic power as a whole under the condition of different absorption rate attenuations of the photovoltaic panel for different light segments. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present application. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0085] Figure 1 It is a schematic diagram of the module composition of a new energy photovoltaic power real-time monitoring and prediction system based on the Internet of Things according to Embodiment 1 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0086] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0087] Embodiment 1: As Figure 1As shown in the figure, this embodiment provides a real-time monitoring and prediction system for new energy photovoltaic power based on the Internet of Things. The system includes: a data acquisition module, an energy density conversion module, an attenuation identification module, a global fitting module, and a power prediction module, and the modules are connected in sequence.

[0088] The data acquisition module is used to collect the spectral incident data of the photovoltaic group area in real time through multi-spectral sensors arranged in a preset area of the photovoltaic panel group, and collect the current parameters and voltage parameters of each photovoltaic panel in real time through current and voltage sensors to obtain the photovoltaic power of each photovoltaic panel.

[0089] For example: A photovoltaic assembly in a certain area has 50 photovoltaic panels, each with an area of 1.92 m². Each row consists of 10 photovoltaic panels connected in series to form a string, and a total of 5 parallel strings together form a 10×5 matrix structure photovoltaic group. The Ocean Optics STS-UV multi-spectral sensor with a wavelength range of 280 - 1100 nm and a spectral resolution of 2 nm is arranged at the center of the photovoltaic array to collect spectral incident data, and the ABB-VSN800 current and voltage sensors are installed on the photovoltaic panels to collect current parameters and voltage parameters. Taking the No. 1 photovoltaic panel as an example, at a certain time point, the current parameter of the photovoltaic panel is 8.73 A, and the voltage parameter is 38.21 V. Multiplying the current parameter and the voltage parameter can obtain the photovoltaic power of this photovoltaic panel as 333.57 W.

[0090] The energy density conversion module is used to divide the spectral incident data according to the light segment type to obtain a light segment data set and perform energy density conversion to obtain light segment energy density distribution data; the light segment types include ultraviolet light segment, visible light segment, and near-infrared light segment; the ultraviolet light segment is the light segment with a wavelength in the nanometer range, the visible light segment is the light segment with a wavelength in the nanometer range, and the near-infrared light segment is the light segment with a wavelength in the nanometer range.

[0091] The attenuation identification module is used to calculate the effective spectral energy value of the light segment according to the light segment energy density distribution data and the installation angle of the photovoltaic panel, and load the effective spectral energy value of the light segment and the power data of each photovoltaic panel into the photovoltaic attenuation identification model to obtain the spectral attenuation factor vector of each photovoltaic panel.

[0092] The global fitting module is used to perform hierarchical recursive analysis according to the photovoltaic group topology structure to obtain a global contribution degree matrix, and calculate the global attenuation degree according to the global contribution degree matrix and the spectral attenuation factor vector of each photovoltaic panel.

[0093] The power prediction module is used to obtain the predicted weather index data at the target prediction time point, determine the reference time point through the similarity judgment of historical weather indexes, use the light segment energy density distribution data corresponding to the reference time point as the target light segment energy density distribution data, and perform power prediction according to the global attenuation degree and the target light segment energy density distribution data to obtain the overall predicted power of the photovoltaic group.

[0094] It should be noted that the energy density conversion module further includes a data preprocessing module, an energy density calculation module, a discrete integration module, and a data integration module that are connected in sequence.

[0095] The data preprocessing module is used to divide the spectral incident data according to the light segment type for each discrete wavelength point to obtain the irradiance data of each discrete wavelength point contained in each light segment as the light segment data set.

[0096] For example: divide the 340nm wavelength point and its corresponding irradiance into the ultraviolet light segment data set, and divide the 460nm wavelength point and its corresponding irradiance into the visible light segment data set.

[0097] The energy density calculation module is used to calculate the corresponding energy density value according to the irradiance data of the discrete wavelength point as:

[0098] .

