Sand and dust weather solar irradiance combined forecasting method and system and storage medium

Abstract

The invention provides a sand and dust weather solar irradiance combined forecasting method and system and a storage medium. The method comprises the steps that meteorological station observation data and sand and dust monitoring data are acquired; inputting meteorological station observation data and dust monitoring data into the solar irradiance forecasting model; processing the meteorological station observation data and the sand and dust monitoring data through a solar irradiance forecasting model, and constructing an independent long-short-term memory network sub-forecasting model; carrying out cumulative summation on forecasting results of the plurality of long-short-term memory network sub-forecasting models for reconstruction to obtain a solar irradiance short-term and imminent forecasting value in sand and dust weather; wherein the solar irradiance forecasting model is obtained by taking meteorological station observation data and sand and dust monitoring data as input, performing variational mode decomposition on a pre-acquired solar irradiance observation value, then taking the solar irradiance observation value as output, and performing training based on a long short-term memory network. According to the invention, the problem of low photovoltaic power generation output caused by inaccurate solar irradiance forecast in the existing sand and dust weather is solved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of meteorological forecasting, and in particular to a method, system and storage medium for combined forecasting of solar irradiance in sandstorm weather. Background Art

[0002] At present, numerical models such as CMA and ECWMF are mainly used, combined with statistical and dynamic downscaling techniques such as machine learning, to forecast meteorological elements such as solar irradiance for photovoltaic power station power generation conditions. However, there are large deviations in the solar irradiance forecast for photovoltaic power generation equipment under weather and dust scenarios.

[0003] Due to the sparse, unevenly distributed and insufficient ground meteorological observation stations at solar power generation bases in some areas, there is a lack of quantitative observation data on atmospheric pollutants transported by sandstorms, such as PM10. It is impossible to effectively use the observation data and empirical models of PM10 and other stations to make a more accurate estimate of the solar irradiance of photovoltaic power generation clusters under sandstorm conditions. As a result, there are large deviations in the solar irradiance forecast in areas with dense distribution of photovoltaic power stations under the influence of sandstorm particles, and the output of photovoltaic power stations is low. Summary of the invention

[0004] The present invention provides a combined forecast method, system and storage medium for solar irradiance in sandstorm weather, which are used to solve the problem that the inaccurate solar irradiance forecast in sandstorm weather leads to low photovoltaic power generation output.

[0005] The present disclosure provides a combined forecasting method for solar irradiance in sandstorm weather, comprising:

[0006] Obtain meteorological station observation data and sand and dust monitoring data;

[0007] Inputting the meteorological station observation data and dust monitoring data into a pre-trained solar irradiance forecast model;

[0008] The meteorological station observation data and dust monitoring data are processed by the solar irradiance forecast model to construct multiple independent long-short term memory network sub-forecast models;

[0009] The forecast results of multiple long-short-term memory network sub-forecast models are accumulated and reconstructed to obtain the short-term forecast value of solar irradiance under sandstorm weather.

[0010] The solar irradiance forecast model takes the meteorological station observation data and the dust monitoring data as input, and performs variational mode decomposition on the pre-acquired solar irradiance observation values, which are then trained based on the long short-term memory network as output.

[0011] According to a combined forecasting method for solar irradiance in sandstorm weather provided by the present disclosure, the solar irradiance forecasting model is obtained by taking meteorological station observation data and sandstorm monitoring data as input, performing variational mode decomposition on pre-acquired solar irradiance observation values ​​and training them based on long short-term memory networks as output, and specifically includes:

[0012] Obtain historical meteorological station observation data, dust monitoring data and solar irradiance observation time series;

[0013] Decomposing the solar irradiance observation time series by using a variational mode decomposition method, and preprocessing the meteorological station observation data and sand and dust monitoring data;

[0014] The preset long short-term memory network is trained based on the preprocessed meteorological station observation data and dust monitoring data as well as the solar irradiance observation time series after variational mode decomposition to obtain a multivariable and multi-step solar irradiance prediction model.

[0015] According to a sandstorm weather solar irradiance combined forecasting method provided by the present disclosure, the solar irradiance observation time series is decomposed by a variational mode decomposition method, specifically comprising:

[0016] Based on the solar irradiance observation time series, a constrained variational problem is constructed by a variational mode decomposition method, and a constrained variational model is established;

[0017] Based on the constrained variational model, a penalty factor and a Lagrangian operator are introduced to convert the constrained variational problem into an unconstrained variational problem. The variational problem is solved by the multiplication operator alternating direction method to decompose and extract each eigenmode function and center frequency.

