Photovoltaic output prediction method, system, device and medium

Abstract

The invention discloses a photovoltaic output prediction method, system and device and a medium, and provides an enhanced kernel ridge regression photovoltaic output data prediction method of ensemble empirical mode decomposition. According to the method, the output influence factor information is subjected to noise reduction processing by using ensemble empirical mode decomposition, the reliability of the data is improved, and the interference of data noise on the data is reduced. A kernel ridge regression method is used for learning a nonlinear mapping relation between input information and a prediction residual error to obtain a real-time prediction residual error so as to improve the accuracy of a prediction model, and a long-short-term memory network of a convolutional neural network based on an attention mechanism is combined with a DBCSAN method for screening and removing when measurement is abnormal. Noise reduction is carried out on the measurement information through ensemble empirical mode decomposition, the average error and the root-mean-square value of historical data of photovoltaic influence factors are reduced, the information is closer to a true value, more reliable training data are provided for follow-up prediction work, and then the prediction effect is effectively improved.

Description

Technical Field

[0001] The present application relates to the technical field of photovoltaic output prediction, and in particular to a photovoltaic output prediction method, system, device and medium. Background Art

[0002] At present, there are two mainstream methods for predicting photovoltaic output data, namely physical method and statistical method. The physical method needs to establish a complex parameter model, which has extremely high requirements on the accuracy of parameters and data, and the accuracy of the established model is not good. The statistical method uses artificial intelligence algorithms such as neural networks for prediction, but the current statistical methods have poor adaptability to nonlinear problems. Moreover, when there are errors in the input data, the prediction accuracy of photovoltaic output data is also insufficient, making it difficult to fully explore the relationship between the factors affecting photovoltaic output and the actual photovoltaic output data.

[0003] At the same time, since the statistical method requires a large amount of historical data, and some abnormal information in the data will affect the prediction accuracy of the statistical model, there is currently a lack of a photovoltaic output data prediction solution with high prediction accuracy. Summary of the invention

[0004] The present application provides a photovoltaic output prediction method, system, device and medium to solve the above-mentioned problems.

[0005] In one aspect, the present application provides a photovoltaic output prediction method, the method comprising the following steps:

[0006] Step S1: Collect relevant impact data of photovoltaic power station operation, and use historical operation data to establish a mapping model between photovoltaic output impact data and photovoltaic output at corresponding time;

[0007] Step S2: collecting real-time photovoltaic output data and inputting it into a pre-built and trained photovoltaic output prediction model for prediction; wherein the photovoltaic output prediction model is associated with an enhanced kernel ridge regression model of ensemble empirical mode decomposition;

[0008] Step S3: Outputting predicted photovoltaic process data.

[0009] In one implementation of the present application, in step S2, multiple intrinsic mode functions and a residual component function are separated from the photovoltaic output-related influencing data through collective empirical mode decomposition; wherein the intrinsic mode function represents the dynamic change law of a certain instantaneous frequency of the influencing factor information, highlighting the local characteristics of the photovoltaic output influencing data, and the residual component reflects the overall change trend of the influencing factor information.

[0010] In one implementation of the present application, the process of collective empirical mode decomposition is specifically as follows:

[0011] Calculate the influencing factor information for a specific period of time e All extreme points of , and use the cubic spline interpolation method to connect the selected maximum and minimum values ​​to obtain x e The upper and lower envelopes of the curves are calculated, and the mean curve of the upper and lower envelopes is calculated. Using z e minus A new influencing factor component with low frequency removed is obtained, as shown in the following formula:

[0012]

[0013] If the amount If the number of extreme points and zero-crossing points in the entire photovoltaic output impact data is equal or differs by at most one, and the average value of the upper and lower envelopes determined by the cubic spline difference method is 0, then this component is Photovoltaic output impact data e If the first-order IMF of The calculation is repeated as input, as shown in the following formula:

[0014]

[0015] In the formula, is the intermediate process component, for The mean curve of the upper and lower envelopes obtained using cubic spline difference. It is worth noting that the abnormal data collected by the influencing factor system will interfere with the division of the upper and lower envelopes and affect the decomposition effect. Therefore, it is still necessary to use the abnormal data processing method proposed in Chapter 2 to reconstruct the abnormal data. Assuming that the IMF requirements are met after k operations, the influencing factor information z e The first-order IMF component L 1 The residual influencing factor component of the separated first-order IMF They are:

[0016] L 1 =h 1k

[0017]

[0018] After separating the first-order IMF, the original influencing factor x e It may also contain IMF components, so it is necessary to Continue the above operation until the remaining component is a monotonic function. Assuming that the requirements are met after n times, the separation process is:

[0019]

[0020] In the formula, r nis the nth-order residual component, which is a monotonic function. At this time, the original photovoltaic output affects the data z e It can be expressed as several IMF components and the final residual component The sum is as follows:

[0021] In one implementation of the present application, the ensemble empirical mode decomposition uses a noise-assisted method to suppress the modal aliasing caused by the presence of noise data in the influencing data, specifically:

[0022] The original influencing factor z e Add Gaussian white noise with a mean of 0 to obtain the influencing factor information z after white noise processing υ,e , to cover up the original noise in the influencing factor information.

[0023] z υ,e =z e +v

[0024] Information on influencing factors after white noise processing x υ,e Perform EEMD decomposition, and z υ,e Decomposed into the following formula:

[0025]

[0026] In the formula, v i is the white noise added for the i-th time, and then the IMF component of each white noise added can be obtained Considering that the mean value of Gaussian noise is 0, the above IMF components can be averaged to obtain the final IMF components of EEMD decomposition as follows: In the formula, m is the number of times white noise is added, and generally m is less than 100. After adding m times of Gaussian white noise, the statistical rules are as follows: Where ΔE m is the influencing factor information υ,e and the difference between the final IMF component and; ε is the amplitude of white noise. The more times noise is added, the closer the IMF component is to the original decomposition when the amplitude of white noise is constant, and the final decomposition result of influencing factor information is obtained.

[0027] In one implementation of the present application, the method further includes:

[0028] To calculate the sample entropy of the influencing factor components, firstly, the influencing factor information IMF components are combined into a vector sequence as follows:

[0029] L n (i) = [L υ,e (i),L υ,e (i+1),...,L υ,e(i+m-1)]

[0030] Definition L υ,e (i),L υ,e The distance between (i+1) is the absolute value of the maximum difference between the corresponding elements of each IMF component as follows: d[L n (i),L n (j)]=max|L υ,e (i+k)-L υ,e (j+k)|, where 1≤k≤n-1; 1≤i, j≤m-n+1, ​​i≠j. For a given L n (i) Calculate L n (i) With L n (j) The distance between n (i),L n (j)] is less than the ratio of the number of j components to the total components, and is recorded as B i,n (r) is as follows,

[0031]

[0032] Calculate B i,n (r) All mean values ​​for i:

[0033] For the n+1 point vector, the ratio and average value are calculated as follows:

[0034]

[0035] The obtained sequence sample entropy is:

[0036]

[0037] When m takes a finite value: SampEn(n,r,m)=-ln[A n (r) / B n (r)], where n is the embedding dimension and r is the similarity tolerance.

[0038] In one implementation of the present application, the nonlinear mapping relationship between photovoltaic output influencing factor information and photovoltaic output data information is analyzed based on the ridge regression method, specifically:

[0039] The mapping relationship is established based on the following formula: t ′=h -1 (Z t ′), where Z t ′ is the factor vector of photovoltaic output influencing factors with non-Gaussian distribution; X t ′ is the corresponding prediction result; h -1 is a nonlinear mapping function from the photovoltaic output influencing factors to the photovoltaic output data;

[0040] For the ordinary linear regression model, we have: k =w k Z k , where X k is the predicted value of photovoltaic output data, w k is the weight parameter, Z k is the information matrix of factors affecting the photovoltaic output of the node, and the weight parameter w is obtained by the least squares method. k When the objective function of the least squares linear regression is set to: J(w) = ∥Xw k Z k ∥ 2 , where X is the true value of photovoltaic output data information. Considering the multicollinearity of the influencing factors in the distribution network, the weight parameter w k The value of is too large, which leads to large fluctuations in the prediction results of photovoltaic output data. Therefore, in the ridge regression method, a penalty function is used to constrain the least squares model. Specifically, l 2 norm, the objective function of the ridge regression model becomes: J(w) = ∥X k -w k Z k ∥ 2 +λ||w k || 2 , where λ is added to balance the regularization term and the loss function. To minimize the loss function J(w), the weight parameter is derived and taken as 0, as shown in the following formula;

[0041]

[0042] is equivalent to:

[0043]