[0099] Where is the discrete wavelength point is the spectral irradiance of is the Planck constant, is the speed of light, is the discrete wavelength point is the wavelength of is the discrete wavelength point is the energy density value of

[0100] The value of the Planck constant is and the value of the speed of light is .

[0101] For example: taking the 340nm wavelength point as an example, the measured irradiance , and the energy density of this wavelength point can be obtained .

[0102] The discrete integration module is used to perform discrete integration on the energy density values of discrete wavelength points in the ultraviolet light segment, visible light segment, and near-infrared light segment to obtain the light segment energy density value as:

[0103] .

[0104] Where is the spectral resolution of the multispectral sensor, is the energy density value of the optical band.

[0105] A data integration module for summarizing the energy density values of the ultraviolet, visible, and near-infrared optical bands to obtain optical band energy density distribution data.

[0106] For example: in the ultraviolet optical band, the discrete integration of the energy density values at discrete wavelength points within the wavelength range is performed. When the spectral resolution of the multispectral sensor is 2 nm, then for all wavelength points with a resolution of 2 nm within the wavelength range, such as wavelength points 280, 282, 284, etc. up to 398 wavelength points, the discrete integration of the energy density values is performed to obtain the ultraviolet optical band energy density value as , and similarly, the visible optical band energy density is 435.20 J / m², and the near-infrared optical band energy density is 268.10 J / m². Then the optical band energy density distribution data is: [83.70, 435.20, 268.10] J / m².

[0107] It should be noted that the attenuation identification module further includes an angle conversion module, an energy analysis module, and a generative adversarial network module connected in sequence.

[0108] The angle conversion module is used to obtain solar position data from the regional weather station through the API interface, and obtain the light incident angle according to the solar position data and the installation angle of the photovoltaic panel as:

[0109] .

[0110] where is the solar altitude angle, is the solar azimuth angle, is the installation inclination angle of the photovoltaic panel, is the azimuth angle of the photovoltaic panel, is the light incident angle.

[0111] The light incident angle is the angle between the light and the normal of the photovoltaic panel surface, which determines how much sunlight can be effectively received by the photovoltaic panel. For example: when the solar position data at a certain time point is obtained through the API interface of the regional weather station as: the solar altitude angle is α = 68.3°, the solar azimuth angle is γ = 135.7°, the installation inclination angle of the photovoltaic panel is β = 30°, and the azimuth angle is δ = 180° (i.e., due south), then the light incident angle .

[0112] The energy analysis module is used to obtain the effective spectral energy value of the optical band according to the energy density value of each optical band in the optical band energy density distribution data and the light incident angle is:

[0113] .

[0114] Wherein is the area of a single photovoltaic panel, is the effective spectral energy value of the optical band.

[0115] Combined with the light incident angle and the optical band energy density, the effective spectral energy value of the ultraviolet light band , similarly, the effective spectral energy value of the visible light band can be obtained as 777.09 J, and the effective spectral energy value of the near-infrared light band is 478.72 J.

[0116] The generative adversarial network module is used to load the effective spectral energy value of the optical band and the power data of each photovoltaic panel into the photovoltaic attenuation recognition model established by the generative adversarial network algorithm to obtain the spectral attenuation factor vector of each photovoltaic panel.

[0117] It should be noted that the generative adversarial network module further includes an input unit, a spectral band response unit, a power reconstruction unit, and a discrimination unit that are connected in sequence.

[0118] The input unit is used to generate a fusion feature through a fully connected layer with the effective spectral energy value of the optical band and the power data of each photovoltaic panel as inputs is:

[0119] .

[0120] Wherein is the power of the photovoltaic panel, and are the weight matrix and bias term of the fully connected layer, is the fusion feature.

[0121] The spectral band response unit is used to set up independent sub-networks for the ultraviolet light band, visible light band, and near-infrared light band respectively to perform spectral band response on the fusion feature and obtain the spectral attenuation factor vector of each optical band through LeakyReLU activation is:

[0122] .

[0123] Wherein is the spectral attenuation factor vector, and are the weight matrix and bias term of each optical band sub-network, and LeakyReLU is the activation function.

