Distributed photovoltaic output prediction method and system

Abstract

The invention relates to the technical field of photovoltaic output prediction, in particular to a distributed photovoltaic output prediction method and system.The distributed photovoltaic output prediction method comprises the steps that historical time sequence data and historical photovoltaic power station output data of a distributed photovoltaic power station are collected and preprocessed, and the historical time sequence data comprise the solar irradiance, the temperature and humidity and the photovoltaic panel angle; establishing a photovoltaic output LSTM prediction model; and training the photovoltaic output LSTM prediction model by using the preprocessed historical time sequence data and historical photovoltaic power station output data, performing parameter optimization, and inputting the time sequence data acquired in real time to perform output prediction of the distributed photovoltaic power station. According to the invention, the precision of distributed photovoltaic output prediction is improved by processing historical data and optimizing the parameters of the photovoltaic output LSTM prediction model. According to the invention, not only weather-related data such as solar irradiance, temperature and humidity are considered, but also the angle of the photovoltaic panel is considered during distributed photovoltaic output, so that the accuracy of distributed photovoltaic output prediction is further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic output prediction, and particularly relates to a distributed photovoltaic output prediction method and system. Background Art

[0002] In recent years, with the development of technology, the solar energy-related industries have developed rapidly. The advantages of solar energy being green, clean, and renewable make it one of the new energy sources with the most promising application prospects. From the development trend of energy utilization, photovoltaic power generation will become an important source of electric energy. The distributed photovoltaic output has typical probability distribution characteristics, which are important basic data for the optimal planning and design, output prediction, and optimal dispatching management of photovoltaic power stations. However, the photovoltaic output is affected by the solar irradiance cycle and the random changes of various meteorological factors, and has obvious characteristics of changing with time. The quality of meteorological data plays an important role in photovoltaic power generation prediction.

[0003] The proportion of photovoltaic power generation in the power system increases year by year, and the difficulty of power grid dispatching also gradually increases. The predicted value of the output of a photovoltaic power station will be used as the basis for power grid dispatching. Therefore, accurately predicting the power generation power of a photovoltaic power station can timely adjust the power grid dispatching plan and further ensure the power quality of the power grid.

[0004] In view of this, it is necessary to study a distributed photovoltaic output prediction method and system with high prediction accuracy. Summary of the Invention

[0005] Aiming at the problems in the prior art, the present invention provides a distributed photovoltaic output prediction method and system, and the specific technical solutions are as follows:

[0006] On the one hand, the present invention provides a distributed photovoltaic output prediction method, including the following steps:

[0007] Step S1, collecting historical time-series data and historical photovoltaic power station output data of a distributed photovoltaic power station, where the historical time-series data includes solar irradiance, temperature and humidity, and the angle of photovoltaic panels;

[0008] Step S2, preprocessing the historical time-series data and historical photovoltaic power station output data of the distributed photovoltaic power station to obtain the preprocessed historical time-series data and historical photovoltaic power station output data;

[0009] Step S3, establishing a photovoltaic output LSTM prediction model;

[0010] Step S4, training the photovoltaic output LSTM prediction model with the preprocessed historical time-series data and historical photovoltaic power station output data, and optimizing the parameters of the photovoltaic output LSTM prediction model to obtain an optimized photovoltaic output LSTM prediction model;

[0011] Step S5: Input the time series data of solar irradiance, temperature, humidity, and photovoltaic panel angle collected in real time, and use the optimized LSTM prediction model for photovoltaic output to make a prediction to obtain the output of the distributed photovoltaic power station.

[0012] Preferably, the preprocessing of the historical time series data of the distributed photovoltaic power station in step S2 includes the following steps:

[0013] Step S21: Eliminate outliers from the historical time series data using the 3σ principle to obtain the historical time series data after elimination.

[0014] Step S22: Perform linear interpolation on the missing values of the historical time series data after elimination to obtain complete historical time series data.

[0015] Step S23: Denoise the complete historical time series data to obtain the historical time series data after denoising.

[0016] Step S24: Normalize the historical time series data after denoising.

[0017] Preferably, the following wavelet denoising methods are specifically used to denoise the solar irradiance, temperature, humidity, and photovoltaic panel angle in the historical time series data:

[0018] 1) Collect the original signal of the corresponding historical time series data and determine the number of layers to be decomposed for the historical time series data.

