Photoelectric output prediction method and system based on data center

Abstract

The invention discloses a photoelectric output prediction method and system based on a data center, and relates to the field of photoelectric output prediction, and the method comprises the steps: obtaining an illumination prediction model based on the training of historical illumination data; dividing historical illumination data according to seasons, and fitting seasonal factors to correct the illumination prediction model; training based on the photovoltaic power generation output data to obtain a photovoltaic power generation output prediction model; predicting illumination duration and intensity by using an illumination prediction model, and calculating first photovoltaic output data based on the predicted illumination duration and intensity; predicting photovoltaic output data at the next moment by using the photovoltaic power generation output prediction model, and recording the photovoltaic output data as second photovoltaic output data; and fitting the first photovoltaic output data and the second photovoltaic output data to obtain third photovoltaic output data, and taking the third photovoltaic output data as a final prediction result. According to the method, the power grid load and the power dispatching strategy can be optimized by accurately predicting the photoelectric output, so that the light abandoning phenomenon caused by power grid capacity limitation is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the field of photovoltaic output prediction, and in particular to a photovoltaic output prediction method and system based on a data center. Background Art

[0002] Abandonment refers to the phenomenon that during the photovoltaic power generation process, due to insufficient grid load or grid scheduling restrictions, part of the generated electricity cannot be received by the grid, thus being forced to stop generating electricity or reduce the power generation capacity.

[0003] The reasons for abandoned solar power include: limited grid acceptance capacity: during peak sunlight periods, photovoltaic output may be far greater than the grid acceptance capacity or the actual local electricity demand.

[0004] Inadequate power dispatch and management: The grid dispatch system fails to effectively predict and utilize renewable energy, resulting in the inability to maximize the use of photovoltaic power generation.

[0005] Abandoning solar power will result in the power generation efficiency of photovoltaic power stations not being fully utilized, resulting in direct economic losses. Renewable energy is not fully utilized, which runs counter to the goal of energy conservation and emission reduction. Frequent adjustment of photovoltaic power generation output may bring challenges to the stability and security of the power grid.

[0006] Therefore, photovoltaic output prediction is very necessary. By accurately predicting photovoltaic output, the grid load and power dispatching strategy can be adjusted in advance, the coordination between photovoltaic power stations and the grid can be optimized, and the phenomenon of abandoned light can be reduced.

[0007] Ensure efficient use of photovoltaic power generation, improve the economic return of photovoltaic power stations, and reduce losses caused by abandoned light. Accurate photovoltaic output forecasts can help grid operators better dispatch electricity and ensure the stability and security of the grid.

[0008] In order to solve the above-mentioned problem of abandoned photovoltaic power generation and improve the utilization rate of photovoltaic power generation, the present invention provides a photovoltaic output prediction method and system based on a data center. The system can integrate environmental data, photovoltaic equipment status and historical power data, and use advanced prediction models to provide accurate photovoltaic output prediction for power grid operations, thereby effectively reducing the phenomenon of abandoned photovoltaic power generation and enhancing the economy and stability of the power grid. Summary of the invention

[0009] In view of the above-mentioned problems, the present invention is proposed.

[0010] Therefore, the problem to be solved by the present invention is: how to predict photovoltaic output more accurately.

[0011] To solve the above technical problems, the present invention provides the following technical solutions: a photovoltaic output prediction method based on a data center, which includes: the data center obtains historical light data, loads the historical light data into a neural network algorithm for training, and obtains a light prediction model; divides the historical light data by season, fits seasonal factors, and corrects the light prediction model based on the seasonal factors; the data center obtains historical photovoltaic power generation output data, loads the historical photovoltaic power generation output data into a neural network algorithm for training, and obtains a photovoltaic power generation output prediction model; uses the light prediction model to predict the light duration and intensity at the next moment, and calculates the first photovoltaic output data based on the predicted light duration and intensity; uses the photovoltaic power generation output prediction model to predict the photovoltaic output data at the next moment, recorded as the second photovoltaic output data; fits the first photovoltaic output data with the second photovoltaic output data to obtain the third photovoltaic output data as the final prediction result.