[0124] For example, for the first photovoltaic panel, the system inputs the effective spectral energy values [149.45, 777.09, 478.72] of three optical bands and the measured power of 333.57 into a pre-trained photovoltaic attenuation recognition model. This model first integrates the input data into a comprehensive feature value through a feature fusion layer, and then calculates the attenuation factors of each optical band as 0.35 for the ultraviolet band, 0.12 for the visible band, and 0.08 for the near-infrared band through three neural networks dedicated to processing different optical bands. These three values form the spectral attenuation factor vector [0.35, 0.12, 0.08] of this photovoltaic panel, indicating the attenuation degree of each optical band.

[0125] A power reconstruction unit, which is used to reconstruct the power of the photovoltaic panel through a fully connected regression network according to the spectral attenuation factor vector of the photovoltaic panel to obtain the reconstructed power of the photovoltaic panel.

[0126] The expression is:

[0127] .

[0128] and are the weight matrix and bias term of the fully connected regression network, is the reconstructed power of the photovoltaic panel.

[0129] A discrimination unit, which is used to input the absolute error between the reconstructed power of the photovoltaic panel and the power data of the photovoltaic panel into a multi-layer perceptron to obtain an attenuation discrimination probability. When the attenuation discrimination probability is less than a preset discrimination threshold, it is determined that the spectral attenuation factor vector of this photovoltaic panel is valid and stored. Otherwise, parameter update is performed through backpropagation of the photovoltaic attenuation generator loss function to regenerate the spectral attenuation factor vector.

[0130] The multi-layer perceptron uses the Sigmoid function, and the expression is:

[0131] ;

[0132] where is the power of the photovoltaic panel, , are the weight matrix and bias term of the photovoltaic attenuation discriminator, is the attenuation discrimination probability;

[0133] The weight matrix and bias term parameters in each of the above expressions are automatically iteratively optimized during the operation of the model.

[0134] For example: To verify the effectiveness of the spectral attenuation factor vector [0.35, 0.12, 0.08], the discriminator reconstructs the photovoltaic panel power through this attenuation factor vector, and the result is 114.34. The discrimination probability obtained by the discriminator is 0.71, which is higher than the system preset threshold of 0.15. Therefore, parameter update iteration is performed to regenerate the spectral attenuation factor vector. After multiple iterations, a new spectral attenuation factor vector [0.23, 0.08, 0.06] is obtained, and the corresponding discrimination probability drops to 0.13, which is less than the preset discrimination threshold of 0.15, indicating accurate reconstruction. Therefore, it is determined that the spectral attenuation factor vector [0.23, 0.08, 0.06] is effective and saved. In this model, the attenuation discrimination probability represents the degree of reconstruction error, and the smaller the value, the more accurate and reliable the attenuation factor is.

[0135] It should be noted that the training steps of the photovoltaic attenuation recognition model are as follows:

[0136] Use the effective spectral energy values in the optical band and the corresponding photovoltaic panel power data within the historical operation period as training data.

[0137] Initialize the parameters of the photovoltaic attenuation generator and the photovoltaic attenuation discriminator through the He initialization method, set the initial learning rates of the photovoltaic attenuation generator and the photovoltaic attenuation discriminator, and update the learning rates through the cosine annealing learning rate scheduling strategy.

[0138] The parameters of the photovoltaic attenuation generator and the photovoltaic attenuation discriminator include the weight matrices and bias terms of each layer in the photovoltaic attenuation generator and the photovoltaic attenuation discriminator. Set the initial learning rate of the photovoltaic attenuation generator to , and set the initial learning rate of the photovoltaic attenuation discriminator to .

[0139] Perform iterative training on the photovoltaic attenuation generator and the photovoltaic attenuation discriminator through the alternating training strategy. Each training iteration includes:

[0140] Fix the parameters of the photovoltaic attenuation generator, continuously train the photovoltaic attenuation discriminator, update the discriminator parameters according to the loss function of the photovoltaic attenuation discriminator, and perform parameter update through the optimizer with the discriminator learning rate.

[0141] .