[0019] 2) Use the wavelet threshold function for calculation to construct the scaling function.

[0020] 3) Decompose the original signal of the corresponding historical time series data using the real part tree and the imaginary part tree according to the number of decomposition layers corresponding to the historical time series data and the constructed scaling function.

[0021] 4) Set the high-frequency coefficients to be filtered, and use the threshold function to perform denoising suppression on the high-frequency coefficients to obtain the estimated value of the wavelet coefficient matrix after denoising suppression.

[0022] 5) Judge whether the denoising is completed. If not, go back to step 2. If completed, output the original signal of the corresponding historical time series data after denoising suppression.

[0023] Preferably, the threshold function is:

[0024]

[0025] where x is the input value, and λ j is the selected threshold at scale j.

[0026] Preferably, the following method is specifically used to denoise the historical photovoltaic power station output data:

[0027] 1) Perform VMD decomposition on the historical output data of the photovoltaic power station to obtain each IMF component;

[0028] 2) Calculate the energy E of each IMF component si and the energy E of each component after VMD decomposition of the noise component ni , and define D i as the energy proportion coefficient of the useful signal component of each component;

[0029] 3) When D i > 0.65, IMF i is the signal component, otherwise it is the noise component;

[0030] 4) Perform wavelet denoising on the obtained signal components and mean filtering denoising on the noise components;

[0031] 5) Reconstruct the denoised signal components and noise components to obtain the denoised historical output data of the photovoltaic power station.

[0032] Preferably, the calculation method of the energy proportion coefficient D of the useful signal component of each component i is as follows:

[0033]

[0034] Preferably, in step S4, the parameter optimization of the photovoltaic output LSTM prediction model is specifically to optimize the number of cell units, the initial learning rate, and the L2 regularization parameter of the photovoltaic output LSTM prediction model by using the Bayesian optimization method.

[0035] On the other hand, the present invention provides a distributed photovoltaic output prediction system, applying the method described above, including: a data acquisition module for collecting historical time series data and historical output data of the distributed photovoltaic power station, and the historical time series data includes solar irradiance, temperature and humidity, and the angle of the photovoltaic panel;

[0036] A data processing module for preprocessing the historical time series data and historical output data of the distributed photovoltaic power station to obtain the preprocessed historical time series data and historical output data of the distributed photovoltaic power station;

[0037] A model establishment module for establishing a photovoltaic output LSTM prediction model;

[0038] A model training and optimization module for training the photovoltaic output LSTM prediction model by using the preprocessed historical time series data and historical output data of the distributed photovoltaic power station, and optimizing the parameters of the photovoltaic output LSTM prediction model to obtain an optimized photovoltaic output LSTM prediction model;

[0039] The output prediction module inputs the time series data of solar irradiance, temperature and humidity, and photovoltaic panel angle collected in real time, and uses the optimized photovoltaic output LSTM prediction model for prediction to obtain the output of the distributed photovoltaic power station.

[0040] On the other hand, the present invention provides a computer-readable storage medium, which includes a stored program. When the program runs, it controls the device where the computer-readable storage medium is located to execute the distributed photovoltaic output prediction method described above.

[0041] Finally, the present invention provides a processor, which is used to run a program. When the program runs, it executes the distributed photovoltaic output prediction method described above.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0043] The present invention provides a distributed photovoltaic output prediction method, which includes collecting historical time series data and historical photovoltaic power station output data of a distributed photovoltaic power station, and performing preprocessing to establish a photovoltaic output LSTM prediction model; training the model, optimizing parameters, and inputting the time series data collected in real time for the output prediction of the distributed photovoltaic power station. By processing historical data and optimizing the parameters of the photovoltaic output LSTM prediction model, the accuracy of distributed photovoltaic output prediction is improved.

[0044] When the present invention predicts the distributed photovoltaic output, it not only considers weather-related data such as solar irradiance, temperature and humidity, but also considers the photovoltaic panel angle, further improving the accuracy of distributed photovoltaic output prediction.