[0012] As a preferred solution of the photovoltaic output prediction method based on a data center described in the present invention, the historical light data includes light duration and light intensity data of different periods in several historical years.

[0013] As a preferred solution of the photovoltaic output prediction method based on a data center described in the present invention, the loading of historical light data into the neural network algorithm for training includes, starting from the first day of each year, taking 30 days as a time period, obtaining the average daily light duration and intensity data within the 30-day time period, and training based on the daily light duration and intensity data to obtain the light prediction model represented as follows: Let X d,t represents the light intensity on the tth day of the dth year, Y d,t represents the duration of sunshine on the tth day of the dth year. For each 30-day period, the average sunshine intensity is calculated. and average sunshine duration The formula is expressed as,

[0014]

[0015]

[0016] Where i is the time period index, starting from 0; for the input sequence Design a self-attention layer to enhance the model’s understanding of the importance of different parts of the time series, with attention weight α ij The calculation formula is expressed as:

[0017]

[0018] e ij =v T tanh(W atth j +b att )

[0019] Among them, N represents the sequence length, h j Represented as the hidden state of the input sequence, v, W att , b att It is represented as learning parameters, and the superscript T represents transposition. After the input sequence is processed by the self-attention layer, it enters the bidirectional LSTM layer for feature extraction, and uses the fully connected layer and activation function to predict the light intensity and duration at the next moment. It is expressed as,

[0020]

[0021]

[0022] in, It represents the predicted light intensity at the next moment. It represents the predicted illumination duration at the next moment. They are respectively represented as the forward and backward final states of the bidirectional LSTM, Represented as a connection operation, W X , W Y Expressed as a weight matrix, b X , b Y Indicated as bias.

[0023] As a preferred solution of the photovoltaic output prediction method based on a data center described in the present invention, the seasonal factor includes processing the historical light number using Fourier series to establish a function S(t) expressing seasonal periodicity, and serving as a correction term of the light prediction model, the formula is expressed as follows:

[0024]

[0025]

[0026] Among them, M represents the number of terms in the series, A m , B m denoted as the coefficients of a Fourier series determined by fitting historical light data, T is the length of the annual seasonal cycle, x t It is expressed as the light intensity or light duration data of the tth day. If it is light intensity data, S(t) is the correction item of light intensity prediction. If it is light duration data, S(t) is the correction item of light duration prediction. The correction of the light prediction model based on seasonal factors includes: in the light prediction model, the predicted light intensity and duration The correction is made by multiplying the seasonal factor S(t), expressed as,

[0027]

[0028]

[0029] in, They are respectively represented as the predicted values ​​of the corrected light intensity and light duration.

[0030] As a preferred solution of the photovoltaic output prediction method based on a data center described in the present invention, wherein: the first photovoltaic output data includes calculating the first photovoltaic output data using the corrected predicted light intensity and duration data; the calculation step includes setting the photovoltaic efficiency parameter η according to the ambient temperature adj , expressed as,

[0031] η adj =η(1-α(T amb -T ref ))

[0032] Among them, η represents the conversion efficiency of the photovoltaic panel, α represents the temperature coefficient, which means the percentage of photovoltaic panel efficiency decrease for every unit increase in temperature, T amb Expressed as ambient temperature, T ref It is expressed as the reference temperature; the actual total radiation E received is calculated according to the tilt angle of the photovoltaic panel and is expressed as,

[0033]

[0034] Among them, θ represents the installation angle of the photovoltaic panel; finally, the first photovoltaic output data P1 is calculated according to the photovoltaic efficiency parameter and the total radiation, which is expressed as,

[0035] P1=η adj AE

[0036] Where A represents the area of ​​the photovoltaic panel.