[0142] Where is the attenuation discrimination probability, is the value of the loss function of the photovoltaic attenuation discriminator.

[0143] Fix the parameters of the photovoltaic attenuation discriminator, train the photovoltaic attenuation generator, update the generator parameters according to the loss function of the photovoltaic attenuation generator, and perform parameter update through the Adam optimizer with the generator learning rate. The loss function of the photovoltaic attenuation generator is:

[0144] 。

[0145] Among them is the reconstructed power of the photovoltaic panel, is the reconstruction error weight coefficient, is the power of the photovoltaic panel, is the loss function value of the photovoltaic attenuation generator.

[0146] The reconstruction error weight coefficient is used to balance the adversarial loss and power accuracy. The initial value of the reconstruction error weight coefficient is set to 0.5 and is gradually increased to 1 through linear update. The linear update is as follows:

[0147] 。

[0148] Among them is the reconstruction error weight coefficient at the th iteration, is the initial reconstruction error weight coefficient, is the final reconstruction error weight coefficient, is the iteration round, is the maximum number of iterations.

[0149] For example: when the maximum number of iterations is set to 200, at the 50th iteration, the reconstruction error weight coefficient is 。

[0150] When the training reaches the preset maximum number of iterations, the iteration is terminated.

[0151] The training of the photovoltaic attenuation identification model is implemented using the PyTorch framework, and the Optuna tool is used for hyperparameter optimization.

[0152] It should be noted that the global fitting module further includes a hierarchical parsing module, a contribution degree analysis module, and an attenuation degree analysis module connected in sequence.

[0153] The hierarchical parsing module is used to determine the series level and parallel level of each photovoltaic panel according to the photovoltaic group topology structure, and obtain the voltage contribution coefficient of the photovoltaic panel according to the series level of the photovoltaic panel is:

[0154] 。

[0155] Among them is the number of photovoltaic panels in the series level where the photovoltaic panel is located, is the photovoltaic panel 's voltage contribution coefficient.

[0156] Obtain the current contribution coefficient of the photovoltaic panel according to the parallel level of the photovoltaic panel It is:

[0157] .

[0158] Wherein is the number of photovoltaic panels in the parallel level where the photovoltaic panel is located, and is the current contribution coefficient of the photovoltaic panel .

[0159] For example: Taking the No. 1 photovoltaic panel as an example, since the topology structure of the photovoltaic group is composed of 10 photovoltaic panels in series in each row to form a string, and a total of 5 parallel strings jointly form a 10×5 matrix structure photovoltaic group, the number of photovoltaic panels in its series level is 10, and the number of photovoltaic panels in its parallel level is 5. Thus, the voltage contribution coefficient of this photovoltaic panel and the current contribution coefficient of this photovoltaic panel can be obtained.

[0160] The contribution degree analysis module is used to obtain the contribution degree of the photovoltaic panel according to the voltage contribution coefficient and the current contribution coefficient of the photovoltaic panel It is:

[0161] .

[0162] Wherein is the contribution degree of the photovoltaic panel .

[0163] For example: When the voltage contribution coefficient of the No. 1 photovoltaic panel and the current contribution coefficient , the contribution degree of this photovoltaic panel . Similarly, the contribution degrees of the remaining photovoltaic panels can be obtained.

[0164] The attenuation degree analysis module is used to summarize the contribution degrees of all photovoltaic panels to obtain the global contribution degree matrix, and sequentially call the contribution degree of each photovoltaic panel in the global contribution degree matrix and multiply it by its corresponding spectral attenuation factor vector and then sum to obtain the global attenuation degree It is:

[0165] .

[0166] Wherein is the total number of photovoltaic panels in the photovoltaic group, is the spectral attenuation factor corresponding to the th optical band in the spectral attenuation factor vector, and is the global attenuation degree of the th optical band.

