Photovoltaic power prediction method and system and computer equipment

Abstract

The invention relates to the technical field of photovoltaic systems, solves the technical problems that in the prior art, irradiance and temperature cannot be accurately predicted, and gradient disappearance or gradient explosion is prone to occurring in the prediction process, and particularly relates to a photovoltaic power prediction method and system and computer equipment. The method comprises the following steps: S1, obtaining a cloud picture data set from a meteorological database of a place where photovoltaic equipment is located; s2, selecting a plurality of zenith cloud pictures from the cloud picture data set and preprocessing the zenith cloud pictures to obtain a foundation cloud picture set consisting of a plurality of foundation cloud pictures; according to the method, the movement curve of the cloud layer in the sky is obtained through image analysis of the zenith cloud picture, and the prediction model is established according to the movement curve, so that the problem of gradient disappearance or gradient explosion of the neural network model can be avoided, real-time cloud layer detection is performed on the zenith cloud picture, the prediction result is more accurate, the accuracy of irradiance is improved, and the prediction efficiency is improved. The prediction process is smoother, and the working efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic systems, and in particular to a photovoltaic power prediction method, system and computer equipment. Background Art

[0002] Solar irradiance is the radiation energy per unit area per unit time that reaches the ground after being absorbed, scattered and reflected by the atmosphere. Its unit is watt per square meter (W / m2). The power of photovoltaic power generation is related to the solar irradiance falling on its surface. Solar irradiance and temperature are very critical to the power generation of photovoltaic systems. The existing technology uses neural network algorithm models to predict irradiance and temperature, which is prone to inaccurate predictions. This is mainly because the neural network algorithm is prone to gradient disappearance or gradient explosion problems when predicting nonlinear irradiance. Once the temperature prediction takes into account many factors, the processing speed is slow, which further leads to inaccurate photovoltaic power prediction and reduces the working efficiency of photovoltaic power generation. Summary of the invention

[0003] In view of the shortcomings of the prior art, the present invention provides a photovoltaic power prediction method, system and computer equipment, which solve the technical problems that the prior art cannot accurately predict irradiance and temperature, and the prediction process is prone to gradient disappearance or gradient explosion. The irradiance and temperature at the location of the photovoltaic system are accurately predicted through a new prediction model, further improving the accuracy of photovoltaic power prediction.

[0004] In order to solve the above technical problems, the present invention provides the following technical solutions: a photovoltaic power prediction method, the method comprising the following steps:

[0005] S1. Obtain a cloud image data set consisting of a number of zenith cloud images from a meteorological database at the location of the photovoltaic equipment;

[0006] S2, preprocessing the cloud image data set to obtain a ground-based cloud image set consisting of a number of ground-based cloud images;

[0007] S3. Select any point on the ground-based cloud map as the cloud moving point coordinate (B a , C a ), according to the cloud moving point coordinates (B a , C a )'s motion trajectory generates a predicted trajectory S', and optimizes the predicted trajectory S' to obtain an optimized trajectory S;

[0008] S4. Select a prediction time H and find the prediction point E that matches the prediction time H on the optimization trajectory S. a (F e , K f ), and according to the predicted point E a(F e , K f ) to obtain the predicted irradiation value J;

[0009] S5. Calculate the wind speed value V in the area where the photovoltaic equipment is located at the current stage s ;

[0010] S6, according to the wind speed value V s Calculate the surface heat value T;

[0011] S7. Predict the power P of the photovoltaic equipment based on the surface heat value T and the predicted irradiation value J y .

[0012] Preferably, in step S2, the specific implementation steps are as follows:

[0013] S21, obtaining a first zenith cloud image from a plurality of zenith cloud images, and filtering the first zenith cloud image by a mean filtering method to obtain a filtered image;

[0014] S22, define the value of 365 days as day number N, and define the value of 24 hours as hour value L;

[0015] S23. Calculate the solar declination angle ε and the solar hour angle ω according to the day number N and the hour value L. The calculation formula is as follows:

[0016]

[0017] ω=15×(L-12)

[0018] Among them, ε represents the solar declination angle, ω represents the solar hour angle;

[0019] S24. Obtain the latitude value of the location of the photovoltaic equipment from the meteorological database Calculate the sun position angle γ based on the sun's declination angle ε and the sun's hour angle ω s and the zenith angle θ a , the calculation formula is as follows:

[0020]

[0021] Among them, θ a represents the zenith angle, γ s represents the sun position angle;

[0022] S25, according to the sun position angle γ s and the zenith angle θ a The time of the first zenith cloud image is taken as the starting point, and several zenith cloud images of consecutive adjacent times are selected to form a zenith cloud image set;

[0023] S26, setting the area with missing information in the first zenith cloud image as the missing area Ω, setting the area with valid information in the zenith cloud image I as the valid area Φ, and setting the boundary line between the missing area Ω and the valid area Φ as the boundary line ΦΩ;

[0024] S27, convert the first day top cloud image into an RGB image through the image processing library, and select any pixel point on the boundary line ΦΩ as the point p to be repaired. a , waiting for the repair point p a Select the block k to be repaired as the center;

[0025] S28. Randomly select a pixel point in the effective area Φ as a matching point And obtain the matching block k' which is the same size as the image block k to be repaired, and calculate the point p to be repaired a with matching points The color difference CCM between them is calculated as follows:

[0026]

[0027] Among them, CCM represents the point to be repaired p a with matching points The color difference between a )、G(p a ) and B(p a ) represents the point p to be repaired a The red, green, and blue color values ​​of and Indicates matching point The red, green, and blue color values ​​of

[0028] S29, obtain the color difference threshold Y by empirical method s And according to the color difference threshold Y s Repair the image block k to be repaired;

[0029] If CCM <Y s , then copy the matching block k' and replace the block k to be repaired;

[0030] If CCM ≥ Y s , it does not meet the repair requirements and returns to step S28;

[0031] S210, repeating step S28 to step S29 until the zenith cloud atlas is repaired and the ground cloud atlas is obtained.

