Low-weather-resolution photovoltaic prediction method and system based on similar day training set

Abstract

The invention provides a low-weather-resolution photovoltaic prediction method and system based on a similar day training set, and belongs to the technical field of photovoltaic prediction, and the method comprises the steps: obtaining the weather forecast data of a to-be-predicted day; constructing a latest similar day similar to the weather forecast data of the day to be predicted, and obtaining meteorological data and photovoltaic power generation data of the day; reconstructing a similar day training set; decomposing two modal components for the similar day training set, establishing a mixed deep learning model for the two modal components to perform learning training, during prediction, reasoning modal component output corresponding to two modals by using the photovoltaic output of the modal component of the last set day number of the similar day training set and weather forecast data of a day to be predicted, and adding the two modal component outputs to obtain a prediction result; obtaining a trend photovoltaic output prediction curve; obtaining a volatility prediction curve based on the similar day training set; and carrying out weighted average on the obtained volatility prediction curve and trend prediction curve to obtain a final prediction result.

Description

Technical Field

[0001] The present invention belongs to the technical field of photovoltaic prediction, and in particular relates to a low-weather resolution photovoltaic prediction method and system based on a similar day training set. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Fossil fuels have long been the cornerstone of global economic growth, but their overuse has caused serious environmental problems, forcing the world to turn to exploring cleaner, low-carbon and sustainable energy. Among these energy sources, photovoltaic power generation stands out due to its low cost, strong sustainability and good environmental benefits, and has achieved rapid growth worldwide. The global photovoltaic installed capacity increased from 400GW in 2010 to 843GW in 2021, and is expected to reach 1,700GW by 2030.

[0004] Although photovoltaic energy has many advantages, its output is highly unstable due to its high dependence on weather, which poses a huge challenge to the safe and stable operation of the power grid. This situation has become more serious as more and more distributed photovoltaic systems are connected to the power grid. Accurate prediction of photovoltaic output on the day before is crucial to maintaining the security, stability and reliability of the power grid. Compared with the one-hour forecast, the 15-minute interval forecast provides finer granularity, which is more conducive to improving the dispatch of the power system, energy management planning and assisting photovoltaic participation in power market transactions. However, photovoltaic power generation is highly sensitive to frequent and rapid changes in weather, which increases the uncertainty and volatility of photovoltaic output, especially in small distributed photovoltaic power generation sites. In addition, due to cost considerations, a large number of distributed sites do not have high-precision, fine-grained weather information acquisition methods, and the weather information that can be obtained at low cost is mostly one-hour intervals, which further increases the difficulty of photovoltaic output prediction at 96 points on the day before.

[0005] The day-ahead PV output forecast is essentially a time series forecasting problem. The current existing methods are generally based on deep learning models that divide the training set and validation set into training sets, adjust parameters, and finally output the predicted value. However, the current methods are difficult to achieve the desired effect under coarse-grained weather conditions, and the assessment penalties for distributed PV sites are very severe.

[0006] The existing technologies for photovoltaic output forecasting can only solve the situation when the sampling granularity of weather conditions is the same as the forecast granularity of demand, and the prediction accuracy is not strong enough. Therefore, it is necessary to solve the problem of how to achieve accurate photovoltaic forecasts at 96 points in the day ahead at low weather resolution. Summary of the invention

[0007] To overcome the above-mentioned deficiencies of the prior art, the present invention provides a low-weather-resolution photovoltaic prediction method based on a similar day training set to achieve accurate photovoltaic prediction at 96 points before the day.

[0008] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: In a first aspect, a low-weather-resolution photovoltaic prediction method based on a similar-day training set is disclosed, comprising: Get the weather forecast data for the day to be predicted; Construct the nearest similar day with weather forecast data similar to the day to be predicted, and obtain the meteorological data and photovoltaic power generation data of that day; Calculate the similarity error between the photovoltaic power generation of the most recent similar day and the photovoltaic power generation of the set historical day, retain the photovoltaic output data and weather data of the historical day with errors below the specified threshold, and sort them from large to small according to the error size to reconstruct a similar day training set; The similar day training set is decomposed into two modal components, and a hybrid deep learning model is established for the two modal components for learning and training. When predicting, the modal component photovoltaic output of the last set number of days in the similar day training set is added to the weather forecast data of the predicted day to infer the modal component output corresponding to the two modes, and the two are added together to obtain the trend photovoltaic output prediction curve; Obtain volatility prediction curve based on similar day training set; The obtained volatility prediction curve and trend prediction curve are weighted averaged to obtain the final prediction result.