[0018] According to a combined forecast method for solar irradiance in sandstorm weather provided by the present disclosure, the preprocessing of the meteorological station observation data and the sandstorm monitoring data specifically includes:

[0019] Filling missing values ​​of the meteorological station observation data, converting visibility units, and setting the still wind observation data to zero;

[0020] The dust monitoring data is interpolated based on the latitude and longitude location information of the monitoring station to obtain the station dust score data.

[0021] According to a sandstorm weather solar irradiance combined forecasting method provided by the present disclosure, the preset long short-term memory network is trained based on the pre-processed meteorological station observation data and sandstorm monitoring data and the solar irradiance observation time series after variational mode decomposition to obtain a multivariate multi-step solar irradiance prediction model, which specifically includes:

[0022] The training and testing datasets are constructed based on the preprocessed meteorological station observation data and dust monitoring data as well as the solar irradiance observation time series after variational mode decomposition.

[0023] Based on the training data set, a preset long short-term memory network is trained, and multiple intrinsic mode functions obtained by variational mode decomposition of solar irradiance observation time series data are used as prediction labels;

[0024] Based on each intrinsic mode function, an independent long short-term memory network sub-forecast model is constructed, and the long short-term memory network sub-forecast models are combined and reconstructed to obtain a multivariable and multi-step solar irradiance prediction model.

[0025] According to a combined forecasting method for solar irradiance in sandstorm weather provided by the present disclosure, the training data set and the test data set are constructed based on the pre-processed meteorological station observation data and sandstorm monitoring data and the solar irradiance observation time series after variational mode decomposition, specifically including:

[0026] Experiment on setting sliding window parameters of time series data based on solar irradiance observation time series after variational mode decomposition;

[0027] Through the preset grid search method, the window size is set according to the set value, the training effect and the test effect are compared, the optimal sliding window parameter value of the model is selected, and the construction of the training data set and the test data set is completed.

[0028] The present disclosure also provides a sandstorm weather solar irradiance combined forecasting system, the system comprising:

[0029] Data acquisition module, used to obtain meteorological station observation data and sand and dust monitoring data;

[0030] A data input module, used for inputting the meteorological station observation data and dust monitoring data into a pre-trained solar irradiance forecast model;

[0031] A forecast submodule, used to process the meteorological station observation data and dust monitoring data through the solar irradiance forecast model, and construct multiple independent long-short term memory network sub-forecast models;

[0032] The irradiance prediction module is used to reconstruct the cumulative sum of the forecast results of multiple long-short-term memory network sub-forecast models to obtain the short-term forecast value of solar irradiance under sandstorm weather;

[0033] The solar irradiance forecast model takes the meteorological station observation data and the dust monitoring data as input, and performs variational mode decomposition on the pre-acquired solar irradiance observation values, which are then trained based on the long short-term memory network as output.

[0034] The present disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the combined forecasting method for solar irradiance in sandstorm weather as described in any one of the above-mentioned methods is implemented.

[0035] The present disclosure also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the combined forecasting method for solar irradiance in sandstorm weather as described in any one of the above is implemented.

[0036] The present disclosure also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned methods for combined forecasting of solar irradiance in sandstorm weather.

[0037] The present invention provides a combined forecast method, system and storage medium for solar irradiance in sandstorm weather. The system adopts a variational mode decomposition method and a long short-term memory network combination method to model and forecast solar irradiance in sparse areas of PM2.5 and other atmospheric pollutant observation stations under sandstorm weather conditions, which can effectively improve the prediction accuracy of solar irradiance. Based on the observation and forecast data of photovoltaic power stations or nearby meteorological observation stations, the advantages of imager sandstorm monitoring DST data, such as wide coverage, high frequency, and continuous sandstorm monitoring capabilities, are fully utilized. The variational mode decomposition method and the long short-term memory network combination algorithm are used for modeling and training, and multi-step forecasts of short-term solar irradiance are made, so as to obtain the solar irradiance situation in the short term in the future under the influence of the current sandstorm weather.