[0044] After simplifying by taking 0, we can get the parameter w k The optimal solution is as follows:

[0045]

[0046] Where I is the unit matrix and the optimal solution is is the positive definite sum of a semi-positive definite matrix and a diagonal matrix;

[0047] Optimize the parameters To transform, first use the matrix inversion theorem:

[0048] (U -1 +V T W -1 V) -1 V T W -1=UV T (VUV T +W) -1

[0049] In the formula V=Z k ,W=I, which is equivalent to:

[0050]

[0051] The weight parameter becomes:

[0052]

[0053] Define a nonlinear mapping function φ(·) and transform the ridge regression model objective function into: The optimal solution is:

[0055] The kernel function K is defined as follows:

[0056] X(S i ,S j )=[φ(S i ),φ(S j )]=φ(S i )φ T (S j )

[0057] make:

[0058] K=K(Z k ,Z k )=φ(Z k )φ(Z k ) T

[0059] α=(φ(Z k )φ(Z k ) T +λI) -1 X

[0060] The intermediate variable α becomes: α=(K+λI) -1 X k ;

[0061] The weight parameters after kernel function processing become:

[0062]

[0063] Then the photovoltaic output forecast value is obtained as follows:

[0064]

[0065] In one implementation of the present application, the method further includes: optimizing the gap between the predicted result and the actual result based on the KRRSE learning method, wherein the KRRSE photovoltaic output prediction model is assumed to be:

[0066]

[0067] In the formula, is a kernel ridge regression model, which is used to learn the mapping relationship between historical influencing factor data z and photovoltaic output data x, where: The data set used is D k ={(z 1 ,x 1 ),(z 2 ,x 2 )...(z t-1 ,x t-1 )}, To approximate unknown prediction residuals;

[0068] Learning to Predict Using Kernel Ridge Regression With the true prediction model h -1 The nonlinear relationship between the prediction residual error information is shown in Figure 2. At this time, the photovoltaic output prediction model is expressed as:

[0069]

[0070] In the formula, z t-1 Represents the input of factors affecting photovoltaic output, Represents the error between the KRRSE model and the true model, model noise

[0071] The present application also provides a photovoltaic output prediction system, the system comprising:

[0072] A data collection unit is used to collect relevant influencing data of the operation of the photovoltaic power station, and use the historical operation data to establish a mapping model between the photovoltaic output influencing data and the photovoltaic output at the corresponding moment;

[0073] A prediction calculation unit, used to collect real-time photovoltaic output data and input it into a pre-built and trained photovoltaic output prediction model for prediction; wherein the photovoltaic output prediction model is associated with an enhanced kernel ridge regression model of ensemble empirical mode decomposition;

[0074] The data output unit is used to output the predicted photovoltaic processing data.

[0075] The present application also provides a photovoltaic output prediction device, the device comprising:

[0076] at least one processor; and,

[0077] a memory communicatively connected to the at least one processor; wherein,

[0078] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can complete the aforementioned photovoltaic output prediction method.

[0079] The present application also provides a non-volatile computer storage medium for photovoltaic output prediction, which stores computer executable instructions. The computer executable instructions are executed by a processor to implement the aforementioned photovoltaic output prediction method.

[0080] The photovoltaic output prediction method, system, device and medium provided in this application have the following beneficial effects:

[0081] (1) By using ensemble empirical mode decomposition to reduce the noise of the measurement information, the average error and root mean square value of the historical data of photovoltaic influencing factors are reduced, making the information closer to the true value, providing more reliable training data for subsequent prediction work, and thus effectively improving the prediction effect;

[0082] (2) After information denoising, KRRSE is used to learn the relationship between the measurement information and the residual during the training process, and the prediction residual is integrated into the real-time prediction value after the real-time measurement is input, thus obtaining an EKRRSE prediction model that is closer to the true state model, thereby improving the prediction accuracy and stability. Compared with KRRSE, EKRRSE has better prediction performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] 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:

[0084] Figure 1 A flow chart of a photovoltaic output prediction method provided in an embodiment of the present application;

[0085] Figure 2 This is a graph of 16-node photovoltaic output results when the variance is 0.6 in the embodiment of the present application;

[0086] Figure 3 This is a 16-node photovoltaic output result diagram when the variance is 1 in the embodiment of the present application;

[0087] Figure 4 A composition diagram of a photovoltaic output prediction system provided in an embodiment of the present application;