[0032] Preferably, in step S3, the specific implementation steps are as follows:

[0033] S31, taking the selected ground-based cloud image as the first ground-based cloud image, establishing a two-dimensional coordinate system with the center point (e, f) of the first ground-based cloud image as the origin, and obtaining a first cloud layer image on the first ground-based cloud image by an edge detection method;

[0034] S32, randomly select a cloud moving point a at the center of the cloud layer in the first cloud layer image. S (B a , C a );

[0035] S33, taking the time of the first cloud layer image as ∈, selecting the second ground-based cloud image at time ∈-1 in the ground-based cloud image collection and obtaining the second cloud layer image through step S31, taking the midpoint of the second cloud layer image as the reference point to calculate the cloud motion point a of the second cloud layer image. S Mapped into the first cloud layer image as mapping point a S1 (B b , C b );

[0036] S34, repeat step 33 to obtain several mapping points a Sm =[a S1 , a S2 ……aS m] , according to several mapping points a Sm The coordinates of are used to obtain the predicted trajectory S' through curve fitting method. The expression of the predicted trajectory S' is:

[0037] S'=W g B x 2 +W k B x +b

[0038] Among them, W g and W k represents the trajectory coefficient, b represents the bias;

[0039] S35, find the prediction point (Y d , Z d ), find the ground cloud map of the predicted time in the ground cloud map collection and get the actual point (Y d’ , Z d’ );

[0040] S36, using the loss function to calculate the predicted point (Y d , Z d ) and the actual point (Y d’ , Z d’ ) between the difference V;

[0041] S37, optimizing the predicted trajectory S' according to the difference threshold M;

[0042] If V < M, it meets the expectation, ends, and obtains the optimized trajectory S;

[0043] If V ≥ M, it does not meet the expectation and returns to step S33.

[0044] Preferably, the calculation formula for the difference V is as follows:

[0045]

[0046] where V represents the difference between the predicted point (Y d , Z d ) and the actual point (Y d’ , Z d’ ), and n represents the number of pairs of the predicted point (Y d , Z d ) and the actual point (Y d’ , Z d’ ).

[0047] Preferably, in step S4, the specific implementation steps are as follows:

[0048] In step S4, the specific implementation steps are as follows:

[0049] S41. Calculate the distance Δd a (F e , K f ) from the predicted point E to the center point (e, f), and the calculation formula is as follows: 1j where Δd 1j represents the distance from the jth predicted point E a (F e , K f ) to the center point (e, f);

[0052] S42. Obtain the sun center point SU z (L h , Q k ) in the first cloud image, calculate the distance Δd 2j from the predicted point E a (F e , K f ) to the sun center point SU z (L h , Q k ), and the calculation formula is as follows:

[0053]

[0054] where Δd​​​​​​2j Indicates the jth prediction point E a (F a , K f ) to the center of the sun SU z (L h , Q k ) distance;

[0055] S43, according to the predicted point E a (F e , K f ) corresponds to the gray value Gr b , distance Δd 1j and distance Δd 2j Calculate the cloud cover value LO f , the calculation formula is as follows:

[0056]

[0057] Among them, Lo f represents the cloud cover value, w represents the number of data;

[0058] S44, according to the cloud cover value LO f Calculate the predicted irradiation value J, the calculation formula is as follows:

[0059] J=I po ×cosθ gd ×(1-LO f )

[0060] Where J represents the predicted irradiance value, I po represents the clear sky irradiance, θ gd Represents the sun's altitude angle.

[0061] Preferably, in step S5, the specific implementation steps are as follows:

[0062] S51, selecting a first basic direction μ from a plurality of directions of a horizontal plane where the photovoltaic system is located, and taking a direction in the plane perpendicular to the first basic direction μ as a second basic direction π;

[0063] S52, measuring the detection distance S between two detection points a and b in the first basic direction μ μ , and the detection distance D between the two detection points c and e in the second basic direction π π ;

[0064] S53, recording the downwind transmission time t of the ultrasonic wave between two detection points in the first basic direction μ ab and upwind transmission time t ba ;

[0065] S54, according to the detection distance S μ, downwind transmission time t ab and upwind transmission time t ba Calculate the average wind speed in the first base direction μ during the photovoltaic generation time The calculation formula is as follows:

[0066]

[0067] in, represents the average wind speed during the photovoltaic power generation period, l represents the detection distance S μ , downwind transmission time t ab and upwind transmission time t ba The number of groups, S μ represents the μth detection distance;

[0068] S55, obtaining the downwind transmission time t of the ultrasonic wave between two detection points in the second basic direction π ce and upwind transmission time t ec ;

[0069] S56, calculating the average wind speed in the second basic direction π The calculation formula is as follows:

[0070]

[0071] in, represents the average wind speed in the second basic direction π, and y represents the detection distance D π , downwind transmission time t ce and upwind transmission time t ec The number of groups, D π represents the πth detection distance;

[0072] S57, based on average wind speed and average wind speed Calculate the wind speed value V s , the calculation formula is as follows:

[0073]

[0074] Among them, V s Indicates wind speed value.

[0075] Preferably, in step S6, the specific implementation steps are as follows:

[0076] S61, obtaining the power generation time period of the photovoltaic power generation equipment, and dividing the power generation time period into a number of time elements t f ;

[0077] S62. Calculate the surface reflected heat value R based on the surface thermal energy data s , the calculation formula is as follows:

[0078] R s =α×J

[0079] Among them, R s represents the surface reflected heat value, α represents the surface reflectivity, and J represents the predicted irradiance value;

[0080] S63, according to the surface reflected heat value R s , wind speed value V s Calculate the surface heat value T, the calculation formula is as follows:

[0081]

[0082] Among them, t f1 and t f2 Represents two time elements, dT represents the differential of the surface heat value T, dt represents the differential of time, ∈ represents the surface emissivity, σ c represents the Stepan-Boltzmann constant, C p represents the specific heat of air, ρ represents the air density, T b represents the atmospheric temperature, Δz s Represents the effective value of the surface thickness layer, C m Represents the surface heat capacity.