[0037] As a preferred solution of the photovoltaic output prediction method based on a data center described in the present invention, the photovoltaic power generation output prediction model includes loading historical photovoltaic power generation output data into a neural network algorithm for training to obtain a photovoltaic power generation output prediction model; normalizing the historical photovoltaic power generation output data to represent it as follows:

[0038]

[0039] Among them, P norm Represented as normalized data, P tIt is represented by the photovoltaic power output data of the tth day, μ is represented by the average value of the historical photovoltaic power output data, and σ is represented by the standard deviation of the historical photovoltaic power output data; the normalized data is substituted into the single-layer LSTM to be expressed as,

[0040] h t =LSTM(P norm ,h t-1 )

[0041] The conversion from LSTM output to predicted value is expressed as,

[0042] P pred =ξ(W p ·h t +b p )

[0043] P pred Inverse normalization to obtain the predicted value of photovoltaic power generation output It is expressed as,

[0044]

[0045] Among them, h t is represented as the hidden state of the photovoltaic power output layer, ξ is represented as the activation function, W p Denotes the output layer weight, b p It is represented as the output layer bias; the moment to be predicted is substituted into the photovoltaic power generation output prediction model to obtain the second photovoltaic output data P2.

[0046] As a preferred solution of the photovoltaic output prediction method based on a data center described in the present invention, wherein: the fitting of the first photovoltaic output data with the second photovoltaic output data includes performing mean processing on the obtained P1 and P2 to obtain the third photovoltaic output data P3 expressed as,

[0047]

[0048] The third photovoltaic output data is set as the final photovoltaic output prediction value.

[0049] Another object of the present invention is to provide a photovoltaic output prediction system based on a data center, which predicts the output of photovoltaic power generation.

[0050] To solve the above technical problems, the present invention provides the following technical solutions: a system for a photovoltaic output prediction method based on a data center, comprising: an illumination prediction model construction module, a seasonal factor correction module, a power generation output prediction model construction module and a photovoltaic output prediction module; the illumination prediction model construction module obtains historical illumination data based on the data center, loads the historical illumination data into a neural network algorithm for training, and obtains an illumination prediction model; the seasonal factor correction module divides the historical illumination data according to seasons, fits seasonal factors, and corrects the illumination prediction model based on seasonal factors; the power generation output prediction model construction module obtains historical photovoltaic power generation output data based on the data center, loads the historical photovoltaic power generation output data into a neural network algorithm for training, and obtains a photovoltaic power generation output prediction model; the photovoltaic output prediction module predicts the illumination duration and intensity at the next moment by using the illumination prediction model, and calculates the first photovoltaic output data based on the predicted illumination duration and intensity; the photovoltaic power generation output prediction model is used to predict the photovoltaic output data at the next moment, recorded as the second photovoltaic output data; the first photovoltaic output data is fitted with the second photovoltaic output data to obtain the third photovoltaic output data as the final prediction result.

[0051] A computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the photovoltaic output prediction method based on a data center are implemented.

[0052] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a photovoltaic output prediction method based on a data center as described above.

[0053] The beneficial effects of the present invention are as follows: the present invention can help power grid operators optimize power grid load and power dispatching strategies by accurately predicting photovoltaic output, thereby effectively reducing the phenomenon of abandoned light due to power grid capacity limitations and improving the utilization efficiency of renewable energy.

[0054] The accurate prediction results provided by the present invention can help power grid managers to better dispatch and manage power and ensure the stability and safety of the power grid, which is particularly important under high load or unstable climate conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0056] Figure 1This is a flow chart of a photovoltaic output prediction method based on a data center in Example 1.

[0057] Figure 2 This is a module structure diagram of a photovoltaic output prediction system based on a data center in Example 3. DETAILED DESCRIPTION

[0058] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0059] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0060] Example 1, reference Figure 1 , which is the first embodiment of the present invention, and which provides a photovoltaic output prediction method based on a data center, including: Figure 1 As shown:

[0061] Step 1: The data center obtains historical lighting data, loads the historical lighting data into the neural network algorithm for training, and obtains a lighting prediction model.

[0062] Historical light data includes data on light duration and light intensity at different periods over several historical years.