[0038] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the present disclosure and do not constitute a limitation of the present disclosure. The above and other features and advantages will become more apparent to those skilled in the art by describing detailed example embodiments with reference to the accompanying drawings, in which:

[0040] Figure 1 It is a flow chart of a combined forecast method of solar irradiance for sandstorm weather provided by the present disclosure;

[0041] Figure 2 This is a flow chart of a sandstorm weather solar irradiance combined forecasting system provided by the present disclosure;

[0042] Figure 3It is a schematic diagram of prediction factor selection and VMD decomposition test result analysis provided by the present disclosure;

[0043] Figure 4 This is a schematic diagram of the comparison between the solar irradiance observation and forecast at Dunhuang Station in April 2023 using the VMD-LSTM and LSTM models provided by the present disclosure;

[0044] Figure 5 It is a schematic diagram comparing the solar irradiance observation and forecast at Minqin Station in April 2023 using VMD-LSTM and LSTM models provided by the present disclosure;

[0045] Figure 6 This is an architecture diagram of a sandstorm weather solar irradiance combined forecasting system provided by the present disclosure;

[0046] Figure 7 It is a schematic diagram of the structure of an electronic device provided by the present disclosure.

[0047] Reference numerals:

[0048] 110: data acquisition module; 120: data input module; 130: forecast submodule; 140: irradiance prediction module;

[0049] 710: processor; 720: communication interface; 730: memory; 740: communication bus. DETAILED DESCRIPTION

[0050] In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0051] In the absence of conflict, the various embodiments of the present disclosure and the various features therein may be combined with each other.

[0052] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0053] The terms used herein are only used to describe specific embodiments and are not intended to limit the present disclosure. As used herein, the singular forms "a" and "the" are also intended to include plural forms, unless the context clearly indicates otherwise. It will also be understood that when the terms "including" and / or "made of" are used in this specification, the presence of the features, wholes, steps, operations, elements and / or components is specified, but the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or groups thereof is not excluded. "Connected" or "connected" and similar words are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect.

[0054] Combine the following Figure 1 A combined forecasting method for solar irradiance in sandstorm weather disclosed in the present invention is described, comprising:

[0055] S100, obtaining meteorological station observation data and sand and dust monitoring data;

[0056] S200, inputting the meteorological station observation data and the dust monitoring data into a pre-trained solar irradiance forecast model;

[0057] S300, processing the meteorological station observation data and dust monitoring data through the solar irradiance forecast model to construct multiple independent long short-term memory network sub-forecast models;

[0058] S400, reconstructing by accumulating and summing the forecast results of multiple long-short term memory network sub-forecast models to obtain the short-term forecast value of solar irradiance under sandstorm weather;

[0059] The solar irradiance forecast model takes the meteorological station observation data and the dust monitoring data as input, and performs variational mode decomposition on the pre-acquired solar irradiance observation values, which are then trained based on the long short-term memory network as output.

[0060] In the present disclosure, the solar radiation, wind speed, visibility and other data of the ground observation station during the daytime of 06:00-18:00 from March 1st to May 31st in the spring of the dust peak season in 2021-2023 and the DST data of dust monitoring by the FY4A AGRI imager are used as examples, and the observation station is Dunhuang Station.

[0061] Based on the observation and forecast data of photovoltaic power stations or nearby meteorological observation stations, the advantages of FY-4A AGRI imager's sand and dust monitoring DST data, such as wide coverage, high frequency, and continuous sand and dust monitoring capabilities, are fully utilized. The VMD-LSTM combined algorithm is used for modeling and training, and multi-step forecasts of short-term solar radiation are made. The solar irradiance conditions for the next 0 to 2 hours every 15 minutes under the influence of the current sandstorm weather are obtained.

[0062] Among them, the solar irradiance forecast model takes the meteorological station observation data and dust monitoring data as input, and the pre-acquired solar irradiance observation values ​​are trained based on the long short-term memory network after variational mode decomposition as output, which specifically includes:

[0063] Obtain historical meteorological station observation data, dust monitoring data and solar irradiance observation time series;

[0064] Decomposing the solar irradiance observation time series by using a variational mode decomposition method, and preprocessing the meteorological station observation data and sand and dust monitoring data;

[0065] The preset long short-term memory network is trained based on the preprocessed meteorological station observation data and dust monitoring data as well as the solar irradiance observation time series after variational mode decomposition to obtain a multivariable and multi-step solar irradiance prediction model.

[0066] In this disclosure, solar irradiance observation data from 06:00 to 18:00 in the spring during the peak period of sandstorm weather in 2021-2023, wind speed and visibility observation data at meteorological stations, FY4A AGRI sandstorm monitoring and other products were selected as sample data, with a total of 26,496 samples.