[0167] For example, taking the ultraviolet light band as an example, since the spectral attenuation factor vector [0.23, 0.08, 0.06] of the No. 1 photovoltaic panel is obtained through the above process, the spectral attenuation factor of this photovoltaic panel corresponding to the ultraviolet light band , the contribution degree of the photovoltaic panel , similarly, the spectral attenuation factors and the contribution degrees of the photovoltaic panels corresponding to the ultraviolet light band of the remaining photovoltaic panels can also be obtained. Thus, the global attenuation degree of the ultraviolet light band can be calculated as follows . Assuming that the solution result is 0.31, it means that the global attenuation degree of the ultraviolet light band is 0.31.

[0168] It should be noted that the power prediction module further includes a similarity determination module, a first prediction unit, a second prediction unit, and a third prediction unit connected in sequence.

[0169] The similarity determination module is used to obtain the weather index prediction data of the target prediction date, calculate the difference coefficient between the weather index prediction data of the target prediction date and the historical weather index data, select the date with the smallest difference coefficient as the reference date, and retrieve the light band energy density distribution data corresponding to the reference date as the target light band energy density distribution data.

[0170] For example: calculating the difference coefficient between the weather index prediction data of the target prediction date and the historical weather index data to obtain the historical date with the smallest difference coefficient and taking the light band energy density distribution data of this day as the target light band energy density distribution data. From the historical record data, it can be known that the light band energy density distribution data of this day is [85.20, 442.50, 273.10], then the target light band energy density distribution data is [85.20, 442.50, 273.10].

[0171] The first prediction unit is used to obtain the solar position prediction data of the target time point from the regional weather station, and obtain the predicted value of the light incident angle according to the solar position prediction data and the installation angle of the photovoltaic panel.

[0172] The obtaining the predicted value of the light incident angle according to the solar position prediction data and the installation angle of the photovoltaic panel is as follows:

[0173] .

[0174] Where is the predicted value of the solar altitude angle, is the predicted value of the solar azimuth angle, is the installation inclination angle of the photovoltaic panel, is the azimuth angle of the photovoltaic panel, is the predicted value of the light incident angle.

[0175] For example: Obtain the predicted solar position data at the target time point from the regional meteorological station, where the predicted value of the solar altitude angle = 68.5°, and the predicted value of the solar azimuth angle = 136.2°. Then the predicted value of the light incident angle .

[0176] The second prediction unit is used to obtain the total predicted value of the effective spectral energy based on the light segment energy density values of each light segment in the target light segment energy density distribution data and the predicted value of the light incident angle which is:

[0177] .

[0178] where is the target light segment energy density of the th light segment, is the total predicted value of the effective spectral energy of the th light segment, is the area of a single photovoltaic panel, is the total number of photovoltaic panels in the photovoltaic group.

[0179] For example: The total predicted value of the effective spectral energy in the ultraviolet light segment , the total predicted value of the effective spectral energy in the visible light segment , and the total predicted value of the effective spectral energy in the infrared light segment .

[0180] The third prediction unit is used to calculate the overall predicted power of the photovoltaic group based on the global attenuation degree and the total predicted value of the effective spectral energy which is:

[0181] .

[0182] where is the energy conversion coefficient of the th light segment, is the overall predicted power of the photovoltaic group.

[0183] The energy conversion coefficient is respectively the energy conversion coefficients of the ultraviolet light segment, visible light segment, and near-infrared light segment, which are calibrated through the spectral segment efficiency experiment of the photovoltaic panel under standard test conditions. The energy conversion coefficients of the ultraviolet light segment, visible light segment, and near-infrared light segment are respectively taken as: , , , the globally obtained attenuation degrees are assumed as follows: when the globally attenuation degree in the ultraviolet light band is 0.31, the globally attenuation degree in the visible light band is 0.09, and the globally attenuation degree in the infrared light band is 0.07, the overall predicted power of the photovoltaic group .

[0184] It should be noted that the weather indicators include: solar irradiance, cloud cover, temperature, visibility; the similarity determination module further includes an index normalization unit and a difference coefficient calculation unit connected in sequence.

[0185] The index normalization unit is used to normalize the predicted weather indicator data of each target prediction date and the historical weather indicators of the corresponding category by the maximum-minimum normalization method to obtain the normalized weather indicator data.