[0083] Preferably, the specific implementation steps in step S7 are as follows:

[0084] S71. Calculate power current I according to predicted irradiation value J p , the calculation formula is as follows:

[0085]

[0086] Among them, I p Indicates power current, S N represents standard irradiance, T represents surface heat value, T1 represents standard temperature, D T Indicates the current coefficient, I s Indicates the standard irradiance s N and the short-circuit current under standard temperature T1, J represents the predicted irradiation value;

[0087] S72, according to the power current I p Calculate power voltage U p , the calculation formula is as follows:

[0088]

[0089] Among them, U p Indicates power voltage, B s Indicates the number of photovoltaic cells connected in series, B prepresents the number of photovoltaic cells in parallel, A represents the diode parameter, K represents the Boltzmann constant, T0 represents the power generation temperature, I0 represents the equivalent reverse diode saturation current inside the photovoltaic cell, and E represents the elementary charge;

[0090] S73, according to the power voltage U p and power current I p Calculate power P y , the calculation formula is as follows:

[0091] P y =U p ×I p

[0092] Among them, P y Indicates the power of the photovoltaic system.

[0093] The technical solution also provides a system for the photovoltaic power prediction method, the system comprising:

[0094] A data acquisition module obtains a cloud map data set from a meteorological database at the location of the photovoltaic equipment;

[0095] A preprocessing module selects a number of zenith cloud images from the cloud image data set and performs preprocessing to obtain a ground cloud image set consisting of a number of ground cloud images;

[0096] Trajectory optimization module, select any point on the ground-based cloud map as the cloud moving point coordinate (B a , C a ), according to the cloud moving point coordinates (B a , C a )'s motion trajectory generates a predicted trajectory S', and optimizes the predicted trajectory S' to obtain an optimized trajectory S;

[0097] The irradiation value prediction module selects a prediction time H and finds the prediction point E that matches the prediction time H on the optimization trajectory S. a (F e , K f ), and according to the predicted point E a (F e , K f ) to obtain the predicted irradiation value J;

[0098] Wind speed analysis module, calculates the wind speed value V in the area where the photovoltaic equipment is located at the current stage s ;

[0099] Heat calculation module, based on wind speed value V s Calculate the surface heat value T;

[0100] Photovoltaic power prediction module predicts the power P of photovoltaic equipment based on the surface heat value T and predicted irradiation value J y .

[0101] The technical solution also provides a computer device, which includes a processor and a memory connected via a system bus.

[0102] By means of the above technical solution, the present invention provides a photovoltaic power prediction method, system and computer device, which have at least the following beneficial effects:

[0103] 1. The present invention obtains the movement curve of clouds in the sky through image analysis of the zenith cloud map, establishes a prediction model based on the movement curve, and then predicts the cloud cover factor based on the relationship between the cloud layer and the sun's position, and obtains the irradiance through the cloud cover factor and the sun's altitude angle. This method can avoid the problem of gradient disappearance or gradient explosion of the neural network model, and also performs real-time cloud detection on the zenith cloud map, so that the prediction result is more accurate, the accuracy of the irradiance is improved, the prediction process is smoother, and the work efficiency is improved.

[0104] 2. The present invention can more accurately obtain the wind speed at the location of the photovoltaic system by establishing mutually perpendicular measurements of the transmission time of ultrasonic waves with or against the wind, and then obtain the temperature of the photovoltaic system through the accurate wind speed, so that the predicted photovoltaic power is more accurate, the accuracy of photovoltaic power prediction is improved, and the operation efficiency of the photovoltaic power station is higher.

[0105] 3. The present invention measures the surface calorific value, calculates the measured value in units of time units, and combines the heat radiated and reflected by the surface, so that the surface calorific value can be more accurate, avoiding the situation where the surface calorific value calculation deviation is too large due to daytime temperature changes, which affects the prediction result of photovoltaic power. BRIEF DESCRIPTION OF THE DRAWINGS

[0106] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0107] Figure 1 is a flow chart of the photovoltaic power prediction method of the present invention;

[0108] Figure 2 is a structural block diagram of the photovoltaic power prediction system of the present invention;

[0109] Figure 3 is a structural block diagram of the computer device of the present invention;

[0110] Figure 4 This is an example diagram of the zenith cloud map of the present invention;

[0111] Figure 5 This is an example diagram of the restored zenith cloud image of the present invention. DETAILED DESCRIPTION

[0112] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific implementation methods, so that the implementation process of how the present application uses technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly.

[0113] Since the existing technology cannot accurately predict irradiance and temperature, the prediction process is prone to gradient disappearance or gradient explosion. Please refer to Figure 1 - Figure 5 This embodiment provides a photovoltaic power prediction method, system and computer device, which can accurately predict the irradiance and temperature of the photovoltaic system location through a new prediction model, further improving the accuracy of photovoltaic power prediction. The method includes the following steps:

[0114] S1. Obtain a cloud image dataset consisting of several zenith cloud images from the meteorological database at the location of the photovoltaic equipment. The cloud image dataset contains several zenith cloud images, such as Figure 4 Shown is an example of a zenith cloud map. The zenith cloud map is taken through a camera on the top, which shoots pictures vertically towards the hemispherical reflective mirror below. The reflection of the hemispherical mirror can capture the cloud state of the entire sky. The superposition of multiple cloud pictures can better reflect the movement trajectory of the clouds. Each zenith cloud map is marked with an accurate recording time to facilitate later predictions.

[0115] S2. Preprocess the cloud image data set to obtain a ground-based cloud image set consisting of several ground-based cloud images. Collect the cloud image data through a cloud image camera. In actual application, the all-sky imager TSI-880 is used to take cloud image photos of the sky. Since a sun position baffle is set in the image of the all-sky imager to block the sun, the camera bracket will also be seen in the photo, such as Figure 4 As shown in FIG. 1 , due to the blocking effect of the sun position baffle and the camera bracket, there must be missing areas in the image, which may cause inconvenience to the later fitting of the cloud movement trajectory. Therefore, the image needs to be preprocessed. In step S2, the specific implementation steps are as follows:

[0116] S21. Obtain a first zenith cloud image from a number of zenith cloud images, and filter the first zenith cloud image by a mean filtering method to obtain a filtered image. The present invention only needs to use a mean filtering method to achieve the expected effect, and the mean filtering method is a commonly used filtering method, which will not be described in detail here.