[0063] Starting from the first day of each year, with 30 days as a time period, the average daily light duration and intensity data within the 30-day period are obtained. Based on the daily light duration and intensity data, the light prediction model is expressed as: d,t represents the light intensity on the tth day of the dth year, Y d,t represents the duration of sunlight on the tth day of the dth year. For each 30-day period, calculate the average light intensity X d,i and average light duration Y d,i The formula is expressed as,

[0064]

[0065]

[0066] Here, i is the time period index, starting from 0.

[0067] In order to improve the prediction accuracy of the model compared to the traditional neural network model, we use a bidirectional LSTM (Bi-LSTM) network combined with a self-attention mechanism. The self-attention mechanism can help the model better understand the correlation between each time point in the time series, while the bidirectional LSTM can capture both past and future information at the same time, thereby more accurately predicting light changes.

[0068] For the input sequence Design a self-attention layer to enhance the model’s understanding of the importance of different parts of the time series, with attention weight α ij The calculation formula is expressed as:

[0069]

[0070] e ij =v T tanh(W att h j +b att )

[0071] Among them, N represents the sequence length, h j Represented as the hidden state of the input sequence, v, W att , b att It is represented as a learning parameter and the superscript T represents the transpose.

[0072] After the input sequence is processed by the self-attention layer, it enters the bidirectional LSTM layer for feature extraction. The bidirectional LSTM can learn the dynamic characteristics of the time series from two directions and enhance the model's ability to capture time dependencies. The fully connected layer and activation function are used to predict the light intensity and duration at the next moment, expressed as:

[0073]

[0074]

[0075] in, It represents the predicted light intensity at the next moment. It represents the predicted illumination duration at the next moment. They are respectively represented as the forward and backward final states of the bidirectional LSTM, Represented as a connection operation, W X , W Y Expressed as a weight matrix, b X , b Y Indicated as bias.

[0076] Step 2: Divide the historical light data by season, fit the seasonal factors, and modify the light prediction model based on the seasonal factors.

[0077] The Fourier series is used to process the historical light data to establish a function S(t) that expresses the seasonal periodicity and serves as a correction term for the light prediction model. The formula is expressed as follows:

[0078]

[0079]

[0080] Where M is the number of terms in the series, usually 3 to 5 to capture the main seasonal components of annual light variation, and A m , B m denoted as the coefficients of the Fourier series, determined by fitting historical light data, and can be determined by minimizing the prediction error in the historical data, for example, using the least squares method; T is the length of the annual seasonal cycle, and for annual seasonal patterns, T = 365 (or 366, depending on the year); x t It is expressed as the light intensity or light duration data of the tth day. If it is light intensity data, S(t) is the correction term for light intensity prediction. If it is light duration data, S(t) is the correction term for light duration prediction. When dealing with data with significant upward or downward trends, A0 helps to adjust these trends and ensure that the other fluctuation terms of the Fourier series (represented by cosine and sine terms) focus on capturing periodic changes relative to this baseline.

[0081] The correction of the light prediction model based on seasonal factors includes: in the light prediction model, the predicted light intensity and duration The correction is made by multiplying the seasonal factor S(t), expressed as,

[0082]

[0083]

[0084] in, They are respectively expressed as the predicted values ​​of the corrected light intensity and light duration. By introducing periodic seasonal changes into light prediction, the present invention can effectively adjust the model output to make it closer to the actual situation. By selecting a suitable M and accurate estimation coefficient, the accuracy of the prediction can be significantly improved.

[0085] Step 3: The data center obtains historical photovoltaic power generation output data, loads the historical photovoltaic power generation output data into the neural network algorithm for training, and obtains a photovoltaic power generation output prediction model.

[0086] The historical photovoltaic power generation output data is loaded into the neural network algorithm for training to obtain the photovoltaic power generation output prediction model. The specific steps are as follows: The historical photovoltaic power generation output data is normalized and expressed as:

[0087]

[0088] Among them, P norm Represented as normalized data, P t It is represented by the photovoltaic power generation output data on the tth day, μ is represented by the average value of the historical photovoltaic power generation output data, and σ is represented by the standard deviation of the historical photovoltaic power generation output data.