[0009] As a further technical solution, N days of historical meteorological data are obtained before obtaining the weather forecast data of the day to be predicted, and a historical period database is constructed. The meteorological data of each day contains C meteorological variables, the time resolution of the meteorological data is 1 hour, and each day contains 24 meteorological segments; For each meteorological hour period, there are 4 corresponding photovoltaic output data points with a resolution of 15 minutes.

[0010] As a further technical solution, the construction process of the historical period database includes: For the N days of historical meteorological data, the weather data for any historical day n can be expressed as a T*C matrix ; The photovoltaic output matrix corresponding to historical day n is expressed as a T*K matrix ; right and Perform reconstruction operation to obtain a two-period matrix and ; Historical weather data and photovoltaic output data for day n and The new weather data matrix is ​​obtained by mirroring the time and photovoltaic output matrix ; right and Perform reconstruction operation to obtain a two-period matrix and ; The matrix and Merge to get the weather data matrix for the historical period ; The matrix and Merge to get the historical period output matrix ; All weather data matrices of historical periods are merged into a three-dimensional weather tensor, and all output matrices of historical periods are merged into a three-dimensional photovoltaic output tensor, thus completing the construction of the historical period database.

[0011] As a further technical solution, constructing the nearest similar day with weather forecast data similar to the day to be predicted specifically includes: The weather forecast data for the predicted day can also be expressed as a T*C matrix ; Reconstruct the weather forecast data matrix to obtain the weather forecast period reconstruction matrix ; Calculate the distance between each row vector in the weather forecast period reconstruction matrix and each historical row vector, sort them in order from large to small, and record the smallest corresponding to each row vector in the weather forecast period reconstruction matrix. The coordinate position corresponding to the distance ; Three-dimensional tensor from meteorology and photovoltaic power output and In the example, the recorded coordinates are used. To reconstruct approximately Weather data matrix for the day and photovoltaic output matrix ; By obtaining Weather data matrix for the day and photovoltaic output matrix Perform weighted averaging to finally obtain the data matrix of the weather period of the nearest similar day and ; The time period data matrix and Rearrange and finally get the weather matrix of the nearest similar day and photovoltaic output matrix .

[0012] As a further technical solution, the specific steps of reconstructing the similar day training set are: Step 1: Matrix the historical daily photovoltaic power generation and the photovoltaic power generation matrix of the most recent similar day Flatten rows into normal time series and ; Step 2: Use the area error method to calculate the similarity error between the photovoltaic power generation data of the most recent similar day and the photovoltaic power generation data of the historical day on a daily basis; Step 3: Extract historical daily values ​​less than the threshold , and sort them from large to small according to the error to complete the construction of the similar day training set.

[0013] As a further technical solution, the process of obtaining the trend photovoltaic output prediction curve is as follows: Using the sliding window method, the similar day training set is divided into multiple subsets every four days, and the sliding window method that advances daily is used to form a two-dimensional matrix; Use the PV output data of the nearest similar day to replace the actual value of the forecast day and construct a complete matrix; The matrix is ​​decomposed using the MVMD algorithm, during which the data covering four days is effectively converted into a one-dimensional time series and the original long one-dimensional time series is converted into a multi-dimensional short time series, thereby ensuring that the training set and the test set have the same central frequency.