[0117] S22. Define the value of 365 days as the day number N, and define the value of 24 hours as the hour value L. For the convenience of calculation, only the values ​​of the days and hours are used for calculation.

[0118] S23, calculate the solar declination angle ε and solar hour angle ω before and after the time to be predicted according to the day number N and the hour value L, and the calculation formula is as follows:

[0119]

[0120] ω=15×(L-12)

[0121] Among them, ε represents the solar declination angle, and ω represents the solar hour angle. When predicting the cloud trajectory in the later stage, the time to be predicted will be determined first. In practice, the solar declination angle ε and solar hour angle ω three days to one week before and after the time to be predicted can be obtained.

[0122] S24. Obtain the latitude value of the location of the photovoltaic equipment from the meteorological database Calculate the sun position angle γ based on the sun's declination angle ε and the sun's hour angle ω s and the zenith angle θ a , the calculation formula is as follows:

[0123]

[0124] Among them, θ a represents the zenith angle, γ s Indicates the sun position angle, which is expressed by the sun position angle γ in the zenith cloud map. s and the zenith angle θ a It can better determine the position of the sun and also facilitate the determination of the sun's position in cloud images later.

[0125] S25, according to the sun position angle γ s and the zenith angle θ a The time of the first zenith cloud image is taken as the starting point, and several zenith cloud images of consecutive adjacent time periods are selected to form a zenith cloud atlas. The images in the cloud image data set must be temporally adjacent and continuous ground-based cloud image images, so that the cloud layer can be predicted more accurately.

[0126] S26. The area with missing information in the first zenith cloud image is set as the missing area Ω, the area with valid information in the zenith cloud image I is set as the valid area Φ, and the boundary line between the missing area Ω and the valid area Φ is set as the boundary line ΦΩ. Due to the occlusion of the equipment, the actual part of the image must be repaired.

[0127] S27, convert the first day top cloud image into an RGB image through the image processing library, and select any pixel point on the boundary line ΦΩ as the point p to be repaired. a , waiting for the repair point p a Selecting the block k to be repaired as the center and defining the surface by point can make the repair process more accurate and improve the prediction accuracy in the later stage.

[0128] S28. Randomly select a pixel point in the effective area Φ as a matching point And obtain the matching block k' which is the same size as the image block k to be repaired, and calculate the point p to be repaired a with matching points The color difference CCM between them is calculated as follows:

[0129]

[0130] Among them, CCM represents the point to be repaired p a with matching points The color difference between a )、G(p a ) and B(p a ) represents the point p to be repaired a The red, green, and blue color values ​​of and Indicates matching point The red, green, and blue color values ​​of

[0131] S29, obtain the color difference threshold Y by empirical method s And according to the color difference threshold Y s Repair the image block k to be repaired;

[0132] If CCM <Y s , then copy the matching block k' and replace the block k to be repaired;

[0133] If CCM ≥ Y s , it does not meet the repair requirements and returns to step S28;

[0134] S210, repeating steps S28 to S29 until the zenith cloud atlas is repaired and the ground cloud atlas is obtained. Here, all zenith cloud images need to be repaired to facilitate the prediction of cloud movement trajectory in the later stage. According to the above steps, the zenith cloud images with missing information can be quickly repaired. Figure 4 The figure shows the repaired zenith cloud image, which makes the zenith cloud image more accurate in the later step of fitting the movement trajectory of the clouds and avoids the situation where the prediction step is terminated or failed due to incomplete pictures.

[0135] S3. Select any point on the ground-based cloud map as the cloud moving point coordinate (B a , C a ), according to the cloud moving point coordinates (B a , C a)'s motion trajectory is used to generate a predicted trajectory S', and the predicted trajectory S' is optimized to obtain an optimized trajectory S. The motion trajectory of the cloud layer is predicted by fitting a curve. The cloud center points at multiple different times are mapped on the same ground-based cloud map using the ground-based cloud map that is continuous and adjacent in time to complete the fitting point selection. Since the existing technology cannot accurately predict the irradiance and temperature, the prediction process is prone to technical problems such as gradient disappearance or gradient explosion. In order to obtain a more accurate prediction result, in step S3, the specific implementation steps are as follows:

[0136] S31. The selected ground-based cloud map is used as the first ground-based cloud map, and a two-dimensional coordinate system is established with the center point (e, f) of the first ground-based cloud map as the origin. The first cloud layer image is obtained on the first ground-based cloud map through the edge detection method. The present invention can achieve the expected effect only through the edge detection method, and the edge detection method is a commonly used method for obtaining edges, which will not be elaborated here. The cloud layer portrait can be obtained through the edge detection method.

[0137] S32, randomly select a cloud moving point a at the center of the cloud layer in the first cloud layer image. S (B a , C a ), in the process of taking points, the geometric center of the cloud layer is generally obtained by geometric methods, and multiple moving points can also be obtained on the cloud layer. You can synthesize multiple curves to make the prediction of the cloud layer more accurate. The present invention takes a moving point as the illustration object.

[0138] S33, taking the time of the first cloud layer image as ∈, selecting the second ground-based cloud image at time ∈-1 in the ground-based cloud image collection and obtaining the second cloud layer image through step S31, taking the midpoint of the second cloud layer image as the reference point to calculate the cloud motion point a of the second cloud layer image. S Mapped into the first cloud layer image as mapping point a S1 (B d , C b );

[0139] S34, repeat step 33 to obtain several mapping points a Sm =[a S1 , a S2 ……a Sm ], according to several mapping points a Sm The coordinates of are used to obtain the predicted trajectory S' through curve fitting method. The expression of the predicted trajectory S' is:

[0140] S'=W g B x 2 +W k B x +b

[0141] Among them, W g and Wk Let \(a\) represent the trajectory coefficient and \(b\) represent the offset. The zenith cloud image is a circular spherical image. The movement of the cloud layer is generally a parabola or an arc. In the present invention, the parabola is used for illustration, and the research on the trajectory of the arc can be modified into a circular curve expression.