[0045] The present invention uses wavelet denoising to denoise data such as solar irradiance, temperature and humidity, and photovoltaic panel angle, and uses VMD decomposition, wavelet denoising and mean filtering methods to process the historical photovoltaic power station output data, reducing the prediction error caused by inaccurate data and improving the accuracy of distributed photovoltaic output prediction. Description of the Drawings

[0046] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts do not necessarily draw according to the actual scale.

[0047] Figure 1 It is the method flow chart of the present invention.

[0048] Figure 2 It is the system schematic diagram of the present invention. Detailed Embodiments

[0049] Next, in combination with the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0050] It should be understood that when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0051] It should also be understood that the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms.

[0052] It should be further understood that the term " / and" used in the specification of the present invention and the appended claims refers to any combination and all possible combinations of one or more of the related listed items, and includes these combinations.

[0053] As Figure 1 shown, this embodiment provides a distributed photovoltaic output prediction method, including the following steps:

[0054] Step S1, collect historical time-series data and historical photovoltaic power plant output data of the distributed photovoltaic power plant. The historical time-series data includes solar irradiance, temperature and humidity, and photovoltaic panel angle.

[0055] Step S2, preprocess the historical time-series data and historical photovoltaic power plant output data of the distributed photovoltaic power plant to obtain the preprocessed historical time-series data and historical photovoltaic power plant output data.

[0056] The preprocessing of the historical time-series data of the distributed photovoltaic power plant in Step S2 includes the following steps:

[0057] Step S21, eliminate outliers from the historical time-series data using the 3σ principle to obtain the historical time-series data after outlier elimination. The probability that the numerical distribution is within (μ - 3σ, μ + 3σ) is 0.9973: almost all data values (99.73%) will fall within this range. Data outside this range is considered an outlier, with a probability of only 0.27%. Therefore, the accuracy of eliminating outliers using the 3σ principle is relatively high. Here, μ is the mean of the historical time-series data, which is the mean of solar irradiance when the historical time-series data is solar irradiance, and so on.

[0058] Step S22, perform linear interpolation on the missing values of the historical time-series data after outlier elimination to obtain complete historical time-series data. Given points (t 0 , x 0 ) and (t 1 , x 1 ), and wishing to estimate the value of the target x at a certain time t between these two points, where x represents the historical time-series data, then:

[0059]

[0060] Step S23, denoise the complete historical time-series data to obtain the denoised historical time-series data.

[0061] Step S24, normalize the denoised historical time-series data. The normalization uses the maximum-minimum normalization formula, which is as follows:

[0062]

[0063] where x* represents the normalized historical time-series data, x min represents the minimum value of the corresponding historical time-series data, and x max represents the maximum value of the corresponding historical time-series data.

[0064] In step S23, the following wavelet denoising methods are specifically used to denoise the solar irradiance, temperature and humidity, and photovoltaic panel angle in the historical time-series data:

[0065] Step S231, collect the original signal of the corresponding historical time-series data and determine the number of decomposition levels corresponding to the historical time-series data. Step S232, use the wavelet threshold function for calculation to construct the scaling function.

[0066] Step S233, according to the decomposition level corresponding to the historical time-series data and the constructed scaling function, decompose the original signal of the corresponding historical time-series data using the real part tree and the imaginary part tree.

[0067] Step S234, set the high-frequency coefficients to be filtered, and use the threshold function to perform denoising suppression on the high-frequency coefficients to obtain the estimated value of the wavelet coefficient matrix after denoising suppression;

[0068] Since only selecting the hard threshold or the soft threshold will cause problems such as the pseudo-Gibbs phenomenon and edge blurring distortion, the threshold function is improved. By combining the advantages of the hard threshold function and the soft threshold function, the threshold function is improved. The threshold function is as follows:

[0069]

[0070] where x is the input value, and λ j is the threshold selected at scale j. This threshold ensures that when there are more useful components in the wavelet coefficients of the j-th layer, the value of λ j will be smaller, and vice versa, so that the selection of the threshold is more in line with the actual noise distribution.

[0071]

[0072] σ j =(median|d j,k |) / 0.6745;

[0073]

[0074] where σ j is the noise intensity at scale j, N is the signal length, median d j, k| represents the median of the wavelet coefficients at scale j, L j represents the length of the wavelet coefficients at scale j, and k is the signal displacement.