[0089] Substitute the normalized data into a single-layer LSTM and express it as:

[0090] h t =LSTM(P norm ,h t-1 )

[0091] The conversion from LSTM output to predicted value is expressed as,

[0092] P pred =ξ(W p ·h t +b p )

[0093] P pred Inverse normalization to obtain the predicted value of photovoltaic power generation output It is expressed as,

[0094]

[0095] Among them, h t is represented as the hidden state of the photovoltaic power output layer, ξ is represented as the activation function, W p Denotes the output layer weight, b p Represented as output layer bias.

[0096] Step 4: Use the light prediction model to predict the light duration and intensity at the next moment, and calculate the first photovoltaic output data based on the predicted light duration and intensity.

[0097] Calculating the first photovoltaic output data using the corrected predicted light intensity and duration data;

[0098] The calculation steps include setting the photovoltaic efficiency parameter η according to the ambient temperature adj , expressed as,

[0099] η adj =η(1-α(T amb -T ref ))

[0100] Among them, η represents the conversion efficiency of the photovoltaic panel, α represents the temperature coefficient, which means the percentage of photovoltaic panel efficiency decrease for every unit increase in temperature, Tamb Expressed as ambient temperature, T ref It is indicated as reference temperature;

[0101] The actual total radiation E received is calculated according to the tilt angle of the photovoltaic panel and is expressed as:

[0102]

[0103] Where, θ represents the installation angle of the photovoltaic panel;

[0104] Finally, the first photovoltaic output data P1 is calculated based on the photovoltaic efficiency parameters and total radiation, expressed as:

[0105] P1=η adj AE

[0106] Where A represents the area of ​​the photovoltaic panel.

[0107] Step 5: Use the photovoltaic power generation output prediction model to predict the photovoltaic output data at the next moment, which is recorded as the second photovoltaic output data.

[0108] Substitute the time to be predicted into the photovoltaic power generation output prediction model to obtain the second photovoltaic output data P2.

[0109] Step 6: Fit the first photovoltaic output data with the second photovoltaic output data to obtain the third photovoltaic output data as the final prediction result.

[0110] The third photovoltaic output data P3 is obtained by averaging P1 and P2, which is expressed as:

[0111]

[0112] The third photovoltaic output data is set as the final photovoltaic output prediction value.

[0113] Embodiment 2, the second embodiment of the present invention, is different from the first embodiment in that: a photovoltaic output prediction method based on a data center also includes: in order to verify and illustrate the technical effects adopted in this method, this embodiment adopts a traditional technical solution and the method of the present invention for comparative testing, and compares the test results by means of scientific argumentation to verify the actual effect of this method.

[0114] The prediction model of the present invention and the existing prediction method are respectively loaded into the system for simulation experiments. The specific test objectives are: Comparison of prediction accuracy: Comparing the prediction accuracy of the present invention and the existing technology (such as a simple time series prediction model, such as ARIMA or a traditional linear regression model) on the same data set.

[0115] Effect of reducing the abandonment rate: Evaluate the change in the abandonment rate in actual grid operation by using the prediction model of the present invention compared with using the existing technology.

[0116] Response time and processing speed: The time required for the present invention and the prior art to process the same amount of data is tested to evaluate the processing speed and real-time response capability.

[0117] The obtained test data are shown in Table 1.

[0118] Table 1: Test data comparison table

[0119]

[0120] Forecast accuracy (MAPE): The mean absolute percentage error (MAPE) shows that the present invention is superior to the existing technology in forecast accuracy, with the error reduced by 50%, which means that the predicted photovoltaic output is closer to the actual value, which helps to manage the power grid and plan resources more effectively.

[0121] Average abandoned light rate: The present invention can significantly reduce the abandoned light rate from 5% to 2%, an improvement of 60%, indicating that the use of the present invention can more effectively utilize renewable energy and reduce waste.

[0122] Average response time: The processing speed of the present invention is faster than that of the existing technology, and the response time is only one-third of that of the existing technology, which is particularly important for power grid operations that require rapid adjustment and decision-making.