[0142] S35. Find the prediction point \((Y d , Z d ) that matches the prediction time on the predicted trajectory \(S'\). Find the ground-based cloud image at the prediction time in the ground-based cloud atlas and obtain the actual point \((Y a’ , Z d’ ) in this ground-based cloud image;

[0143] S36. Calculate the difference \(V\) between the prediction point \((Y d , Z d ) and the actual point \((Y d’ , Z d’ ). The calculation formula is as follows:

[0144]

[0145] where \(V\) represents the difference between the prediction point \((Y d , Z d ) and the actual point \((Y d’ , Z d’ ), and \(n\) represents the logarithm of the prediction point \((Y d , Z d ) and the actual point \((Y d’ , Z d’ ).

[0146] S37. Optimize the predicted trajectory \(S'\) according to the difference threshold \(M\);

[0147] If \(V < M\), it meets the expectation, end and obtain the optimized trajectory \(S\);

[0148] If \(V\geq M\), it does not meet the expectation and return to step S33. Steps S36 - S37 can further optimize the fitting curve, make the fitting curve more accurate, increase the accuracy of the final prediction. This step can obtain the preliminary trajectory curve of the cloud layer in the sky, and establish the preliminary predicted trajectory \(S'\) of the trajectory fitting curve of the cloud center point according to the trajectory curve, and finally obtain the optimized trajectory \(S\) through the optimization of the preliminary trajectory curve, making the prediction result of the method for the cloud layer trajectory more accurate.

[0149] S4. Select a prediction time \(H\), find the prediction point \(E a (F e , K f ) that matches the prediction time \(H\) on the optimized trajectory \(S\), and according to the prediction point \(E a (F e , Kf ) to obtain the predicted irradiance value J. Since the prior art cannot accurately predict the irradiance and temperature, the prediction process is prone to the technical problems of gradient disappearance or gradient explosion. In this embodiment, time is used as a prediction variable to more accurately predict the movement trajectory of the cloud layer. In step S4, the specific implementation steps are as follows:

[0150] S41. Calculate the predicted point E a (F e , K f ) to the center point (e,f) Δd 1j , the calculation formula is as follows:

[0151]

[0152] Where, Δd 1j Indicates the jth prediction point E a (F e , K f ) to the center point (e,f);

[0153] S42, obtaining the sun center point SU in the first cloud layer image z (L h , Q k ), calculate the predicted point E a (F e , K f ) to the center of the sun SU z (L h , Q k ) 2j , the calculation formula is as follows:

[0154]

[0155] Where, Δd 2j Indicates the jth prediction point E a (F e , K f ) to the center of the sun SU z (L h , Q k ) distance;

[0156] S43, according to the predicted point E a (F e , K f ) corresponds to the gray value Gr b , distance Δd 1j and distance Δd 2j Calculate the cloud cover value LO f , the calculation formula is as follows:

[0157]

[0158] Among them, LO f represents the cloud cover value, w represents the number of data;

[0159] S44, according to the cloud cover value LO f Calculate the predicted irradiation value J, the calculation formula is as follows:

[0160] J=I po ×cosθ gd ×(1-LO f )

[0161] Where J represents the predicted irradiance value, I po represents the clear sky irradiance, θ gd Indicates the solar altitude angle, clear sky irradiance I po and the solar altitude angle θ gd It can be directly obtained through the meteorological database. The present invention obtains the movement curve of clouds in the sky through image analysis of the zenith cloud map, and establishes a prediction model based on the movement curve. The cloud cover factor is predicted according to the relationship between the cloud layer and the sun's position. The irradiance is obtained through the cloud cover factor and the sun's altitude angle. This method avoids the problem of gradient vanishing or gradient exploding in the neural network model, and performs real-time cloud detection on the zenith cloud map, so that the prediction result is more accurate, the accuracy of the irradiance is improved, the prediction process is smoother, and the work efficiency is improved.

[0162] S5. Calculate the wind speed value V in the area where the photovoltaic equipment is located at the current stage s In actual measurement, due to the change of wind speed, the wind speed measured using a single-direction method lacks accuracy. In order to obtain a more accurate measurement value, in step S5, the specific implementation steps are as follows:

[0163] S51. Select a first basic direction μ from several directions of the horizontal plane where the photovoltaic system is located, and use a direction in the plane that is perpendicular to the first basic direction μ as the second basic direction π. The first basic direction μ can be selected arbitrarily on the horizontal plane, and the two perpendicular directions can facilitate the final calculation of the wind speed.

[0164] S52, measuring the detection distance S between two detection points a and b in the first basic direction μ μ , and the detection distance D between the two detection points c and e in the second basic direction π π The distance between the detection points a and b or the detection points c and e can be measured by a variety of methods, such as using an acoustic wave ranging method or a manual measurement method to obtain the detection distance.

[0165] S53, recording the downwind transmission time t of the ultrasonic wave between two detection points in the first basic direction μ ab and upwind transmission time t baThe transmission time of the ultrasonic wave is accurately timed by a timing device, the timing device includes a timing chip of model TDC-GP2, and other timing methods are also acceptable. The propagation time of the ultrasonic wave is different in the wind and against the wind. This physical phenomenon can be used to make the wind speed measurement more accurate. The ultrasonic wave in the present invention uses the most commonly used intermediate frequency ultrasonic wave.

[0166] S54, according to the detection distance S μ , downwind transmission time t ab and upwind transmission time t ba Calculate the average wind speed in the first base direction μ during the photovoltaic generation time The calculation formula is as follows:

[0167]

[0168] in, represents the average wind speed during the photovoltaic power generation period, l represents the detection distance S μ , downwind transmission time t ab and upwind transmission time t ba The number of groups, S μ It represents the μth detection distance. Since the measured distance does not take the direction factor into account, the average wind speed is calculated.