[0075] Step S235, determine whether the denoising is completed. If not, go to step S232. If completed, output the original signal of the corresponding historical time series data after denoising suppression.

[0076] The historical photovoltaic power plant output data is denoised in the following specific way:

[0077] 1) Perform VMD decomposition on the historical photovoltaic power plant output data to obtain each IMF component.

[0078] 2) Calculate the energy E si of each IMF component and the energy E ni of the noise components after VMD decomposition of each component. Define D i as the energy ratio coefficient of the useful signal components of each component. According to the spectral characteristics of each component obtained after signal decomposition, some components are mainly composed of noise components, called noise components, and some components are mainly composed of useful signal components, called signal components. The calculation method of the energy ratio coefficient D i of the useful signal components of each component is as follows:

[0079]

[0080] Among them, IMF i,k represents the k-th sample of the i-th IMF component IMF i ; and ni,k represents the k-th sample of the noise component of the i-th IMF component IMF i .

[0081] 3) When D i > 0.65, IMF i is a signal component; otherwise, it is a noise component.

[0082] 4) Perform wavelet noise reduction on the obtained signal components and perform mean filter noise reduction on the noise components.

[0083] 5) Reconstruct the noise-reduced signal components and noise components to obtain the historical photovoltaic power plant output data after noise reduction.

[0084] Step S3: Establish a photovoltaic output LSTM prediction model.

[0085] Step S4: Use the preprocessed historical time series data and historical photovoltaic power plant output data to train the photovoltaic output LSTM prediction model, and optimize the parameters of the photovoltaic output LSTM prediction model to obtain an optimized photovoltaic output LSTM prediction model.

[0086] Specifically, in step S4, the parameter optimization of the photovoltaic output LSTM prediction model is to optimize the number of cell units, initial learning rate, and L2 regularization parameter of the photovoltaic output LSTM prediction model by using the Bayesian optimization method.

[0087] Changing the number of neurons affects the training result of the photovoltaic output LSTM prediction model. If the number of neurons is too small, the learning of data is insufficient, and it does not have effective learning and understanding capabilities. If the number of neurons is too large, the network will fall into local optimum, the network learning speed will slow down, and the generalization ability of the model will be reduced. In this embodiment, the Bayesian optimization method is used to optimize the number of cell units of the photovoltaic output LSTM prediction model, which can enable the photovoltaic output LSTM prediction model to quickly reach the optimum.

[0088] The learning rate is an important parameter that affects the training process of the entire neural network. If the learning rate is small, the model training speed is slow; if the learning rate is large, problems are likely to occur during model training. Therefore, it is necessary to find a balance between the training speed and the training accuracy of the model. In this embodiment, the Bayesian optimization method is used to optimize the learning rate of the photovoltaic output LSTM prediction model, which can enable the photovoltaic output LSTM prediction model to quickly reach the optimum and reduce the probability of problems.

[0089] During the entire model training process, when the prediction accuracy is greater than or equal to the preset value, the training stops. When the prediction accuracy does not reach the preset value and the set maximum number of iterations has been reached, the photovoltaic output LSTM prediction model is adjusted and training continues.

[0090] In this embodiment, the mean absolute error of the prediction is used to characterize the prediction accuracy. When the mean absolute error of the prediction is less than or equal to the preset value, it means that the prediction accuracy meets the requirements.

[0091] The process of the Bayesian optimization method is as follows:

[0092] Gaussian process regression: The Gaussian process is the most commonly used probability model in the application of Bayesian optimization. As a general generalization of the multivariate Gaussian probability distribution, the Gaussian process is mainly composed of a mean function and a positive semi - definite covariance function, that is:

[0093] y = gp[n(x), k(x, x')];

[0094] Among them, n(x) is the mean function, and k(x, x') is the covariance function. x is the training sample of the Gaussian process, and the measure of x is distributed in a finite - dimensional manner in the Gaussian process. When using the Gaussian process to fit the discrete data in the sample, n(x) is generally set to 0, and k(x, x') generally uses the Matern covariance function.