[0123] Example 3, reference Figure 2 , which is the third embodiment of the present invention, and is different from the previous two embodiments in that: a system of a photovoltaic output prediction method based on a data center includes an illumination prediction model construction module, a seasonal factor correction module, a power generation output prediction model construction module and a photovoltaic output prediction module; the illumination prediction model construction module obtains historical illumination data based on the data center, loads the historical illumination data into a neural network algorithm for training, and obtains an illumination prediction model; the seasonal factor correction module divides the historical illumination data according to seasons, fits seasonal factors, and corrects the illumination prediction model based on seasonal factors; the power generation output prediction model construction module obtains historical photovoltaic power generation output data based on the data center, loads the historical photovoltaic power generation output data into a neural network algorithm for training, and obtains a photovoltaic power generation output prediction model; the photovoltaic output prediction module predicts the illumination duration and intensity at the next moment by using the illumination prediction model, and calculates the first photovoltaic output data based on the predicted illumination duration and intensity; the photovoltaic power generation output prediction model is used to predict the photovoltaic output data at the next moment, which is recorded as the second photovoltaic output data; the first photovoltaic output data is fitted with the second photovoltaic output data to obtain the third photovoltaic output data as the final prediction result.

[0124] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0125] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0126] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0127] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0128] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A photovoltaic output prediction method based on a data center, characterized by: include, The data center obtains historical lighting data, loads the historical lighting data into the neural network algorithm for training, and obtains a lighting prediction model; Divide historical light data by season, fit seasonal factors, and modify the light prediction model based on seasonal factors; The data center obtains historical photovoltaic power generation output data, loads the historical photovoltaic power generation output data into the neural network algorithm for training, and obtains a photovoltaic power generation output prediction model; The light prediction model is used to predict the light duration and intensity at the next moment, and the first photovoltaic output data is calculated based on the predicted light duration and intensity; The photovoltaic power generation output prediction model is used to predict the photovoltaic output data at the next moment, which is recorded as the second photovoltaic output data; The first photovoltaic output data is fitted with the second photovoltaic output data to obtain the third photovoltaic output data as the final prediction result.

2. The photovoltaic output prediction method based on a data center according to claim 1, characterized in that: The historical light data includes light duration and light intensity data at different periods in several historical years.

3. The photovoltaic output prediction method based on a data center according to claim 2, characterized in that: The loading of historical light data into the neural network algorithm for training includes: starting from the first day of each year, taking 30 days as a time period, obtaining the average daily light duration and intensity data within the 30-day time period, and training based on the daily light duration and intensity data to obtain the light prediction model represented as: Let X d,t represents the light intensity on the tth day of the dth year, Y d,t represents the duration of sunshine on the tth day of the dth year. For each 30-day period, the average sunshine intensity is calculated. and average sunshine duration The formula is expressed as, Where i is the time period index, starting from 0; For the input sequence Design a self-attention layer to enhance the model’s understanding of the importance of different parts of the time series, with attention weight α ij The calculation formula is expressed as: e ij =v T tanh(W att h j +b att ) Where N is the sequence length, h j Represented as the hidden state of the input sequence, v, W att 、b att It is represented as a learning parameter, and the superscript T represents the transpose; After the input sequence is processed by the self-attention layer, it enters the bidirectional LSTM layer for feature extraction, and uses the fully connected layer and activation function to predict the light intensity and duration at the next moment, expressed as, in, It represents the predicted light intensity at the next moment. It represents the predicted illumination duration at the next moment. They are respectively represented as the forward and backward final states of the bidirectional LSTM, Represented as a connection operation, W X , W Y Expressed as a weight matrix, b X 、b Y Indicated as bias.

4. The photovoltaic output prediction method based on a data center according to claim 3, characterized in that: The seasonal factor includes processing the historical light data using Fourier series to establish a function S(t) that expresses seasonal periodicity and serves as a correction term for the light prediction model. The formula is: Among them, M represents the number of terms in the series, A m , B m denoted as the coefficients of a Fourier series determined by fitting historical light data, T is the length of the annual seasonal cycle, x t It is expressed as the light intensity or light duration data of the tth day. If it is light intensity data, S(t) is the correction term for light intensity prediction. If it is light duration data, S(t) is the correction term for light duration prediction. The modification of the illumination prediction model based on the seasonal factor includes: in the illumination prediction model, the predicted illumination intensity and duration The correction is made by multiplying the seasonal factor S(t), expressed as, in, They are respectively represented as the predicted values ​​of the corrected light intensity and light duration.