[0186] The difference coefficient calculation unit is used to calculate the Euclidean distance of the normalized weather indicator data according to the Euclidean distance calculation method based on the normalized weather indicator data to obtain the difference coefficient.

[0187] For example: when the weather indicators at the predicted target time point are: solar irradiance 950 W / m², cloud cover 15%, temperature 32 °C, visibility 15 km, and the weather indicators on a certain day in history are obtained as: solar irradiance 920 W / m², cloud cover 15%, temperature 31 °C, visibility 14 km, and the range of solar irradiance in the historical data is 600 - 1100 W / m², then the normalized value of the solar irradiance 950 W / m² at the target time point is , and the normalized value of the solar irradiance 920 W / m² at the historical time point is , when the range of cloud cover in the historical data is 0 - 100%, then the normalized values of the cloud cover at the target time point and the cloud cover of 15% at the historical time point are , similarly, when the range of temperature in the historical data is 5 - 40 °C, then the normalized values of the temperature at the target time point and the temperature at the historical time point are 0.77 and 0.74 respectively, when the range of visibility in the historical data is 0.1 - 50 KM, then the normalized values of the visibility at the target time point and the visibility at the historical time point are 0.30 and 0.28 respectively, and the difference coefficient calculated by the Euclidean distance method for the two is:

[0188] .

[0189] If the difference coefficient on this day in the historical date is the smallest, that is, the similarity is the highest, then it is selected as the reference date.