[0088] Figure 5A schematic diagram of a photovoltaic output prediction device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0089] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0090] The embodiments of the present application provide a photovoltaic output prediction method, system, device and medium. The technical solutions proposed in the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0091] Figure 1 A flow chart of a photovoltaic output prediction method provided in an embodiment of the present application. Figure 1 As shown, the method mainly includes the following steps:

[0092] Step S1: Collect relevant impact data of photovoltaic power station operation, and use historical operation data to establish a mapping model between photovoltaic output impact data and photovoltaic output at corresponding time;

[0093] Step S2: collecting real-time photovoltaic output data and inputting it into a pre-built and trained photovoltaic output prediction model for prediction; wherein the photovoltaic output prediction model is associated with an enhanced kernel ridge regression model of ensemble empirical mode decomposition;

[0094] Step S3: Outputting predicted photovoltaic process data.

[0095] In the embodiment of the present application, in step S2, multiple intrinsic mode functions and a residual component function are separated from the photovoltaic output-related influencing data through collective empirical mode decomposition; the intrinsic mode function represents the dynamic change law of a certain instantaneous frequency of the influencing factor information, highlighting the local characteristics of the photovoltaic output influencing data, and the residual component reflects the overall change trend of the influencing factor information.

[0096] In the embodiment of the present application, the process of collective empirical mode decomposition is specifically as follows:

[0097] Calculate the influencing factor information for a specific period of time e All extreme points of , and use the cubic spline interpolation method to connect the selected maximum and minimum values ​​to obtain x e The upper and lower envelopes of the curves are calculated, and the mean curve of the upper and lower envelopes is calculated. Using z e minus A new influencing factor component with low frequency removed is obtained, as shown in the following formula:

[0098]

[0099] If the amount If the number of extreme points and zero-crossing points in the entire photovoltaic output impact data is equal or differs by at most one, and the average value of the upper and lower envelopes determined by the cubic spline difference method is 0, then this component is Photovoltaic output impact data e If the first-order IMF of The calculation is repeated as input, as shown in the following formula:

[0100]

[0101] In the formula, is the intermediate process component, for The mean curve of the upper and lower envelopes obtained using cubic spline difference. It is worth noting that the abnormal data collected by the influencing factor system will interfere with the division of the upper and lower envelopes and affect the decomposition effect. Therefore, it is still necessary to use the abnormal data processing method proposed in Chapter 2 to reconstruct the abnormal data. Assuming that the IMF requirements are met after k operations, the influencing factor information z e The first-order IMF component L 1 The residual influencing factor component of the separated first-order IMF They are:

[0102] L 1 =h 1k

[0103]

[0104] After separating the first-order IMF, the original influencing factor x e It may also contain IMF components, so it is necessary to Continue the above operation until the remaining component is a monotonic function. Assuming that the requirements are met after n times, the separation process is:

[0105]

[0106] In the formula, r n is the nth-order residual component, which is a monotonic function. At this time, the original photovoltaic output affects the data z e It can be expressed as several IMF components and the final residual component The sum is as follows:

[0107] In the embodiment of the present application, the ensemble empirical mode decomposition uses a noise-assisted method to suppress the modal aliasing caused by the presence of noise data in the influencing data, specifically:

[0108] The original influencing factor z e Add Gaussian white noise v with a mean of 0 to obtain the influencing factor information z after white noise processing ν,e , to cover up the original noise in the influencing factor information.

[0109] z υ,e =z e +υ

[0110] Information on influencing factors after white noise processing x υ,e Perform EEMD decomposition, and z υ,e Decomposed into the following formula:

[0111]

[0112] In the formula, v i is the white noise added for the i-th time, and then the IMF component of each white noise added can be obtained Considering that the mean value of Gaussian noise is 0, the above IMF components can be averaged to obtain the final IMF components of EEMD decomposition as follows: In the formula, m is the number of times white noise is added, and generally m is less than 100. After adding m times of Gaussian white noise, the statistical rules are as follows: Where ΔE m is the influencing factor information υ,e and the difference between the final IMF component and; ε is the amplitude of white noise. The more times noise is added, the closer the IMF component is to the original decomposition when the amplitude of white noise is constant, and the final decomposition result of influencing factor information is obtained.