[0067] Specifically, it is necessary to identify whether there is sandstorm weather in the photovoltaic power station. With reference to the national standard (GB / T 20480-2017) and the "Ground Meteorological Observation Specification" published by the China Meteorological Administration, the visibility observation data of the photovoltaic power station or the nearby meteorological station is used to identify whether there is sandstorm weather and its level. The identification standard is shown in Table 1; Identification process: First, if the visibility is less than (greater than) 10km, there is (is not) sandstorm weather in the photovoltaic power generation area of ​​the concerned area; secondly, when it is determined that there is sandstorm weather, it is divided into five levels: floating dust, blowing sand, sandstorm, strong sandstorm, and extremely strong sandstorm according to the horizontal visibility and the wind force.

[0068] Table 1 Standards for judging sandstorm weather

[0069]

[0070] The solar irradiance observation time series is decomposed by a variational mode decomposition method, specifically including:

[0071] Based on the solar irradiance observation time series, a constrained variational problem is constructed by a variational mode decomposition method, and a constrained variational model is established;

[0072] Based on the constrained variational model, a penalty factor and a Lagrangian operator are introduced to convert the constrained variational problem into an unconstrained variational problem. The variational problem is solved by the multiplication operator alternating direction method to decompose and extract each eigenmode function and center frequency.

[0073] In the present disclosure, the variational mode decomposition method (VMD) is used to decompose the time series of solar irradiance observations with nonlinear characteristics to reduce the influence of the complexity and non-stationarity of the original data on the prediction accuracy of the model. The solar irradiance decomposition process is mainly divided into two parts: constructing a constrained variational problem and solving the variational problem. Assuming that each intrinsic mode function (IMF) is a finite bandwidth with a different center frequency, when the constraint condition is that the sum of each intrinsic mode function is equal to the original signal f(t) of the solar irradiance, K intrinsic mode functions are found so that the sum of the estimated bandwidth of each intrinsic mode function is minimized. The constrained variational model expression (1) is:

[0074]

[0075] In the formula, u k = {u1...u k} and w k ={w1...w k} are the K intrinsic mode functions IMF and their central frequencies obtained by VMD decomposition of solar irradiance series; σ(t) is the Dirichlet function.

[0076] In order to obtain the optimal solution of the constrained variational model, the penalty factor α and the Lagrangian operator λ(t) are introduced to transform the constrained variational problem into an unconstrained variational problem. The multiplication operator alternating direction method is used to solve the variational problem, and finally the eigenmode functions and center frequencies are decomposed and extracted. The extended Lagrangian expression (2) is as follows:

[0077]

[0078] The present invention discloses a decomposition test process for the observed original solar irradiance time series: first, the modal number K = {2, ..., N}, α = 2000, τ = 0, N ≤ 10, and the VMD decomposition is performed under different K values ​​to obtain the correlation coefficients of each modal component IMF and the original solar irradiance sequence. As the modal number increases, the correlation coefficient shows a downward trend as a whole; when the modal number is 5, the correlation coefficient decreases slowly. At this time, the center frequencies of the intrinsic modal components IMF are closer as the K value increases (see Figure 3 (a) in the figure), at this time, the maximum value of K is 5, which is the modal number N.

[0079] Preprocessing the meteorological station observation data and dust monitoring data includes:

[0080] Filling missing values ​​of the meteorological station observation data, converting visibility units, and setting the still wind observation data to zero;

[0081] The dust monitoring data is interpolated based on the latitude and longitude location information of the monitoring station to obtain the station dust score data.

[0082] In this disclosure, various types of observation data are preprocessed, including missing value filling, visibility data unit conversion from m to km, stationary wind observation data set to 0 and other preprocessing work; FY4A dust monitoring product DST is collected, and according to the latitude and longitude location information of the concerned station, the DST product is interpolated to obtain the station dust score data; in order to reduce the complexity of the model, the mutual information method is used for feature selection to determine the solar irradiance sequence related factors, and finally 15 factors including sunshine solar altitude angle, hour, dust score, wind speed, visibility, etc. are selected, see Figure 3 Figure (b) in .

[0083] Based on the preprocessed meteorological station observation data and dust monitoring data and the solar irradiance observation time series after variational mode decomposition, the preset long short-term memory network is trained to obtain a multivariate multi-step solar irradiance prediction model, which specifically includes:

[0084] The training and testing datasets are constructed based on the preprocessed meteorological station observation data and dust monitoring data as well as the solar irradiance observation time series after variational mode decomposition.