[0014] In the second aspect, a low-weather-resolution photovoltaic prediction system based on a similar-day training set is disclosed, comprising: The data acquisition module is configured to: acquire weather forecast data for the day to be predicted; The nearest similar day construction module is configured to: construct the nearest similar day with weather forecast data similar to the day to be predicted, and obtain meteorological data and photovoltaic power generation data of the day; The similar day training set reconstruction module is configured to: calculate the similarity error between the photovoltaic power generation of the most recent similar day and the photovoltaic power generation of a set number of historical days, retain the photovoltaic output data and weather data of the historical days with errors below the specified threshold, and sort them from large to small according to the error size, and reconstruct a similar day training set; The prediction curve acquisition module is configured as follows: decomposing the two modal components of the similar day training set, establishing a hybrid deep learning model for the two modal components for learning and training, and when predicting, using the modal component photovoltaic output of the last set number of days in the similar day training set plus the weather forecast data of the day to be predicted to infer the modal component outputs corresponding to the two modes, adding the two together to obtain a trend photovoltaic output prediction curve; obtaining a volatility prediction curve based on the similar day training set; The prediction module is configured to: perform weighted averaging on the obtained volatility prediction curve and trend prediction curve to obtain a final prediction result.

[0015] One or more of the above technical solutions have the following beneficial effects: The technical solution of the present invention changes the traditional similar training set construction method of directly clustering multiple weather type training sets using weather information, but customizes a training set for the future day that needs to be predicted, and attempts to construct the training set directly based on the photovoltaic power generation curve. However, since the actual photovoltaic power generation curve of the second day is still unknown, but the weather forecast information of the second day is known, we can build the most similar day from the historical database based on the weather forecast information of the second day, find the m historical weather information of each time period that is closest to the weather forecast information of the day to be predicted through similar matching of time periods, and obtain the corresponding photovoltaic power generation time period information accordingly, perform weighted average on these time periods and reassemble them into a complete day, so as to obtain the nearest similar day data.

[0016] In order to further improve the prediction accuracy of photovoltaic output on the day before, the technical solution of the present invention proposes a new hybrid deep learning model that combines multi-stream data mining with long short-term memory network (LSTM) and one-dimensional convolutional network (CNN) and self-attention mechanism. Since LSTM network and one-dimensional convolutional layer have different methods and capabilities in mining time features, their collaboration can extract a large number of time implicit features from different angles for subsequent network learning.

[0017] The technical solution of the present invention can not only solve the situation where the weather granularity is the same as the output granularity that needs to be predicted, but also has good prediction capabilities when the weather granularity is coarser than the output granularity that needs to be predicted, thereby helping a large number of distributed photovoltaic power generation sites that are difficult to obtain fine-grained weather data to improve prediction accuracy or establish prediction capabilities. The method of the technical solution of the present invention has predicted the photovoltaic output of multiple distributed sites. Compared with the general LSTM conventional prediction model, the prediction accuracy of the present invention has been improved by more than 15%.

[0018] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0020] Figure 1 This is a schematic diagram of the structure of a hybrid deep learning model according to an embodiment of the present invention; Figure 2 This is a flow chart of the overall prediction method according to an embodiment of the present invention; Figure 3 A flow chart of using MVMD to perform trend prediction in an embodiment of the present invention; Figure 4 The diagram shows the forecast effect on a sunny day; Figure 5 Shown is a schematic diagram of the forecast effect of highly volatile weather with multiple transitions between sunny and cloudy weather. DETAILED DESCRIPTION

[0021] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0022] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0023] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0024] Embodiment 1 This embodiment discloses a low-weather resolution photovoltaic prediction method based on a similar day training set, including: Step 1: Using the weather forecast data of the day to be predicted, a bidirectional time window sliding similarity matching and multi-curve weighting method are adopted to construct a nearest similar day and obtain the meteorological and photovoltaic power generation data of that day.

[0025] Then, the regional error method is used to calculate the similarity error between the photovoltaic power generation on the most recent similar day and the photovoltaic power generation on each historical day. The photovoltaic output data and weather data of historical days with errors below the specified threshold are retained and sorted from large to small according to the error size, thereby reconstructing a similar day training set.

[0026] This step solves the problem of unclear boundaries and loss of a large amount of effective boundary information caused by traditional direct weather clustering. At the same time, because the similar day training set is established for the forecast day, the data utilization is more sufficient and has the ability to update in real time. This method takes into account the similarity of time periods and judges similarity from a more detailed perspective, rather than directly making a general judgment on the overall day. At the same time, by finding the output of the most recent similar day to directly match the output of the historical day, it can also avoid the situation where the weather granularity is coarse and the weather is similar, but the actual output is completely different.