[0169] S55, obtaining the downwind transmission time t of the ultrasonic wave between two detection points in the second basic direction π ce and upwind transmission time t ec ;

[0170] S56, calculating the average wind speed in the second basic direction π The calculation formula is as follows:

[0171]

[0172] in, represents the average wind speed in the second basic direction π, and y represents the detection distance D π , downwind transmission time t ec and upwind transmission time t ec The number of groups, D π represents the πth detection distance,

[0173] S57, based on average wind speed and average wind speed Calculate the wind speed value V s , the calculation formula is as follows:

[0174]

[0175] Among them, V s Indicates the wind speed value, due to the average wind speed and average wind speed The direction is vertical, so the Pythagorean theorem is used to directly calculate the wind speed value. The present invention can more accurately obtain the wind speed at the location of the photovoltaic system by establishing mutually perpendicular measurements of the transmission time of ultrasonic waves with or against the wind, and then obtain the temperature of the photovoltaic system through the accurate wind speed, so that the predicted photovoltaic power is more accurate, the accuracy of photovoltaic power prediction is improved, and the operation efficiency of the photovoltaic power station is higher.

[0176] S6, according to the wind speed value V s Calculate the surface heat value T. The surface heat value at the location of the photovoltaic system has a direct impact on the power. The predicted irradiance value J and wind speed value V are required to calculate the surface heat value. s The prior art is too simple in calculating the calorific value, and too few physical quantities are referenced, which easily leads to the problem of inaccurate calculation of the calorific value. In order to improve the accuracy of the calorific value, in step S6, the specific implementation steps are as follows:

[0177] S61, obtaining the power generation time period of the photovoltaic power generation equipment, and dividing the power generation time period into a number of time elements t f Dividing time into several time units can improve the calculation accuracy, make the surface heat value T more accurate, and avoid the fluctuation of the surface heat value T causing large errors in the results.

[0178] S62. Calculate the surface reflected heat value R based on the surface thermal energy data s , the calculation formula is as follows:

[0179] R s =α×J

[0180] Among them, R s represents the surface reflected heat value, α represents the surface reflectivity, and J represents the predicted irradiance value;

[0181] S63, according to the surface reflected heat value R s , wind speed value V s Calculate the surface heat value T, the calculation formula is as follows:

[0182]

[0183] Among them, t f1 and t f2 Represents two time elements, dT represents the differential of the surface heat value T, dt represents the differential of time, ∈ represents the surface emissivity, σ c represents the Stepan-Boltzmann constant, C p represents the specific heat of air, ρ represents the air density, T b represents the atmospheric temperature, Δz S Represents the effective value of the surface thickness layer, Cm To represent the surface heat capacity, after dividing the time into several time units, the surface heat value T is calculated by the integration method, which can avoid the situation where simple calculation leads to inaccurate results. The present invention measures the surface heat value, calculates the measured value in time units, and combines the heat of surface radiation and reflection, so that the surface heat value can be more accurate, avoiding the situation where the surface heat value calculation deviation is too large due to daytime temperature changes, which affects the prediction result of photovoltaic power.

[0184] S7. Predict the power P of the photovoltaic equipment based on the surface heat value T and the predicted irradiation value J y Photovoltaic power is a relatively complex physical quantity that needs to be calculated based on a large number of values. Since the existing technology cannot accurately predict irradiance and temperature, the prediction process is prone to technical problems such as gradient disappearance or gradient explosion. In order to make the photovoltaic power prediction more accurate, in step S7, the specific implementation steps are as follows:

[0185] S71. Calculate power current I according to predicted irradiation value J p , the calculation formula is as follows:

[0186]

[0187] Among them, I p Indicates power current, S N represents standard irradiance, T represents surface heat value, T1 represents standard temperature, D T Indicates the current coefficient, I s Indicates the standard irradiance S N and the short-circuit current under standard temperature T1 conditions, J represents the predicted irradiation value, and the physical quantities related to current and voltage can be directly obtained from the photovoltaic power generation database without the need for complex calculations.

[0188] S72, according to the power current I p Calculate power voltage U p , the calculation formula is as follows:

[0189]

[0190] Among them, U p Indicates power voltage, B s Indicates the number of photovoltaic cells connected in series, B p represents the number of photovoltaic cells in parallel, A represents the diode parameters, K represents the Boltzmann constant, T0 represents the power generation temperature, I0 represents the equivalent reverse diode saturation current inside the photovoltaic cell, and E represents the elementary charge, which is 1.6×10 -19 C, Boltzmann constant is 1.380649×10 -23J / K, diode parameters are obtained from the nameplate of the diode in the photovoltaic system.

[0191] S73, according to the power voltage U p and power current I p Calculate power P y , the calculation formula is as follows:

[0192] P y =U p ×I p

[0193] Among them, P y Indicates the power of the photovoltaic system. The present invention obtains the movement curve of the cloud layer in the sky through image analysis of the zenith cloud map, and establishes a prediction model based on the movement curve. The cloud cover factor is predicted according to the relationship between the cloud layer and the sun's position. The irradiance is obtained through the cloud cover factor and the sun's altitude angle. This method can avoid the problem of gradient disappearance or gradient explosion of the neural network model, and also performs real-time cloud layer detection on the zenith cloud map, so that the prediction result is more accurate, the accuracy of the irradiance is improved, the prediction process is smoother, and the work efficiency is improved.

[0194] See also Figure 2 , which is a structural block diagram of the photovoltaic power prediction system provided in this embodiment, and the photovoltaic power prediction system includes a data acquisition module, a preprocessing module, a trajectory optimization module, an irradiation value prediction module, a wind speed analysis module, a heat calculation module, and a photovoltaic power prediction module.

[0195] The data acquisition module obtains the cloud image data set from the meteorological database where the photovoltaic equipment is located. The preprocessing module selects several zenith cloud images from the cloud image data set and performs preprocessing to obtain a ground cloud image set composed of several ground cloud images. The trajectory optimization module selects any point on the ground cloud image as the cloud moving point coordinate (B a , C a ), according to the cloud moving point coordinates (B a , C a ) generates a predicted trajectory S', and optimizes the predicted trajectory S' to obtain the optimized trajectory S. The irradiation value prediction module selects a prediction time H and finds a prediction point E on the optimized trajectory S that matches the prediction time H. a (F e , K f ), and according to the predicted point E a (F e , K f ) to obtain the predicted irradiance value J, and the wind speed analysis module calculates the wind speed value V in the area where the photovoltaic equipment is located at the current stage s , heat calculation module, according to the wind speed value V sCalculate the surface heat value T, photovoltaic power prediction module, predict the power P of photovoltaic equipment according to the surface heat value T and predicted irradiation value J y .