[0095] Acquisition function: The expected improvement is the most commonly used acquisition function. The function takes the point that can maximize the current maximum value f(x max ) as the next point. Its expression is as follows:

[0096]

[0097] Among them, f(x) is the objective function value, is the distribution function, μ(x) and σ(x) are the mean and variance of the Gaussian process regression probability model respectively, and α is a hyperparameter to explore the maximum value space. α = 0 makes the value converge to f(x max ), avoiding local optima.

[0098] The general steps for prediction using Bayesian - optimized LSTM are as follows: 1) Set the range of parameters to be optimized, and the optimization objective is RMSE; 2) Use Bayesian method to optimize the hidden layer and initial learning rate of LSTM; 3) Return the minimum value of the optimized RMSE and the corresponding hyperparameters of the LSTM model; 4) Use the hyperparameters of the Bayesian - optimized LSTM model as the parameters of the final LSTM model.

[0099] In step S5, input the time - series data of the real - time collected solar irradiance, temperature and humidity, and photovoltaic panel angle, and use the optimized photovoltaic output LSTM prediction model for prediction to obtain the output of the distributed photovoltaic power station.

[0100] Embodiment 2:

[0101] As Figure 2 shown, based on the same inventive concept as Embodiment 1, this embodiment provides a distributed photovoltaic output prediction system, which applies the method described above and includes:

[0102] A data acquisition module, which is used to acquire the historical time-series data and historical photovoltaic power station output data of the distributed photovoltaic power station. The historical time-series data includes solar irradiance, temperature and humidity, and the angle of the photovoltaic panel;

[0103] A data processing module, which is used to preprocess the historical time-series data and historical photovoltaic power station output data of the distributed photovoltaic power station to obtain the preprocessed historical time-series data and historical photovoltaic power station output data;

[0104] A model establishment module, which is used to establish a photovoltaic output LSTM prediction model;

[0105] A model training and optimization module, which uses the preprocessed historical time-series data and historical photovoltaic power station output data to train the photovoltaic output LSTM prediction model and optimize the parameters of the photovoltaic output LSTM prediction model to obtain an optimized photovoltaic output LSTM prediction model;

[0106] An output prediction module, which inputs the time-series data of solar irradiance, temperature and humidity, and the angle of the photovoltaic panel collected in real time, and uses the optimized photovoltaic output LSTM prediction model to make a prediction to obtain the output of the distributed photovoltaic power station.

[0107] Embodiment 3:

[0108] Based on the same inventive concept as Embodiment 1, this embodiment provides a computer-readable storage medium, which includes a stored program. When the program runs, it controls the device where the computer-readable storage medium is located to execute the described distributed photovoltaic output prediction method.

[0109] Embodiment 4:

[0110] Based on the same inventive concept as Embodiment 1, this embodiment provides a processor, which is used to run a program. When the program runs, it executes the described distributed photovoltaic output prediction method.

[0111] Those of ordinary skill in the art can realize that the modules of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition of each example has been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present invention.

[0112] In the embodiments provided by the present invention, it should be understood that the division of modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple modules can be combined into one module, one module can be split into multiple modules, or some features can be ignored, etc.

[0113] In addition, the functional modules in each embodiment of the present invention can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module. The above-mentioned integrated modules can be implemented in the form of hardware or in the form of software functional modules.

[0114] If the above-mentioned integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, read-only memories (ROMs), random access memories (RAMs), external hard drives, magnetic disks, or optical discs that can store program codes.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of each embodiment of the present invention, and they should all be covered by the scope of the claims and the description of the present invention.

Claims

1. A distributed photovoltaic output prediction method, characterized in that: The following steps are involved: Step S1, collecting historical time series data and historical photovoltaic power station output data of distributed photovoltaic power stations, the historical time series data including solar irradiance, temperature and humidity, and photovoltaic panel angle; Step S2, preprocessing the historical time series data of the distributed photovoltaic power station and the historical photovoltaic power station output data to obtain the preprocessed historical time series data and the historical photovoltaic power station output data; Step S3, establishing a photovoltaic output LSTM prediction model; Step S4, using the preprocessed historical time series data and the historical photovoltaic power station output data to train the photovoltaic output LSTM prediction model, and optimizing the parameters of the photovoltaic output LSTM prediction model to obtain an optimized photovoltaic output LSTM prediction model; Step S5, input the real-time collected solar irradiance, temperature and humidity, and photovoltaic panel angle time series data, use the optimized photovoltaic output LSTM prediction model to make predictions, and obtain the output of the distributed photovoltaic power station.