[0085] Based on the training data set, a preset long short-term memory network is trained, and multiple intrinsic mode functions obtained by variational mode decomposition of solar irradiance observation time series data are used as prediction labels;

[0086] Based on each intrinsic mode function, an independent long short-term memory network sub-forecast model is constructed, and the long short-term memory network sub-forecast models are combined and reconstructed to obtain a multivariable and multi-step solar irradiance prediction model.

[0087] In the present disclosure, time series data are prepared, and training and test matrices are determined. The main parameters of the model input data set characteristics include the selected dust fraction DST and other parameter characteristics, and the prediction label is the future 2h solar irradiance VMD intrinsic mode component IMF (one-dimensional output parameter).

[0088] The data set is divided as follows: the data from March 1 to May 31, 2021-2022 is used as the training data set, and the data from March 1 to May 31, 2023 is used as the test data set.

[0089] The window_size parameter of the sliding window for time series data is set experimentally. The grid search method is used with sliding window sizes of 8, 12, 16, 20, and 24, respectively. The training and test results are compared, and the optimal sliding window parameter value of the model is selected as 20, and the forecast step size is 8 (15 minutes per 2 hours in the future).

[0090] In the present disclosure, the long short-term memory network (LSTM network) includes two LSTM layers, with 32 and 16 layers respectively; and one fully connected layer with 8 layers. The number of network nodes (num_units), learning rate (learning_rate), training step size (steps), and number of training (epochs) are determined. The loss function of the model is MAE, which dynamically expresses the error change process between the IMF predicted value and the true value during the training process.

[0091] The five intrinsic mode functions (IMFs) obtained by VMD decomposition of solar irradiance time series data are used as prediction labels, and independent LSTM sub-prediction models are built for each IMF. The optimal parameters are obtained through model comparison experiments (see Table 2 for details), where the data batch size (batch_size) is set to 128, the learning rate (learning_rate) is 0.01, the optimizer uses Rmsprop, the activation function is ReLu, and the number of training (epochs) is 10.

[0092] Reconstruct the solar irradiance forecast value. Reconstruct the cumulative sum of the IMF forecast values ​​of each intrinsic mode component to obtain the short-term forecast value of solar irradiance every 15 minutes in the next 2 hours under the conditions of sandstorm weather process.

[0093] In this disclosure, after obtaining the solar irradiance forecast model, the mean absolute error (MAE), root mean square error (RMSE), goodness of fit (R 2 ) is used to evaluate the prediction performance of the model. The calculation expressions (3)-(7) are as follows:

[0094]

[0095] Where TSS represents the observed value y i and expected value The sum of squares of deviations represents the total degree of change of the variable; the RSS observation value y i With the predicted value The sum of squares of the deviations represents the unknown degree of change in the variable.

[0096] In a specific embodiment, the disclosed technology is used to test the forecast effect during the period of frequent sandstorms in the Hexi Corridor area of ​​Gansu Province. Figure 4 and Figure 5This is a comparison chart of the solar irradiance forecast and observation curve effects for the daytime (06:00-18:00) solar irradiance of 2h every 15min under the two sandstorm weather conditions on April 12-13 and 22-23, 2023 in a certain area. Figure 4 (a) and (b) are schematic diagrams comparing the observed and predicted solar irradiance values ​​of Dunhuang Station in April 2023 using the VMD-LSTM model. Figure 4 (c) and (d) are schematic diagrams comparing the observed and predicted values ​​of solar irradiance at Dunhuang Station in April 2023 using the LSTM model; Figure 5 (a) and (b) are schematic diagrams comparing the observed and predicted values ​​of solar irradiance at Minqin Station in April 2023 using the VMD-LSTM model. Figure 5 (c) and (d) are schematic diagrams comparing the observed and predicted values ​​of solar irradiance at Minqin Station in April 2023 using the LSTM model. Referring to Table 2, the performance evaluation indicators of the VMD-LSTM and LSTM models are shown. The evaluation indicators of VMD-LSTM at Dunhuang Station under the two sandstorm weather conditions are MAE, RMSE, and R 2 90.80~51.58W / m 2 , 70.53~121.91W / m 2 , 0.72~0.84W / m 2 Minqin Station Evaluation Indicators MAE, RMSE, R 2 44.96~96.32W / m 2 , 60.69~124.99W / m 2 , 0.443~0.896W / m 2 , which are better than the prediction performance of the LSTM single model.