[0027] Step 2: Includes two parallel forecasting branches - volatility and trend, to generate two different forecast curves. Volatility forecasts are directly trained from the similar day training set mentioned above; in contrast, trend forecasts require a method called multivariate variational mode decomposition (MVMD) to decompose and reconstruct the similar day training set before forecasting to avoid information leakage.

[0028] To further improve the prediction accuracy, a multi-stream parallel hybrid deep learning architecture is proposed: MSLCnet, which combines a long short-term memory network (LSTM), a convolutional neural network (CNN), a residual block network, and a self-attention mechanism.

[0029] The multi-stream parallel structure allows the use of multiple neural networks for feature mining and extraction at the same time, making the mining more comprehensive and diverse in scale, and no longer restricted to the perspective of a single neural network. The residual block network can effectively avoid problems such as gradient disappearance caused by the convolutional network being too deep. The self-attention mechanism automatically screens and weights a large number of mined features to highlight important features.

[0030] Step 3: Take a weighted average of the two curves obtained in step 2 to get the final prediction result. This step can fully combine the volatility prediction and trend prediction results. The mutual correction of the two curves makes the prediction result more accurate from a statistical perspective.

[0031] Regarding the construction of similar day training sets in step one: The sub-technical solution of this embodiment changes the traditional similar training set construction method of using weather information to directly cluster multiple weather type training sets. Instead, it customizes a training set for the future day that needs to be predicted, and attempts to construct the training set directly based on the photovoltaic power generation curve.

[0032] However, since the actual photovoltaic power generation curve of the next day is still unknown, but the weather forecast information of the next day is known, the most similar day can be constructed from the historical database based on the weather forecast information of the second day. Through similar matching of time periods, the m historical weather information of each time period closest to the weather forecast information of the predicted day can be found, and the corresponding photovoltaic power generation time period information can be obtained accordingly. These time periods are weighted averaged and reassembled into a complete day, thereby obtaining the data of the nearest similar day.

[0033] Then, the photovoltaic power generation curve of the nearest similar day is used as a benchmark, and the area error method is used to perform similar matching in the historical database to find out the data of several days similar to the predicted day, and construct a similar day training set corresponding to the predicted day.

[0034] The above steps include: Step 1-1) Mathematically represent the available data.

[0035] Assuming you have Historical weather data for the day, daily meteorological data includes meteorological variables such as temperature, irradiance, wind speed, etc. The time resolution of meteorological data is 1 hour, and each day contains In addition, for each meteorological hour period, there are The data points of photovoltaic output correspond to it, with a resolution of 15 minutes. Of course, there are also weather forecast data with predicted days, and the composition and format of this data are exactly the same as the historical weather data.

[0036] The weather data of historical day n can be expressed as a Matrix, C is the type of meteorological variable, T is the number of meteorological segments:

[0037] in, Representative Day's The time period data on various weather variables.

[0038] The photovoltaic output matrix corresponding to historical day n can be expressed as a matrix:

[0039] in, Representative Tiandi The first The photovoltaic output value is displayed at 15-minute intervals.

[0040] The weather forecast data for the day to be predicted can also be expressed as a matrix:

[0041] in, Represents the day to be predicted The time period data on various weather variables.

[0042] In order to prevent serious errors in subsequent similar matching due to the different dimensions of each meteorological variable, all the above meteorological variable data have been normalized to values ​​between 0 and 1. The normalization formula is as follows:

[0043] in, is the normalized data, is the minimum value in the data, is the maximum value in the data.

[0044] Step 1-2) Reconstruct the historical period database based on dual-period sliding and data enhancement.

[0045] In getting the matrix and After that, the two matrices are reconstructed to obtain the two-period matrix and , the reconstruction method is as follows: ; ; In order to enhance the data set, the weather and photovoltaic output data are mirrored by time, so that a new weather data matrix is ​​obtained for each historical day n. and photovoltaic output matrix .