[0190] Finally, it should be noted that although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A real-time monitoring and prediction system for new energy photovoltaic power based on the Internet of Things, characterized in that, The system includes: a data acquisition module, an energy density conversion module, an attenuation identification module, a global fitting module, and a power prediction module, and the modules are connected in sequence; The data acquisition module is used to collect the spectral incident data of the photovoltaic group area in real time through multi-spectral sensors arranged in a preset area of the photovoltaic panel group, and collect the current parameters and voltage parameters of each photovoltaic panel in real time through current and voltage sensors to obtain the photovoltaic power of each photovoltaic panel; The energy density conversion module is used to divide the spectral incident data according to the light segment type to obtain a light segment data set and perform energy density conversion to obtain light segment energy density distribution data; the light segment types include ultraviolet light segment, visible light segment, and near-infrared light segment; The attenuation identification module is used to calculate the effective spectral energy value of the light segment according to the light segment energy density distribution data and the installation angle of the photovoltaic panel, and load the effective spectral energy value of the light segment and the power data of each photovoltaic panel into the photovoltaic attenuation identification model to obtain the spectral attenuation factor vector of each photovoltaic panel; The global fitting module is used to perform hierarchical recursive analysis according to the photovoltaic group topology structure to obtain a global contribution matrix, and calculate the global attenuation degree according to the global contribution matrix and the spectral attenuation factor vector of each photovoltaic panel; The power prediction module is used to obtain the weather index prediction data at the target prediction time point and perform a historical weather index similarity determination to obtain a reference time point, use the light segment energy density distribution data corresponding to the reference time point as the target light segment energy density distribution data, and perform power prediction according to the global attenuation degree and the target light segment energy density distribution data to obtain the overall predicted power of the photovoltaic group; The energy density conversion module further includes a data preprocessing module, an energy density calculation module, a discrete integration module, and a data integration module that are connected in sequence; The data preprocessing module is used to divide the spectral incident data according to the light segment type for each discrete wavelength point to obtain the irradiance data of the discrete wavelength points contained in each light segment as the light segment data set; An energy density calculation module, configured to calculate the corresponding energy density value according to the irradiance data at discrete wavelength points as follows: ； wherein is the spectral irradiance at discrete wavelength points ; is the Planck constant is the speed of light is the wavelength at discrete wavelength points ; is the energy density value at discrete wavelength points ; A discrete integration module is used to perform discrete integration on the energy density values at discrete wavelength points within the ultraviolet, visible, and near-infrared light bands to obtain the light band energy density value. Namely: ； Among them is the spectral resolution of the multispectral sensor, is the energy density value of the optical band; The data integration module is used to summarize the light segment energy density values of the ultraviolet light segment, visible light segment, and near-infrared light segment to obtain the light segment energy density distribution data; The attenuation identification module further includes an angle conversion module, an energy analysis module, and a generative adversarial network module that are connected in sequence; An angle conversion module, which is used to obtain solar position data from a regional weather station through an API interface, and obtain the light incident angle according to the solar position data and the installation angle of the photovoltaic panel It is: ； where is the solar altitude angle, is the solar azimuth angle, is the installation tilt angle of the photovoltaic panel, is the azimuth angle of the photovoltaic panel, is the incident angle of light; An energy analysis module, configured to obtain an effective spectral energy value of an optical segment based on the optical segment energy density value and the light incident angle of each optical segment in the optical segment energy density distribution data It is: ； Among them is the area of a single photovoltaic panel, is the effective spectral energy value of the optical segment; The generative adversarial network module is used to load the effective spectral energy value of the light segment and the power data of each photovoltaic panel into the photovoltaic attenuation identification model established by the generative adversarial network algorithm to obtain the spectral attenuation factor vector of each photovoltaic panel; The global fitting module further includes a hierarchical parsing module, a contribution analysis module, and an attenuation analysis module that are connected in sequence; Hierarchical parsing module, configured to determine the series level and parallel level of each photovoltaic panel according to the photovoltaic group topology structure, and obtain the voltage contribution coefficient of the photovoltaic panel according to the series level of the photovoltaic panel Namely: ； Among them is the number of photovoltaic panels in the series level where the photovoltaic panel is located ; is the voltage contribution coefficient of the photovoltaic panel . Obtain the current contribution coefficient of the photovoltaic panel according to the parallel level of the photovoltaic panel as follows: ； Among them is the number of photovoltaic panels in the parallel level where the photovoltaic panels are located, is the current contribution coefficient of the photovoltaic panels; A contribution analysis module, which is used to obtain the contribution degree of a photovoltaic panel according to the voltage contribution coefficient and the current contribution coefficient of the photovoltaic panel as follows: ； Among them is the contribution degree of the photovoltaic panel ; The attenuation degree analysis module is used to summarize the contribution degrees of all photovoltaic panels to obtain a global contribution degree matrix, and sequentially retrieve the contribution degree of each photovoltaic panel in the global contribution degree matrix, multiply it by its corresponding spectral attenuation factor vector, and then sum them up to obtain the global attenuation degree as follows: ； Among them is the total number of photovoltaic panels in the photovoltaic array, is the spectral attenuation factor corresponding to the th optical band in the spectral attenuation factor vector, is the global attenuation degree of the th optical band.

2. The new energy photovoltaic power real-time monitoring and prediction system based on the Internet of Things according to claim 1, wherein, The generative adversarial network module further includes an input unit, a light segment response unit, a power reconstruction unit, and a discriminant unit that are connected in sequence; An input unit for generating a fused feature by passing the optical-section effective spectral energy value and each photovoltaic panel power data as inputs through a fully connected layer is as follows: ； Among them is the power of the photovoltaic panel, and are the weight matrix and bias term of the fully connected layer, is the fused feature; The spectral band response unit is used to separately set up independent sub-networks for the ultraviolet band, visible light band, and near-infrared light band to perform spectral band response on the fused features, and obtain the spectral attenuation factor vectors of each light band through LeakyReLU activation It is as follows: ； where is the spectral attenuation factor vector, and are the weight matrix and bias term of each optical section subnet, and LeakyReLU is the activation function; The power reconstruction unit is used to reconstruct the photovoltaic panel power through a fully connected regression network according to the spectral attenuation factor vector of the photovoltaic panel to obtain the reconstructed power of the photovoltaic panel; A discrimination unit is used to input the absolute error between the reconstructed power of the photovoltaic panel and the power data of the photovoltaic panel into a multi-layer perceptron to obtain an attenuation discrimination probability. When the attenuation discrimination probability is less than a preset discrimination threshold, it is determined that the spectral attenuation factor vector of the photovoltaic panel is valid and stored. Otherwise, parameter update is performed through backpropagation of the photovoltaic attenuation generator loss function to regenerate the spectral attenuation factor vector.