[0097] Table 2 Comparison of performance indicators of VMD-LSTM and LSTM prediction models

[0098]

[0099] By establishing solar irradiance forecasting technology based on FY4A AGRI imager dust monitoring DST and other product data and observation data such as solar irradiance, the advantages of FY4A remote sensing observation products such as wide coverage, high update frequency, and continuous dynamic monitoring are fully utilized, which improves the forecast accuracy of solar irradiance in sandstorm weather in areas with sparse observations of atmospheric pollutants such as PM2.5, thereby enhancing the support capacity of photovoltaic power generation condition forecast for power dispatching decisions; and for the forecast of solar irradiance time series with strong nonlinearity under the influence of sandstorm weather, the VMD-LSTM combination method is adopted to effectively reduce the complexity and non-stationarity of the original data on the prediction accuracy of the model, which can effectively improve the accuracy of short-term forecast of observed solar irradiance under sandstorm weather conditions.

[0100] refer to Figure 2 and Figure 6 The present disclosure also discloses a sandstorm weather solar irradiance combined forecasting system, the system comprising:

[0101] The data acquisition module 110 is used to acquire weather station observation data and sand and dust monitoring data;

[0102] A data input module 120, for inputting the meteorological station observation data and the dust monitoring data into a pre-trained solar irradiance forecast model;

[0103] A forecast submodule 130 is used to process the meteorological station observation data and dust monitoring data through the solar irradiance forecast model to construct multiple independent long-short term memory network sub-forecast models;

[0104] The irradiance prediction module 140 is used to reconstruct the forecast results of multiple long-term and short-term memory network sub-forecast models by accumulating and summing them to obtain the short-term forecast value of solar irradiance under sandstorm weather;

[0105] The solar irradiance forecast model takes the meteorological station observation data and the dust monitoring data as input, and performs variational mode decomposition on the pre-acquired solar irradiance observation values, which are then trained based on the long short-term memory network as output.

[0106] Among them, the solar irradiance forecast model takes the meteorological station observation data and dust monitoring data as input, and the pre-acquired solar irradiance observation values ​​are trained based on the long short-term memory network after variational mode decomposition as output, which specifically includes:

[0107] Obtain historical meteorological station observation data, dust monitoring data and solar irradiance observation time series;

[0108] Decomposing the solar irradiance observation time series by using a variational mode decomposition method, and preprocessing the meteorological station observation data and sand and dust monitoring data;

[0109] The preset long short-term memory network is trained based on the preprocessed meteorological station observation data and dust monitoring data as well as the solar irradiance observation time series after variational mode decomposition to obtain a multivariable and multi-step solar irradiance prediction model.

[0110] The solar irradiance observation time series is decomposed by a variational mode decomposition method, specifically including:

[0111] Based on the solar irradiance observation time series, a constrained variational problem is constructed by a variational mode decomposition method, and a constrained variational model is established;

[0112] Based on the constrained variational model, a penalty factor and a Lagrangian operator are introduced to convert the constrained variational problem into an unconstrained variational problem. The variational problem is solved by the multiplication operator alternating direction method to decompose and extract each eigenmode function and center frequency.

[0113] Preprocessing the meteorological station observation data and dust monitoring data includes:

[0114] Filling missing values ​​of the meteorological station observation data, converting visibility units, and setting the still wind observation data to zero;

[0115] The dust monitoring data is interpolated based on the latitude and longitude location information of the monitoring station to obtain the station dust score data.

[0116] Based on the preprocessed meteorological station observation data and dust monitoring data and the solar irradiance observation time series after variational mode decomposition, the preset long short-term memory network is trained to obtain a multivariate multi-step solar irradiance prediction model, which specifically includes:

[0117] The training and testing datasets are constructed based on the preprocessed meteorological station observation data and dust monitoring data as well as the solar irradiance observation time series after variational mode decomposition.

[0118] Based on the training data set, a preset long short-term memory network is trained, and multiple intrinsic mode functions obtained by variational mode decomposition of solar irradiance observation time series data are used as prediction labels;

[0119] Based on each intrinsic mode function, an independent long short-term memory network sub-forecast model is constructed, and the long short-term memory network sub-forecast models are combined and reconstructed to obtain a multivariable and multi-step solar irradiance prediction model.

[0120] The training and test datasets are constructed based on the preprocessed meteorological station observation data and dust monitoring data and the solar irradiance observation time series after variational mode decomposition, including:

[0121] Experiment on setting sliding window parameters of time series data based on solar irradiance observation time series after variational mode decomposition;

[0122] Through the preset grid search method, the window size is set according to the set value, the training effect and the test effect are compared, the optimal sliding window parameter value of the model is selected, and the construction of the training data set and the test data set is completed.