[0046]

[0047]

[0048] right and Perform the same reconstruction method as above to obtain a two-period matrix and , the matrix and ;matrix and Merge to get the weather data matrix for the historical period and output matrix .

[0049]

[0050]

[0051] All weather data matrices and output matrices for all historical periods can be combined into two three-dimensional tensors, thus completing the construction of the historical period database, which includes the weather tensor and the photovoltaic output tensor.

[0052] Weather Tensor: ,The weather tensor is specifically a historical meteorological information database, including the combination of historical weather matrices for N days; Photovoltaic output tensor: .

[0053] Steps 1-3) Build the most recent similar day.

[0054] After obtaining the reconstructed historical meteorological information database After that, the weather forecast information is also reconstructed to facilitate similar matching of meteorological periods. Since each meteorological period is represented by a point of data, the weather forecast data matrix is ​​reconstructed in the following way to avoid repeated matching periods: .

[0055] The weather forecast period reconstruction matrix can be expressed as:

[0056] Compute each row vector With each row vector The distance between , and sort them in order from largest to smallest, and record the The corresponding smallest The coordinate position corresponding to the distance The distance is calculated as follows:

[0057] A three-dimensional tensor database of meteorological and photovoltaic power output and In the example, the recorded coordinates are used. To reconstruct approximately Weather data matrix for the day and photovoltaic output matrix The construction method is as follows:

[0058] .

[0059] By obtaining The weather and photovoltaic output data matrix of the day are weighted averaged to finally obtain the data matrix of the weather period of the nearest similar day. and .

[0060] Since the weather granularity is coarse, the weighted average of the output fluctuations corresponding to multiple close weather periods can make the final result as similar to the true value as possible, thereby improving the similarity of similar matching outputs.

[0061]

[0062] The time period data matrix and Rearrange and finally get the weather matrix of the nearest similar day and photovoltaic output matrix . The rearrangement method is as follows:

[0063]

[0064] in: Representation Matrix Middle Line Elements of a column; Representation Matrix Middle Line Elements of a column.

[0065] Steps 1-4) Construction of similar day training set.

[0066] Different from the traditional method of similarity clustering using all-day meteorological data, the similarity training set establishment method proposed in the present invention is based on the photovoltaic power generation data of the most recent similar day. The specific method is as follows: Step 1-4-1): Convert the historical daily photovoltaic power generation matrix and the photovoltaic power generation matrix of the most recent similar day Flatten rows into normal time series and .

[0067] Step 1-4-2): Use the area error method to calculate the similarity error between the photovoltaic power generation data of the most recent similar day and the photovoltaic power generation data of the historical day on a daily basis. The calculation method of the area error method is as follows:

[0068] represents the similarity area error between two curves, Indicates the start time of photovoltaic output, Indicates the end time of photovoltaic output, corresponds to the full-day photovoltaic output curve of the nearest similar day, and This corresponds to the full-day photovoltaic output curve for a certain day in history.

[0069] Step 1-4-3): Extract historical daily values ​​less than the threshold , and sort them from large to small according to the error to complete the construction of the similar day training set. It is determined by the actual situation, but it is generally not recommended to be too small to avoid making the training set too small and resulting in a decrease in prediction accuracy. Therefore, the similar day training set specifically includes historical daily weather and historical photovoltaic output selected in accordance with the above conditions.

[0070] Step 2: MSLCnet construction.

[0071] In order to further improve the prediction accuracy of photovoltaic output on the day before, this embodiment proposes a new hybrid deep learning model called MSLCnet, which combines multi-stream data mining with long short-term memory network (LSTM) and one-dimensional convolutional network (CNN) and self-attention mechanism. The architecture of the model is as follows: Figure 1 shown.