[0196] See also Figure 3 , this embodiment also provides a computer device, which includes a processor and a memory connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory includes a storage medium and an internal memory. The storage medium can be a non-volatile storage medium or a volatile storage medium. The storage medium stores an operating system and can also store computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor can implement a photovoltaic power prediction method. The internal memory provides an environment for the operation of the operating system and computer-readable instructions in the storage medium. The internal memory can also store computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor can execute the photovoltaic power prediction method. The network interface of the computer device is used to communicate with an external server through a network connection. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the shell of the computer device, or an external keyboard, touchpad or mouse.

[0197] Those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a program, so the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application can be in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0198] The above implementation methods have been described in detail. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A photovoltaic power prediction method, characterized in that: The method includes the following steps: S1. Obtain a cloud map data set composed of several zenith cloud maps from the meteorological database at the location of the photovoltaic device; S2. Preprocess the cloud map data set to obtain a ground-based cloud map set composed of several ground-based cloud maps; S3. Select any point on the ground-based cloud map as the cloud moving point coordinate (B a , C a ), according to the cloud moving point coordinates (B a , C a )'s motion trajectory generates a predicted trajectory S', and optimizes the predicted trajectory S' to obtain an optimized trajectory S; S4. Select a prediction time H and find the prediction point E that matches the prediction time H on the optimization trajectory S. a (F e , K f ), and according to the predicted point E a (F e , K f ) to obtain the predicted irradiation value J; S5. Calculate the wind speed value V in the area where the photovoltaic equipment is located at the current stage s ; S6, according to the wind speed value V s Calculate the surface heat value T; S7. Predict the power P of the photovoltaic equipment based on the surface heat value T and the predicted irradiation value J y .

2. The photovoltaic power prediction method according to claim 1, characterized in that: In step S2, the specific implementation steps are as follows: S21. Obtain the first zenith cloud map from several zenith cloud maps, and filter the first zenith cloud map by the mean filtering method to obtain a filtered image; S22. Define the value corresponding to 365 days as the day number N, and the value corresponding to 24 hours as the hour value L; S23. Calculate the solar declination ε and the solar hour angle ω according to the day number N and the hour value L. The calculation formulas are as follows: ω = 15×(L - 12) where ε represents the solar declination and ω represents the solar hour angle; S24. Obtain the latitude value of the location of the photovoltaic equipment from the meteorological database Calculate the sun position angle γ based on the sun's declination angle ε and the sun's hour angle ω s and the zenith angle θ a , the calculation formula is as follows: Among them, θ a represents the zenith angle, γ s represents the sun position angle; S25, according to the sun position angle γ s and the zenith angle θ a The time of the first zenith cloud image is taken as the starting point, and several zenith cloud images of consecutive adjacent times are selected to form a zenith cloud image set; S26. Set the area with missing information in the first zenith cloud map as the missing area Ω, set the area with valid information in the zenith cloud map I as the valid area Φ, and set the boundary line between the missing area Ω and the valid area Φ as the boundary line ΦΩ; S27, convert the first day top cloud image into an RGB image through the image processing library, and select any pixel point on the boundary line ΦΩ as the point p to be repaired. a , waiting for the repair point p a Select the block k to be repaired as the center; S28. Randomly select a pixel point in the effective area Φ as a matching point And obtain the matching block k' which is the same size as the image block k to be repaired, and calculate the point p to be repaired a with matching points The color difference CCM between them is calculated as follows: Among them, CCM represents the point to be repaired p a with matching points The color difference between a )、G(p a ) and B(p a ) represents the point p to be repaired a The red, green, and blue color values ​​of and Indicates matching point The red, green, and blue color values ​​of S29, obtain the color difference threshold Y by empirical method s And according to the color difference threshold Y s Repair the image block k to be repaired; If CCM <Y s , then copy the matching block k' and replace the block k to be repaired; If CCM ≥ Y s , it does not meet the repair requirements and returns to step S28; S210. Repeat steps S28 to S29 until the zenith cloud map set is repaired and a ground-based cloud map set is obtained.

3. The photovoltaic power prediction method according to claim 1, characterized in that: In step S3, the specific implementation steps are as follows: S31. Take the selected ground-based cloud map as the first ground-based cloud map, establish a two-dimensional coordinate system with the center point (e, f) of the first ground-based cloud map as the origin, and obtain the first cloud layer image on the first ground-based cloud map by the edge detection method; S32, randomly select a cloud moving point a at the center of the cloud layer in the first cloud layer image. S (B a , C a ); S33, taking the time of the first cloud layer image as ∈, selecting the second ground-based cloud image at time ∈-1 in the ground-based cloud image collection and obtaining the second cloud layer image through step S31, taking the midpoint of the second cloud layer image as the reference point to calculate the cloud motion point a of the second cloud layer image. S Mapped into the first cloud layer image as mapping point a S1 (B b , C b ); S34, repeat step 33 to obtain several mapping points a Sm =[a S1 , a S2 ......a Sm ], according to several mapping points a Sm The coordinates of are used to obtain the predicted trajectory S' through curve fitting method. The expression of the predicted trajectory S' is: S’=W g B x 2 +W k B x +b Among them, W g and W k represents the trajectory coefficient, b represents the bias; S35, find the prediction point (Y d , Z d ), find the ground-based cloud map of the predicted time in the ground-based cloud map set and get the actual point (Y d’ , Z d’ ); S36, calculate the predicted point (Y d , Z d ) and the actual point (Y d’ , Z d’ ) between the difference V; S37. Optimize the prediction trajectory S according to the difference threshold M; If V < M, it meets the expectation, end and obtain the optimized trajectory S; If V ≥ M, it does not meet the expectation and return to step S33.

4. The photovoltaic power prediction method according to claim 3, characterized in that: The calculation formula of the difference V is as follows: Where V represents the predicted point (Y d , Z d ) and the actual point (Y d’ , Z d’ ), n represents the difference between the predicted point (Y d , Z d ) and the actual point (Y d’ , Z d’ ) is the logarithm of .