[0123] A combined forecast system for solar irradiance in dust weather provided by the present invention can effectively improve the prediction accuracy of solar irradiance by modeling and forecasting solar irradiance in sparse areas of PM2.5 and other atmospheric pollutant observation stations under dust weather conditions using a variational mode decomposition method and a long short-term memory network combination method. The system is based on the observation and forecast data of photovoltaic power stations or nearby meteorological observation stations, and gives full play to the advantages of imager dust monitoring DST data, such as wide coverage, high frequency, and continuous dust monitoring capabilities. A variational mode decomposition method and a long short-term memory network combination algorithm are used for modeling and training, and multi-step forecasts are made for short-term solar irradiance, so as to obtain the solar irradiance situation in the short term in the future under the influence of the current dust weather.

[0124] Figure 7 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 7 As shown, the electronic device may include: a processor (processor) 710, a communication interface (Communications Interface) 720, a memory (memory) 730 and a communication bus, wherein the processor 710, the communication interface 720, and the memory 730 communicate with each other through the communication bus. The processor 710 can call the logic instructions in the memory 730 to execute a combined forecasting method for solar irradiance in sandstorm weather, the method comprising: obtaining meteorological station observation data and sandstorm monitoring data; inputting the meteorological station observation data and sandstorm monitoring data into a pre-trained solar irradiance forecast model; processing the meteorological station observation data and sandstorm monitoring data through the solar irradiance forecast model to construct multiple independent long-short-term memory network sub-forecast models; reconstructing the forecast results of multiple long-short-term memory network sub-forecast models by accumulating and summing them to obtain the short-term forecast value of solar irradiance in sandstorm weather; wherein the solar irradiance forecast model takes the meteorological station observation data and sandstorm monitoring data as input, and trains the pre-acquired solar irradiance observation values ​​based on the long-short-term memory network as output after variational mode decomposition.

[0125] In addition, the logic instructions in the above-mentioned memory 730 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present disclosure is essentially 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, and the computer software product is stored in a storage medium, including a number of instructions to enable 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 method described in each embodiment of the present disclosure. 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 and other media that can store program codes.

[0126] On the other hand, the present disclosure also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute a combined forecasting method for solar irradiance in sandstorm weather provided by the above-mentioned methods, the method including: obtaining meteorological station observation data and sandstorm monitoring data; inputting the meteorological station observation data and sandstorm monitoring data into a pre-trained solar irradiance forecast model; processing the meteorological station observation data and sandstorm monitoring data through the solar irradiance forecast model to construct multiple independent long-short-term memory network sub-forecast models; reconstructing the forecast results of multiple long-short-term memory network sub-forecast models by accumulating and summing them to obtain the short-term forecast value of solar irradiance in sandstorm weather; wherein the solar irradiance forecast model takes the meteorological station observation data and sandstorm monitoring data as input, and trains the pre-acquired solar irradiance observation values ​​based on the long-short-term memory network as output after variational mode decomposition.

[0127] On the other hand, the present disclosure also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute a method for combined forecasting of solar irradiance in sandstorm weather provided by the above-mentioned methods, the method comprising: obtaining meteorological station observation data and sandstorm monitoring data; inputting the meteorological station observation data and sandstorm monitoring data into a pre-trained solar irradiance forecast model; processing the meteorological station observation data and sandstorm monitoring data through the solar irradiance forecast model to construct multiple independent long-short-term memory network sub-forecast models; reconstructing the forecast results of multiple long-short-term memory network sub-forecast models by accumulating and summing them to obtain a short-term forecast value of solar irradiance in sandstorm weather; wherein the solar irradiance forecast model is obtained by taking the meteorological station observation data and sandstorm monitoring data as input, and training the pre-acquired solar irradiance observation values ​​based on the long-short-term memory network as output after variational mode decomposition.

[0128] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0129] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than to limit them. Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present disclosure.

Claims

1. A combined forecasting method for solar irradiance in sandstorm weather, characterized in that: include: Obtain meteorological station observation data and sand and dust monitoring data; Inputting the meteorological station observation data and dust monitoring data into a pre-trained solar irradiance forecast model; The meteorological station observation data and dust monitoring data are processed by the solar irradiance forecast model to construct multiple independent long-short term memory network sub-forecast models; The forecast results of multiple long-short-term memory network sub-forecast models are accumulated and reconstructed to obtain the short-term forecast value of solar irradiance under sandstorm weather. The solar irradiance forecast model takes the meteorological station observation data and the dust monitoring data as input, and performs variational mode decomposition on the pre-acquired solar irradiance observation values, which are then trained based on the long short-term memory network as output.