[0072] After receiving the input of various meteorological data and three-day historical photovoltaic output information, MSLCnet uses a multi-stream parallel method to extract and process sequence features. Since the LSTM network and the one-dimensional convolutional layer have different methods and capabilities in mining temporal features, their collaboration can extract a large number of temporal implicit features from different angles for subsequent network learning. In addition, in order to alleviate the negative impact of the network being too deep, MSLCnet introduces a residual block component to ensure that the depth of the convolutional layer does not affect the performance of the entire network. After parallel feature extraction, the importance of different feature quantities in a large number of obtained feature vectors will inevitably differ, and even misleading or fictitious features may appear. To solve this problem, a self-attention layer is introduced, which automatically filters and selects relevant important feature vectors through the self-attention mechanism to increase their weights, and finally inputs them into the fully connected layer to establish a regression relationship. Therefore, the MSLCnet architecture can not only overcome the convergence challenges brought by the linear series stacking network, but also flexibly utilize the temporal features of different scales and modes extracted from the input by different networks.

[0073] Step 2-1) Use the MVMD algorithm to perform trend prediction without information leakage.

[0074] The main application of the MVMD algorithm is that it can decompose multivariate time series into modal components that share a common center frequency. Its constrained objective function is as follows:

[0075] in: represents the number of decomposed modal components; represents the number of raw signals fed into the algorithm; Represents the center frequency of each modal component after MVMD decomposition; A set of multi-dimensional time series signals representing an input; represents the decomposed multi-dimensional time series signal set, which is the photovoltaic output matrix after transformation in this implementation example.

[0076] Then, the Lagrange multiplier method is used to transform the constrained problem into an unconstrained problem. The Lagrange function form is as follows:

[0077] In this equation, is a penalty factor, which is a constant; and is a time-varying Lagrange multiplier.

[0078] The ADMM method is used to solve the unconstrained optimization problem, which transforms the optimization problem into several simpler sub-optimization problems. The update equations for the modal components and center frequencies are as follows:

[0079]

[0080] The following sections provide guidance for using the MVMD algorithm while preventing spatiotemporal information leakage.

[0081] Step 2-2) In the traditional time series forecasting and inference process, data from the previous few days are usually used to predict future data, and then the training set is divided into smaller training units.

[0082] This embodiment of the sub-technical solution uses the PV output of the past three days to predict the PV output of the next day. Therefore, the effective data span of the real single input training set is four days. Therefore, in the first stage, the sliding window method is used to divide the similar day training set into four subsets every four days, and the sliding window method is used to advance daily, so as to form a two-dimensional matrix .

[0083]

[0084] However, it should be noted that the last row of the matrix lacks data from the last day, because that day is the prediction target of our test set. Including the photovoltaic output data on the prediction day will lead to spatiotemporal information leakage. To solve this problem, the photovoltaic output data on the nearest similar day is used to replace the actual value on the prediction day, thereby preventing spatiotemporal information leakage and constructing a complete matrix for decomposition.

[0085] Next, we simply decompose the matrix using the MVMD algorithm. In this process, the data covering four days is effectively converted into a one-dimensional time series, and the original long one-dimensional time series is converted into a multi-dimensional short time series, thereby ensuring that the training set and the test set have the same central frequency. When decomposing, it is generally decomposed into two modal components of different frequencies.

[0086]

[0087] During training, an MSLCnet is established for the two modal components for learning and training. The reasoning method is to use the modal component photovoltaic output of the past three days plus the weather data of the day to predict the fourth day. During prediction, the modal component photovoltaic output of the last three days of the training set plus the weather forecast data of the predicted day are used to infer the modal component output corresponding to the two modes. The two are added together to obtain the trend photovoltaic output prediction curve, see the attached Figure 3 shown.

[0088] Step 2-3) Directly use the similar day training set to make volatility predictions.

[0089] Similarly, the above method divides the training set of similar days into small units of four days, and the output data of the first three days plus the weather data of the fourth day are used to infer the output data of the fourth day, but no MVMD decomposition is performed. Instead, training is performed directly on the training set. When predicting and reasoning, the PV output of the last three days of the training set and the weather forecast data of the day to be predicted are input to obtain the volatility prediction curve. During training, an MSLCnet model can be established.

[0090] Step 3-1) The final measurement curve is generated.

[0091] After obtaining the trend and volatility forecast curves, the final forecast result can be obtained by weighted averaging the two curves. The formula is as follows:

[0092] It is usually recommended to and They are all set to 0.5, but the specific parameters can be adjusted through experiments according to the specific power station.