5. The photovoltaic output prediction method based on a data center according to claim 4, characterized in that: The first photovoltaic output data includes calculating the first photovoltaic output data using the corrected predicted light intensity and duration data; The calculation steps include setting the photovoltaic efficiency parameter η according to the ambient temperature adj , expressed as, or adj =η(1-α(T amb -T ref )) Among them, η represents the conversion efficiency of the photovoltaic panel, α represents the temperature coefficient, which means the percentage of photovoltaic panel efficiency decrease for every unit increase in temperature, T amb Expressed as ambient temperature, T ref It is indicated as reference temperature; The actual total radiation E received is calculated according to the tilt angle of the photovoltaic panel and is expressed as: Where, θ represents the installation angle of the photovoltaic panel; Finally, the first photovoltaic output data P1 is calculated based on the photovoltaic efficiency parameters and total radiation, expressed as: P1=η adj AE Where A represents the area of ​​the photovoltaic panel.

6. The photovoltaic output prediction method based on a data center according to claim 5, characterized in that: The photovoltaic power generation output prediction model includes loading historical photovoltaic power generation output data into a neural network algorithm for training to obtain a photovoltaic power generation output prediction model; The historical photovoltaic power generation output data is normalized and expressed as: Among them, P norm Represented as normalized data, P t It is represented by the photovoltaic power generation output data on the tth day, μ is represented by the average value of the historical photovoltaic power generation output data, and σ is represented by the standard deviation of the historical photovoltaic power generation output data; Substitute the normalized data into a single-layer LSTM and express it as: h t =LSTM(P norm ,h t-1 ) The conversion from LSTM output to predicted value is expressed as, P pred =ξ(W p ·h t +b p ) P pred Inverse normalization to obtain the predicted value of photovoltaic power generation output It is expressed as, Among them, h t is represented as the hidden state of the photovoltaic power output layer, ξ is represented as the activation function, W p Denotes the output layer weight, b p Represented as output layer bias; Substitute the time to be predicted into the photovoltaic power generation output prediction model to obtain the second photovoltaic output data P2.

7. The photovoltaic output prediction method based on a data center according to claim 6, characterized in that: The fitting of the first photovoltaic output data and the second photovoltaic output data includes performing mean processing on the obtained P1 and P2 to obtain the third photovoltaic output data P3 expressed as: The third photovoltaic output data is set as the final photovoltaic output prediction value.

8. A system using a photovoltaic output prediction method based on a data center as claimed in any one of claims 1 to 7, characterized in that: It includes a light forecasting model building module, a seasonal factor correction module, a power generation output forecasting model building module and a photovoltaic output forecasting module; The illumination prediction model building module obtains historical illumination data based on the data center, loads the historical illumination data into the neural network algorithm for training, and obtains an illumination prediction model; The seasonal factor correction module divides the historical light data by season, fits the seasonal factors, and corrects the light prediction model based on the seasonal factors; The power generation output prediction model building module obtains historical photovoltaic power generation output data based on the data center, loads the historical photovoltaic power generation output data into the neural network algorithm for training, and obtains a photovoltaic power generation output prediction model; The photovoltaic output prediction module uses the light prediction model to predict the light duration and intensity at the next moment, and calculates the first photovoltaic output data based on the predicted light duration and intensity; uses the photovoltaic power generation output prediction model to predict the photovoltaic output data at the next moment, recorded as the second photovoltaic output data; fits the first photovoltaic output data with the second photovoltaic output data to obtain the third photovoltaic output data as the final prediction result.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the processor implements the steps of a photovoltaic output prediction method based on a data center according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a photovoltaic output prediction method based on a data center according to any one of claims 1 to 7 are implemented.

Images