2. The combined forecasting method for solar irradiance in sandstorm weather according to claim 1 is characterized in that: The solar irradiance forecast model is obtained by taking the meteorological station observation data and the dust monitoring data as input, and performing variational mode decomposition on the pre-acquired solar irradiance observation values ​​as output through training based on the long short-term memory network, specifically including: Obtain historical meteorological station observation data, dust monitoring data and solar irradiance observation time series; Decomposing the solar irradiance observation time series by using a variational mode decomposition method, and preprocessing the meteorological station observation data and sand and dust monitoring data; The preset long short-term memory network is trained based on the preprocessed meteorological station observation data and dust monitoring data as well as the solar irradiance observation time series after variational mode decomposition to obtain a multivariable and multi-step solar irradiance prediction model.

3. The combined forecasting method for solar irradiance in sandstorm weather according to claim 2 is characterized in that: Decomposing the solar irradiance observation time series by using a variational mode decomposition method specifically includes: Based on the solar irradiance observation time series, a constrained variational problem is constructed by a variational mode decomposition method, and a constrained variational model is established; Based on the constrained variational model, a penalty factor and a Lagrangian operator are introduced to convert the constrained variational problem into an unconstrained variational problem. The variational problem is solved by the multiplication operator alternating direction method to decompose and extract each eigenmode function and center frequency.

4. The combined forecasting method for solar irradiance in sandstorm weather according to claim 2 is characterized in that: The preprocessing of the meteorological station observation data and the dust monitoring data specifically includes: Filling missing values ​​of the meteorological station observation data, converting visibility units, and setting the still wind observation data to zero; The dust monitoring data is interpolated based on the latitude and longitude location information of the monitoring station to obtain the station dust score data.

5. The combined forecasting method of solar irradiance for sandstorm weather according to claim 2 is characterized in that: The method trains the preset long short-term memory network based on the pre-processed meteorological station observation data and dust monitoring data and the solar irradiance observation time series after variational mode decomposition to obtain a multivariate multi-step solar irradiance prediction model, which specifically includes: The training and testing datasets are constructed based on the preprocessed meteorological station observation data and dust monitoring data as well as the solar irradiance observation time series after variational mode decomposition. Based on the training data set, a preset long short-term memory network is trained, and multiple intrinsic mode functions obtained by variational mode decomposition of solar irradiance observation time series data are used as prediction labels; Based on each intrinsic mode function, an independent long short-term memory network sub-forecast model is constructed, and the long short-term memory network sub-forecast models are combined and reconstructed to obtain a multivariable and multi-step solar irradiance prediction model.

6. The combined forecasting method for solar irradiance in sandstorm weather according to claim 5 is characterized in that: The training data set and the test data set are constructed based on the pre-processed meteorological station observation data and dust monitoring data and the solar irradiance observation time series after variational mode decomposition, specifically including: Experiment on setting sliding window parameters of time series data based on solar irradiance observation time series after variational mode decomposition; Through the preset grid search method, the window size is set according to the set value, the training effect and the test effect are compared, the optimal sliding window parameter value of the model is selected, and the construction of the training data set and the test data set is completed.

7. A combined forecast system for solar irradiance in sandstorm weather, characterized in that: The system comprises: Data acquisition module, used to obtain meteorological station observation data and sand and dust monitoring data; A data input module, used for inputting the meteorological station observation data and dust monitoring data into a pre-trained solar irradiance forecast model; A forecast submodule, used to process the meteorological station observation data and dust monitoring data through the solar irradiance forecast model, and construct multiple independent long-short term memory network sub-forecast models; The irradiance prediction module is used to reconstruct the cumulative sum of the forecast results of multiple long-short-term memory network sub-forecast models to obtain the short-term forecast value of solar irradiance under sandstorm weather; The solar irradiance forecast model takes the meteorological station observation data and the dust monitoring data as input, and performs variational mode decomposition on the pre-acquired solar irradiance observation values, which are then trained based on the long short-term memory network as output.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the combined forecasting method of solar irradiance for sandstorm weather as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the combined forecasting method for solar irradiance in sandstorm weather as claimed in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the combined forecasting method for solar irradiance in sandstorm weather as claimed in any one of claims 1 to 6 is implemented.

Images