2. A distributed photovoltaic output prediction method according to claim 1, characterized in that: The preprocessing of the historical time series data of the distributed photovoltaic power station in step S2 includes the following steps: Step S21, using the 3σ principle to remove abnormal values ​​from the historical time series data to obtain the historical time series data after removal; Step S22, performing linear interpolation on the missing values ​​of the historical time series data after elimination to obtain complete historical time series data; Step S23, denoising the complete historical time series data to obtain denoised historical time series data; Step S24, normalizing the historical time series data after noise reduction.

3. A distributed photovoltaic output prediction method according to claim 2, characterized in that: The following wavelet denoising methods are used to denoise the solar irradiance, temperature and humidity, and photovoltaic panel angle in the historical time series data: 1) Collect the original signal of the corresponding historical time series data and determine the number of layers that need to be decomposed corresponding to the historical time series data; 2) Use the wavelet threshold function to calculate and construct the scaling function; 3) According to the number of decomposition layers corresponding to the historical time series data and the constructed scaling function, the real part tree and the imaginary part tree are used to decompose the original signal of the corresponding historical time series data; 4) Setting the high-frequency coefficients to be filtered, using the threshold function to perform denoising on the high-frequency coefficients, and obtaining an estimate of the wavelet coefficient matrix after denoising; 5) Determine whether denoising is completed. If not, go to step 2. If completed, output the original signal of the corresponding historical time series data after denoising.

4. A distributed photovoltaic output prediction method according to claim 3, characterized in that: The threshold function is: Among them, x is the input value, λ j is the threshold selected on scale j.

5. A distributed photovoltaic output prediction method according to claim 2, characterized in that: The following methods are used to reduce the noise of historical photovoltaic power station output data: 1) Perform VMD decomposition on historical PV power station output data to obtain various IMF components; 2) Calculate the energy E of each IMF component si The energy E of each component after VMD decomposition of the noise component ni , define D i is the energy proportion coefficient of each useful signal component; 3) When D i >0.65, IMF i is the signal component, otherwise it is the noise component; 4) Perform wavelet denoising on the obtained signal component and mean filter denoising on the noise component; 5) Reconstruct the signal component and noise component after noise reduction to obtain the historical photovoltaic power station output data after noise reduction.

6. A distributed photovoltaic output prediction method according to claim 5, characterized in that: Energy ratio of each useful signal component D i The calculation method is as follows:

7. A distributed photovoltaic output prediction method according to claim 1, characterized in that: In the step S4, the parameter optimization of the photovoltaic output LSTM prediction model is specifically to optimize the number of cell units, the initial learning rate, and the L2 regularization parameter of the photovoltaic output LSTM prediction model using the Bayesian optimization method.

8. A distributed photovoltaic output prediction system, characterized in that: The method described in any one of claims 1 to 7 includes: a data acquisition module for collecting historical time series data of distributed photovoltaic power stations and historical photovoltaic power station output data, the historical time series data including solar irradiance, temperature and humidity, and photovoltaic panel angle; A data processing module is used to preprocess the historical time series data of the distributed photovoltaic power station and the historical photovoltaic power station output data to obtain the preprocessed historical time series data and the historical photovoltaic power station output data; Model building module, used to build LSTM prediction model for photovoltaic output; The model training optimization module uses the preprocessed historical time series data and historical photovoltaic power station output data to train the photovoltaic output LSTM prediction model, and optimizes the parameters of the photovoltaic output LSTM prediction model to obtain the optimized photovoltaic output LSTM prediction model; The output prediction module inputs the real-time collected solar irradiance, temperature and humidity, and photovoltaic panel angle time series data, and uses the optimized photovoltaic output LSTM prediction model to predict and obtain the output of the distributed photovoltaic power station.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein: When the program is running, the device where the computer-readable storage medium is located is controlled to execute a distributed photovoltaic output prediction method as described in any one of claims 1 to 7.

10. A processor, characterized in that: The processor is used to run a program, wherein the program, when running, executes a distributed photovoltaic output prediction method as described in any one of claims 1 to 7.

Images