[0113] In an embodiment of the present application, the method further includes:

[0114] To calculate the sample entropy of the influencing factor components, firstly, the influencing factor information IMF components are combined into a vector sequence as follows:

[0115] L n (i) = [L υ,e (i),L υ,e (i+1),...,L υ,e (i+m-1)]

[0116] Definition L υ,e (i),L υ,e The distance between (i+1) is the absolute value of the maximum difference between the corresponding elements of each IMF component as follows: d[L n (i),Ln (j)]=max|L υ,e (i+k)-L υ,e (j+k)|, where 1≤k≤n-1; 1≤i, j≤m-n+1, ​​i≠j. For a given L n (i) Calculate L n (i) With L n (j) The distance between n (i),L n (j)] is less than the ratio of the number of j components to the total components, and is recorded as B i,n (r) is as follows,

[0117]

[0118] Calculate B i,n (r) All mean values ​​for i:

[0119] For the n+1 point vector, the ratio and average value are calculated as follows:

[0120]

[0121] The obtained sequence sample entropy is:

[0122]

[0123] When m takes a finite value: SampEn(n,r,m)=-ln[A n (r) / B n (r)], where n is the embedding dimension and r is the similarity tolerance.

[0124] In the embodiment of the present application, the nonlinear mapping relationship between photovoltaic output influencing factor information and photovoltaic output data information is analyzed based on the ridge regression method, specifically:

[0125] The mapping relationship is established based on the following formula: t ′=h -1 (Z t ′), where Z t ′ is the factor vector of photovoltaic output influencing factors with non-Gaussian distribution; X t ′ is the corresponding prediction result; h -1 is a nonlinear mapping function from the photovoltaic output influencing factors to the photovoltaic output data;

[0126] For the ordinary linear regression model, we have: k =w k Z k , where X k is the predicted value of photovoltaic output data, w kis the weight parameter, Z k is the information matrix of factors affecting the photovoltaic output of the node, and the weight parameter w is obtained by the least squares method. k When the objective function of the least squares linear regression is set to: J(w) = ∥Xw k Z k ∥ 2 , where X is the true value of photovoltaic output data information. Considering the multicollinearity of the influencing factors in the distribution network, the weight parameter w k The value of is too large, which leads to large fluctuations in the prediction results of photovoltaic output data. Therefore, in the ridge regression method, a penalty function is used to constrain the least squares model. Specifically, l 2 norm, the objective function of the ridge regression model becomes: J(w) = ∥X k -w k Z k ∥ 2 +λ||w k || 2 , where λ is added to balance the regularization term and the loss function. To minimize the loss function J(w), the weight parameter is derived and taken as 0, as shown in the following formula;

[0127]

[0128] is equivalent to:

[0129]

[0130] After simplifying by taking 0, we can get the parameter w k The optimal solution is as follows:

[0131]

[0132] Where I is the unit matrix and the optimal solution is is the positive definite sum of a semi-positive definite matrix and a diagonal matrix;

[0133] Optimize the parameters To transform, first use the matrix inversion theorem:

[0134] (U -1 +V T W -1 V) -1 V T W -1 =UV T (VUV T +W) -1

[0135] In the formula V=Z k ,W=I, which is equivalent to:

[0136]

[0137] The weight parameter becomes:

[0138]

[0139] Define a nonlinear mapping function φ(·) and transform the ridge regression model objective function into: The optimal solution is:

[0140] The kernel function K is defined as follows:

[0141] K(S i ,S j )=[φ(S i ),φ(S j )]=φ(S i )φ T (S j )

[0142] make:

[0143] K=K(Z k ,Z k )=φ(Z k )φ(Z k ) T

[0144] α=(φ(Z k )φ(Z k ) T +λI) -1 X

[0145] The intermediate variable α becomes: α=(K+λI) -1 X k ;

[0146] The weight parameters after kernel function processing become:

[0147]

[0148] Then the photovoltaic output forecast value is obtained as follows:

[0149]

[0150] In an embodiment of the present application, the method further includes: optimizing the gap between the predicted result and the actual result based on the KRRSE learning method, wherein the KRRSE photovoltaic output prediction model is assumed to be:

[0151]

[0152] In the formula, is a kernel ridge regression model, which is used to learn the mapping relationship between historical influencing factor data z and photovoltaic output data x, where: The data set used is D k ={(z 1 ,x 1 ),(z 2 ,x 2 )...(z t-1 ,x t-1 )}, To approximate unknown prediction residuals;