[0093] Deployment algorithm: The algorithm of this embodiment exists in the form of a Python library service and is integrated into an intelligent computing machine. Whenever a prediction is needed, the relevant historical output and weather data and weather forecast data are passed in. The intelligent computing machine will automatically start the training and prediction services and output the required day-ahead output curve.

[0094] The method described in this embodiment has predicted the photovoltaic output of multiple distributed sites. Compared with the general LSTM conventional prediction model, the prediction accuracy of the present invention has been improved by more than 15%. Figure 4 , 5 shown.

[0095] Embodiment 2 The purpose of this embodiment is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the program.

[0096] Embodiment 3 The purpose of this embodiment is to provide a computer-readable storage medium.

[0097] A computer-readable storage medium stores a computer program, which executes the steps of the above method when executed by a processor.

[0098] Embodiment 4 The purpose of this embodiment is to provide a second aspect, disclosing a low weather resolution photovoltaic prediction system based on a similar day training set, including: The data acquisition module is configured to: acquire weather forecast data for the day to be predicted; The nearest similar day construction module is configured to: construct the nearest similar day with weather forecast data similar to the day to be predicted, and obtain meteorological data and photovoltaic power generation data of the day; The similar day training set reconstruction module is configured to: calculate the similarity error between the photovoltaic power generation of the most recent similar day and the photovoltaic power generation of a set number of historical days, retain the photovoltaic output data and weather data of the historical days with errors below the specified threshold, and sort them from large to small according to the error size, and reconstruct a similar day training set; The prediction curve acquisition module is configured as follows: decomposing the two modal components of the similar day training set, establishing a hybrid deep learning model for the two modal components for learning and training, and when predicting, using the modal component photovoltaic output of the last set number of days in the similar day training set plus the weather forecast data of the day to be predicted to infer the modal component outputs corresponding to the two modes, adding the two together to obtain a trend photovoltaic output prediction curve; obtaining a volatility prediction curve based on the similar day training set; The prediction module is configured to: perform weighted averaging on the obtained volatility prediction curve and trend prediction curve to obtain a final prediction result.

[0099] Embodiment 5 The purpose of this embodiment is to provide a computer program product containing instructions, which, when running on a computer, enables the computer to execute the methods and functions involved in any of the above embodiments. The steps involved in the apparatus of the above embodiment correspond to the method embodiment 1, and the specific implementation method can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0100] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0101] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A low-weather-resolution photovoltaic prediction method based on a similar day training set, characterized by: include: Get the weather forecast data for the day to be predicted; Construct the nearest similar day with weather forecast data similar to the day to be predicted, and obtain the meteorological data and photovoltaic power generation data of that day; Calculate the similarity error between the photovoltaic power generation of the most recent similar day and the photovoltaic power generation of the set historical day, retain the photovoltaic output data and weather data of the historical day with errors below the specified threshold, and sort them from large to small according to the error size to reconstruct a similar day training set; The similar day training set is decomposed into two modal components, and a hybrid deep learning model is established for the two modal components for learning and training. When predicting, the modal component photovoltaic output of the last set number of days in the similar day training set is added to the weather forecast data of the predicted day to infer the modal component output corresponding to the two modes, and the two are added together to obtain the trend photovoltaic output prediction curve; Obtain volatility prediction curve based on similar day training set; The obtained volatility prediction curve and trend prediction curve are weighted averaged to obtain the final prediction result.

2. The low-weather-resolution photovoltaic prediction method based on similar day training sets as claimed in claim 1, characterized in that: Before obtaining the weather forecast data for the day to be predicted, obtain N days of historical meteorological data and build a historical period database. The daily meteorological data contains C meteorological variables. For each meteorological hour period, there is a corresponding photovoltaic output data point.