5. The photovoltaic power prediction method according to claim 1, characterized in that: In step S4, the specific implementation steps are as follows: S41. Calculate the predicted point E a (F e , K f ) to the center point (e, f) Δd 1j , the calculation formula is as follows: Where, Δd 1j Indicates the jth prediction point E a (F e , K f ) to the center point (e, f); S42, obtaining the sun center point SU in the first cloud layer image z (L h , Q k ), calculate the predicted point E a (F e , K f ) to the center of the sun SU z (L h , Q k ) 2j , the calculation formula is as follows: Where, Δd 2j Indicates the jth prediction point E a (F e , K f ) to the center of the sun SU z (L h , Q k ) distance; S43, according to the predicted point E a (F e , K f ) corresponds to the gray value Gr b , distance Δd 1j and distance Δd 2j Calculate the cloud cover value LO f , the calculation formula is as follows: Among them, LO f represents the cloud cover value, w represents the number of data; S44, according to the cloud cover value LO f Calculate the predicted irradiation value J, the calculation formula is as follows: J=I po ×cosθ gd ×(1-LO f ) Where J represents the predicted irradiance value, I po represents the clear sky irradiance, θ gd Represents the sun's altitude angle.

6. The photovoltaic power prediction method according to claim 1, characterized in that: In step S5, the specific implementation steps are as follows: S51. Select the first basic direction μ from several directions on the horizontal plane where the photovoltaic system is located, and take a direction perpendicular to the first basic direction μ in this plane as the second basic direction π; S52, measuring the detection distance S between two detection points a and b in the first basic direction μ μ , and the detection distance D between the two detection points c and e in the second basic direction π π ; S53, recording the downwind transmission time t of the ultrasonic wave between two detection points in the first basic direction μ ab and upwind transmission time t ba ; S54, according to the detection distance S μ , downwind transmission time t ab and upwind transmission time t ba Calculate the average wind speed in the first base direction μ during the photovoltaic generation time The calculation formula is as follows: in, represents the average wind speed during the photovoltaic power generation period, l represents the detection distance S μ , downwind transmission time t ab and upwind transmission time t ba The number of groups, S μ represents the μth detection distance; S55, obtaining the downwind transmission time t of the ultrasonic wave between two detection points in the second basic direction π ce and upwind transmission time t ec ; S56, calculating the average wind speed in the second basic direction π The calculation formula is as follows: in, represents the average wind speed in the second basic direction π, and y represents the detection distance D π , downwind transmission time t ce and upwind transmission time t ec The number of groups, D π represents the πth detection distance; S57, based on average wind speed and average wind speed Calculate the wind speed value V s , the calculation formula is as follows: Among them, V s Indicates wind speed value.

7. The photovoltaic power prediction method according to claim 1, characterized in that: In step S6, the specific implementation steps are as follows: S61, obtaining the power generation time period of the photovoltaic power generation equipment, and dividing the power generation time period into a number of time elements t f ; S62. Calculate the surface reflected heat value R based on the surface thermal energy data s , the calculation formula is as follows: R s =α×J Among them, R s represents the surface reflected heat value, α represents the surface reflectivity, and J represents the predicted irradiance value; S63, according to the surface reflected heat value R s , wind speed value V s Calculate the surface heat value T, the calculation formula is as follows: Among them, t f1 and t f2 Represents two time elements, dT represents the differential of the surface heat value T, dt represents the differential of time, ∈ represents the surface emissivity, σ c represents the Stepan-Boltzmann constant, C p represents the specific heat capacity of air, ρ represents the air density, T b represents the atmospheric temperature, Δz s Represents the effective value of the surface thickness layer, C m Represents the surface heat capacity.

8. The photovoltaic power prediction method according to claim 1, characterized in that: In step S7, the specific implementation steps are as follows: S71. Calculate power current I according to predicted irradiation value J p , the calculation formula is as follows: Among them, I p Indicates power current, S N represents standard irradiance, T represents surface heat value, T1 represents standard temperature, D T Indicates the current coefficient, I s Indicates the standard irradiance S N and the short-circuit current under standard temperature T1, J represents the predicted irradiation value; S72, according to the power current I p Calculate power voltage U p , the calculation formula is as follows: Among them, U p Indicates power voltage, B s Indicates the number of photovoltaic cells connected in series, B p represents the number of photovoltaic cells in parallel, A represents the diode parameter, K represents the Boltzmann constant, T0 represents the power generation temperature, I0 represents the equivalent reverse diode saturation current inside the photovoltaic cell, and E represents the elementary charge; S73, according to the power voltage U p and power current I p Calculate power P y , the calculation formula is as follows: P y =U p ×I p Among them, P y Indicates the power of the photovoltaic system.

9. A system comprising the photovoltaic power prediction method according to any one of claims 1 to 8, characterized in that: The system includes: A data acquisition module, which obtains a cloud map data set from the meteorological database at the location of the photovoltaic device; A preprocessing module, which selects several zenith cloud maps from the cloud map data set and preprocesses them to obtain a ground-based cloud map set composed of several ground-based cloud maps; Trajectory optimization module, select any point on the ground-based cloud map as the cloud moving point coordinate (B a , C a ), according to the cloud moving point coordinates (B a , C a )'s motion trajectory generates a predicted trajectory S', and optimizes the predicted trajectory S' to obtain an optimized trajectory S; The irradiation value prediction module selects a prediction time H and finds the prediction point E that matches the prediction time H on the optimization trajectory S. a (F e , K f ), and according to the predicted point E a (F e , K f ) to obtain the predicted irradiation value J; Wind speed analysis module, calculates the wind speed value V in the area where the photovoltaic equipment is located at the current stage s ; Heat calculation module, based on wind speed value V s Calculate the surface heat value T; Photovoltaic power prediction module predicts the power P of photovoltaic equipment based on the surface heat value T and predicted irradiation value J y .

10. A computer device comprising a processor and a memory connected via a system bus, characterized in that: When the processor executes the computer program, it implements the photovoltaic power prediction method described in claims 1-8.

Images