3. The real-time monitoring and prediction system for new energy photovoltaic power based on the Internet of Things according to claim 1, characterized in that, The training steps of the photovoltaic attenuation recognition model are as follows: Taking the effective spectral energy values in the light section and the corresponding photovoltaic panel power data during the historical operation period as training data; Initializing the parameters of the photovoltaic attenuation generator and the photovoltaic attenuation discriminator by the He initialization method, setting the initial learning rates of the photovoltaic attenuation generator and the photovoltaic attenuation discriminator, and updating the learning rates through the cosine annealing learning rate scheduling strategy; Performing iterative training on the photovoltaic attenuation generator and the photovoltaic attenuation discriminator through an alternating training strategy. Each training iteration includes: Fixing the parameters of the photovoltaic attenuation generator, continuously training the photovoltaic attenuation discriminator, updating the discriminator parameters according to the photovoltaic attenuation discriminator loss function, and performing parameter update with the discriminator learning rate through an optimizer; ； Among them is the attenuation discrimination probability is the loss function value of the photovoltaic attenuation discriminator Fixing the parameters of the photovoltaic attenuation discriminator, training the photovoltaic attenuation generator, updating the generator parameters according to the photovoltaic attenuation generator loss function, and performing parameter update with the generator learning rate through the Adam optimizer. The photovoltaic attenuation generator loss function is: ； Among them is the reconstructed power of the photovoltaic panel, is the reconstruction error weight coefficient, is the power of the photovoltaic panel, is the loss function value of the photovoltaic attenuation generator; Terminating the iteration when the training reaches the preset maximum number of iterations.

4. An Internet of Things-based new energy photovoltaic power real-time monitoring and prediction system according to claim 1, characterized in that, The power prediction module further includes a similarity determination module, a first prediction unit, a second prediction unit, and a third prediction unit connected in sequence; The similarity determination module is used to obtain the weather index prediction data of the target prediction date, calculate the difference coefficient between the weather index prediction data of the target prediction date and the historical weather index data, select the date with the smallest difference coefficient as the reference date, and retrieve the light section energy density distribution data corresponding to the reference date as the target light section energy density distribution data; The first prediction unit is used to obtain the predicted sun position data at the target time point from the regional weather station, and obtain the predicted light incident angle value according to the predicted sun position data and the photovoltaic panel installation angle; A second prediction unit, configured to obtain a total predicted value of effective spectral energy according to the optical segment energy density values of each optical segment in the target optical segment energy density distribution data and the predicted value of the light incident angle which is ； Among them is the target optical segment energy density of the optical segment, is the total predicted value of the effective spectral energy of the optical segment, is the area of a single photovoltaic panel, is the total number of photovoltaic panels in the photovoltaic group; The third prediction unit is used to calculate the overall predicted power of the photovoltaic group based on the global attenuation degree and the predicted total value of the effective spectral energy as follows: ； Among them is the energy conversion coefficient of the nth optical segment, and is the overall predicted power of the photovoltaic array. It should be noted that there seems to be an error in the original text where the variable is used twice in an unclear way in the English translation above. It might need to be corrected in the original text for a more accurate and meaningful translation.

5. The real-time monitoring and prediction system for new energy photovoltaic power based on the Internet of Things according to claim 4, characterized in that, The weather indicators include: solar irradiance, cloud cover, temperature, visibility; the similarity determination module further includes an index normalization unit and a difference coefficient calculation unit connected in sequence; The index normalization unit is used to normalize the weather index prediction data of each target prediction date and the corresponding historical weather index of the corresponding category by the maximum-minimum normalization method to obtain the normalized weather index data; The difference coefficient calculation unit is used to calculate the Euclidean distance of the normalized weather index data according to the Euclidean distance calculation method of the normalized weather index data to obtain the difference coefficient.

Images