[0153] Learning to Predict Using Kernel Ridge Regression With the true prediction model h -1 The nonlinear relationship between the prediction residual error information is shown in Figure 2. At this time, the photovoltaic output prediction model is expressed as:

[0154]

[0155] In the formula, z t-1 Represents the input of factors affecting photovoltaic output, Represents the error between the KRRSE model and the true model, model noise

[0156] The embodiment of the present application also proposes an abnormal data processing method of LSTM (ACL) and DBSCAN based on the convolutional neural network of the attention mechanism. The LSTM neural network is used to perform feature screening again on the fine-grained features after being screened by ATT and CNN, and the fine-grained features are effectively integrated to obtain coarse-grained features, and then all the input measurement information is efficiently utilized to obtain reliable photovoltaic output data information reconstruction results.

[0157] In the embodiment of the present application, the kernel ridge regression DG output data comes from the public data set of the Belgian power grid company. The sampling frequency of all data is 15 minutes, with a total of 35040 groups of data. The training set and the prediction set are divided according to the ratio of 8:2. The computer configuration is Win10 system, the processor is Intel Core i7-8700k series, the main frequency is 3.0GHz, the graphics card is NVIDIA GTX-1660TI, and the memory is 16G. The Keras toolbox is selected for training the DG output prediction load prediction model, in which the kernel function selects the Gaussian function, and the grid search is used to select the regularization strength, and the selection range is 10 -2 , 10 -3 , 10 -4 The learning efficiency is 1×10 -4 .

[0158] In the embodiment of the present application, in order to intuitively show the impact of noise with different variance sizes on the filtering results, Figure 2 , Figure 3 The voltage state estimation results of 16 nodes are plotted when the variance is 0.6 and 1 respectively. Figure 2 It can be seen that when the variance is 0.6, some predicted deviation points appear in the voltage waveform, and the maximum error is 4.67×10 –3 , the average error is 9.77×10 –5 ; When the variance rises to 1, the number of voltage waveform deviation points increases further, and the maximum error is 7.86×10 –3 , the average error is 1.33×10 –4 , and it still remains at a low level. The main reason is that the cross-validation method used in this paper can select reasonable model parameters, thereby limiting the interference of noise data to the greatest extent. It can be seen that when the noise variance is within a reasonable range, the model proposed in this paper can effectively filter out noise.

[0159] The above is a photovoltaic output prediction method provided by the embodiment of the present application. Based on the same inventive concept, the embodiment of the present application also provides a photovoltaic output prediction system. Figure 4 A composition diagram of a photovoltaic output prediction system provided in an embodiment of the present application, such as Figure 4 As shown, the system mainly includes: a data collection unit 401, which is used to collect relevant influencing data of the operation of the photovoltaic power station, and use historical operation data to establish a mapping model between photovoltaic output influencing data and photovoltaic output at the corresponding moment; a prediction calculation unit 402, which is used to collect real-time photovoltaic output data and input it into a pre-built and trained optimized photovoltaic output prediction model for prediction; wherein the photovoltaic output prediction model is associated with an enhanced kernel ridge regression model of ensemble empirical mode decomposition; a data output unit 403, which is used to output predicted photovoltaic processing data.

[0160] The above is a photovoltaic output prediction system provided by the embodiment of the present application. Based on the same inventive concept, the embodiment of the present application also provides a photovoltaic output prediction device. Figure 5 A schematic diagram of a photovoltaic output prediction device provided in an embodiment of the present application is shown in FIG. Figure 5 As shown, the device mainly includes: at least one processor 501; and a memory 502 that is communicatively connected to the at least one processor; wherein the memory 502 stores instructions that can be executed by the at least one processor 501, and the instructions are executed by the at least one processor 501 so that the at least one processor 501 can complete the aforementioned photovoltaic output prediction method.

[0161] In addition, an embodiment of the present application further provides a non-volatile computer storage medium for photovoltaic output prediction, which stores computer executable instructions, and the computer executable instructions are executed by a processor to implement the aforementioned photovoltaic output prediction method.

[0162] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0163] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0164] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0165] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0166] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0167] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0168] The above is only the embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A photovoltaic output prediction method, characterized in that: The method comprises the following steps: Step S1: Collect relevant impact data of photovoltaic power station operation, and use historical operation data to establish a mapping model between photovoltaic output impact data and photovoltaic output at corresponding time; Step S2: collecting real-time photovoltaic output data and inputting it into a pre-built and trained photovoltaic output prediction model for prediction; wherein the photovoltaic output prediction model is associated with an enhanced kernel ridge regression model of ensemble empirical mode decomposition; Step S3: Outputting predicted photovoltaic process data.