3. The low-weather-resolution photovoltaic prediction method based on similar day training sets as claimed in claim 2, characterized in that: The construction process of the historical period database includes: For the N days of historical meteorological data, the weather data for any historical day n can be expressed as a T*C matrix ; The photovoltaic output matrix corresponding to historical day n is expressed as a T*K matrix ; right and Perform reconstruction operation to obtain a two-period matrix and ; Historical weather data and photovoltaic output data for day n and The new weather data matrix is ​​obtained by mirroring the time and photovoltaic output matrix ; right and Perform reconstruction operation to obtain a two-period matrix and ; The matrix and Merge to get the weather data matrix for the historical period ; The matrix and Merge to get the historical period output matrix ; All weather data matrices of historical periods are merged into a three-dimensional weather tensor, and all output matrices of historical periods are merged into a three-dimensional photovoltaic output tensor, thus completing the construction of the historical period database.

4. The low-weather-resolution photovoltaic prediction method based on similar day training sets as claimed in claim 1, characterized in that: Construct the nearest similar day with weather forecast data similar to the day to be predicted, including: The weather forecast data for the predicted day can also be expressed as a T*C matrix ; Reconstruct the weather forecast data matrix to obtain the weather forecast period reconstruction matrix ; Calculate the distance between each row vector in the weather forecast period reconstruction matrix and each historical row vector, sort them in order from large to small, and record the smallest corresponding to each row vector in the weather forecast period reconstruction matrix. The coordinate position corresponding to the distance ; Three-dimensional tensor from meteorology and photovoltaic power output and In the example, the recorded coordinates are used. To reconstruct approximately Weather data matrix for the day and photovoltaic output matrix ; By obtaining Weather data matrix for the day and photovoltaic output matrix Perform weighted averaging to finally obtain the data matrix of the weather period of the nearest similar day and ; The time period data matrix and Rearrange and finally get the weather matrix of the nearest similar day and photovoltaic output matrix .

5. The low-weather-resolution photovoltaic prediction method based on similar day training sets as claimed in claim 1, characterized in that The specific steps of the daily training set are: Step 1: Matrix the historical daily photovoltaic power generation and the photovoltaic power generation matrix of the most recent similar day Flatten rows into normal time series and ; Step 2: Use the area error method to calculate the similarity error between the photovoltaic power generation data of the most recent similar day and the photovoltaic power generation data of the historical day on a daily basis; Step 3: Extract historical daily values ​​less than the threshold , and sort them from large to small according to the error to complete the construction of the similar day training set.

6. The low-weather-resolution photovoltaic prediction method based on similar day training sets as claimed in claim 1, characterized in that: The process of obtaining the trend photovoltaic output prediction curve is as follows: Using the sliding window method, the similar day training set is divided into four subsets every four days, and the sliding window method is used to advance daily, thus forming a two-dimensional matrix; Use the PV output data of the nearest similar day to replace the actual value of the forecast day and construct a complete matrix; The matrix is ​​decomposed using the MVMD algorithm, during which the data covering four days is effectively converted into a one-dimensional time series and the original long one-dimensional time series is converted into a multi-dimensional short time series, thereby ensuring that the training set and the test set have the same central frequency.

7. A low-weather-resolution photovoltaic prediction system based on a similar-day training set, characterized by: include: The data acquisition module is configured to: acquire weather forecast data for the day to be predicted; The nearest similar day construction module is configured to: construct the nearest similar day with weather forecast data similar to the day to be predicted, and obtain meteorological data and photovoltaic power generation data of the day; The similar day training set reconstruction module is configured to: calculate the similarity error between the photovoltaic power generation of the most recent similar day and the photovoltaic power generation of a set number of historical days, retain the photovoltaic output data and weather data of the historical days with errors below the specified threshold, and sort them from large to small according to the error size, and reconstruct a similar day training set; The prediction curve acquisition module is configured as follows: decomposing the two modal components of the similar day training set, establishing a hybrid deep learning model for the two modal components for learning and training, and when predicting, using the modal component photovoltaic output of the last set number of days in the similar day training set plus the weather forecast data of the day to be predicted to infer the modal component outputs corresponding to the two modes, adding the two together to obtain a trend photovoltaic output prediction curve; obtaining a volatility prediction curve based on the similar day training set; The prediction module is configured to: perform weighted averaging on the obtained volatility prediction curve and trend prediction curve to obtain a final prediction result.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method described in any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 6 are performed.

Images