2. A photovoltaic output prediction method according to claim 1, characterized in that: In step S2, multiple intrinsic mode functions and a residual component function are separated from the photovoltaic output-related influencing data through collective empirical mode decomposition; the intrinsic mode function represents the dynamic change law of a certain instantaneous frequency of the influencing factor information, highlighting the local characteristics of the photovoltaic output influencing data, and the residual component reflects the overall change trend of the influencing factor information.

3. A photovoltaic output prediction method according to claim 2, characterized in that: The process of ensemble empirical mode decomposition is as follows: Calculate the influencing factor information for a specific period of time e All extreme points of , and use the cubic spline interpolation method to connect the selected maximum and minimum values ​​to obtain x e The upper and lower envelopes of the curves are calculated, and the mean curve of the upper and lower envelopes is calculated. Using z e minus A new influencing factor component with low frequency removed is obtained, as shown in the following formula: If the amount If the number of extreme points and zero-crossing points in the entire photovoltaic output impact data is equal or differs by at most one, and the average value of the upper and lower envelopes determined by the cubic spline difference method is 0, then this component is Photovoltaic output impact data e If the first-order IMF of The calculation is repeated as input, as shown in the following formula: In the formula, is the intermediate process component, for The mean curve of the upper and lower envelopes obtained using cubic spline difference. It is worth noting that the abnormal data collected by the influencing factor system will interfere with the division of the upper and lower envelopes and affect the decomposition effect. Therefore, it is still necessary to use the abnormal data processing method proposed in Chapter 2 to reconstruct the abnormal data. Assuming that the IMF requirements are met after k operations, the influencing factor information z e The first-order IMF component L1 and the residual influencing factor component of the separated first-order IMF They are: L1=h 1k After separating the first-order IMF, the original influencing factor x e It may also contain IMF components, so it is necessary to Continue the above operation until the remaining component is a monotonic function. Assuming that the requirements are met after n times, the separation process is: In the formula, r n is the nth-order residual component, which is a monotonic function. At this time, the original photovoltaic output affects the data Z e It can be expressed as several IMF components and the final residual component The sum is as follows:

4. A photovoltaic output prediction method according to claim 2, characterized in that: The ensemble empirical mode decomposition uses a noise-assisted method to suppress the modal aliasing caused by the presence of noise data in the influencing data, specifically: The original influencing factor z e Add Gaussian white noise with a mean of 0 to obtain the influencing factor information z after white noise processing v,e , to cover up the original noise in the influencing factor information as follows: With υ,e =from e +υ Information on influencing factors after white noise processing x υ,e Perform EEMD decomposition, and z υ,e Decomposed into the following formula: In the formula, v i is the white noise added for the i-th time, and then the IMF component of each white noise added can be obtained Considering that the mean value of Gaussian noise is 0, the above IMF components can be averaged to obtain the final IMF components of EEMD decomposition as follows: In the formula, m is the number of times white noise is added, and generally m is less than 100. After adding m times of Gaussian white noise, the statistical rules are as follows: Where ΔE m is the influencing factor information υ,e and the difference between the final IMF component and; ε is the amplitude of white noise. The more times noise is added, the closer the IMF component is to the original decomposition when the amplitude of white noise is constant, and the final decomposition result of influencing factor information is obtained.

5. A photovoltaic output prediction method according to claim 4, characterized in that: The method further comprises: To calculate the sample entropy of the influencing factor components, firstly, the influencing factor information IMF components are combined into a vector sequence as follows: L n (i)=[L υ,e (i),L υ,e (i+1),...,L υ,e (i+m-1)] Definition L υ,e (i),L υ,e The distance between (i+1) is the absolute value of the maximum difference between the corresponding elements of each IMF component as follows: d[L n (i),L n (j)]=max|L υ,e (i+k)-L υ,e (j+k)|, where 1≤k≤n-1; 1≤i, j≤m-n+1, ​​i≠j. For a given L n (i) Calculate L n (i) With L n (j) The distance between n (i),L n (j)] is less than the ratio of the number of j components to the total components, and is recorded as B i,n (r) is as follows, Calculate B i,n (r) All mean values ​​for i: For the n+1 point vector, the ratio and average value are calculated as follows: The obtained sequence sample entropy is: When m takes a finite value: SampEn(n,r,m)=-ln[A n (r) / B n (r)], where n is the embedding dimension and r is the similarity tolerance.

6. A photovoltaic output prediction method according to claim 1, characterized in that: The nonlinear mapping relationship between the photovoltaic output influencing factors and photovoltaic output data is analyzed based on the ridge regression method, specifically: The mapping relationship is established based on the following formula: t ′=h -1 (Z t ′), where Z t ′ is the factor vector of photovoltaic output influencing factors with non-Gaussian distribution; X t ′ is the corresponding prediction result; h -1 is a nonlinear mapping function from the photovoltaic output influencing factors to the photovoltaic output data; For the ordinary linear regression model, we have: k =w k Z k , where X k is the predicted value of photovoltaic output data, w k is the weight parameter, Z k is the information matrix of factors affecting the photovoltaic output of the node, and the weight parameter w is obtained by the least squares method. k When the objective function of the least squares linear regression is set to: J(w) = ∥Xw k Z k ∥2, where X is the real value of photovoltaic output data information. Considering the multicollinearity of the influencing factors in the distribution network, the weight parameter w k The value of is too large, resulting in large fluctuations in the prediction results of photovoltaic output data. Therefore, in the ridge regression method, a penalty function is used to constrain the least squares model. Specifically, the l2 norm is used, and the objective function of the ridge regression model becomes: J(w) = ∥X k -w k Z k ∥2+λ||w k ||2, where λ is added to balance the regularization term and the loss function. To minimize the loss function J(w), the weight parameter is derived and taken as 0, as shown in the following formula; is equivalent to: After simplifying by taking 0, we can get the parameter w k The optimal solution is as follows: Where I is the unit matrix and the optimal solution is is the positive definite sum of a semi-positive definite matrix and a diagonal matrix; Optimize the parameters To transform, first use the matrix inversion theorem: (U -1 +V T W -1 V) -1 V T W -1 =UV T (VUV T +W) -1 In the formula V=Z k ,W=I, which is equivalent to: The weight parameter becomes: Define a nonlinear mapping function φ(·) and transform the ridge regression model objective function into: The optimal solution is: The kernel function K is defined as follows: K(S i ,S j )=[φ(S i ),φ(S j )]=φ(S i )φ T (S j ) make: K=K(Z k ,WITH k )=φ(Z k )φ(Z k ) T α=(φ(Z k )φ(Z k ) T +λI) -1 X The intermediate variable α becomes: α=(K+λI) -1 X k ; The weight parameters after kernel function processing become: Then the photovoltaic output forecast value is obtained as follows:

7. A photovoltaic output prediction method according to claim 6, characterized in that: The method further includes: optimizing the gap between the predicted result and the actual result based on the KRRSE learning method, wherein the KRRSE photovoltaic output prediction model is assumed to be: In the formula, is a kernel ridge regression model, which is used to learn the mapping relationship between historical influencing factor data z and photovoltaic output data x, where: The data set used is D k ={(z1,x1),(z2,x2)...(z t-1 ,x t-1 )}, To approximate unknown prediction residuals; Learning to Predict Using Kernel Ridge Regression With the true prediction model h -1 The nonlinear relationship between the prediction residual error information is shown in Figure 2. At this time, the photovoltaic output prediction model is expressed as: In the formula, z t-1 Represents the input of factors affecting photovoltaic output, Represents the error between the KRRSE model and the true model, model noise 8. A photovoltaic output prediction system, characterized in that: The system comprises: A data collection unit is used to collect relevant influencing data of the operation of the photovoltaic power station, and use the historical operation data to establish a mapping model between the photovoltaic output influencing data and the photovoltaic output at the corresponding moment; A prediction calculation unit, used to collect real-time photovoltaic output data and input it into a pre-built and trained photovoltaic output prediction model for prediction; wherein the photovoltaic output prediction model is associated with an enhanced kernel ridge regression model of ensemble empirical mode decomposition; The data output unit is used to output the predicted photovoltaic processing data.

9. A photovoltaic output prediction device, characterized in that: The device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can complete the photovoltaic output prediction method described in any one of claims 1-7.

10. A non-volatile computer storage medium for photovoltaic output prediction, storing computer executable instructions, characterized in that: The computer executable instructions are executed by a processor to implement a photovoltaic output prediction method as described in any one of claims 1-7.

Images