Photovoltaic power prediction method, device, equipment, storage medium and program product

Abstract

The present application relates to the technical field of power prediction, and discloses a photovoltaic power prediction method, device, equipment, storage medium and program product, the photovoltaic power prediction method comprises the following steps: constructing a meteorological time effect regression term based on the influence of low-frequency weather forecast data on high-frequency historical photovoltaic power data; constructing a power lag effect regression term based on the influence of high-frequency historical photovoltaic power data lag on high-frequency historical photovoltaic power data; combining the meteorological time effect regression term and the power lag effect regression term to construct a multivariate reverse MIDAS model; training the multivariate reverse MIDAS model using a data set to obtain a reverse MIDAS photovoltaic power prediction model; and performing photovoltaic power prediction based on the reverse MIDAS photovoltaic power prediction model. The present application can solve the problem of predicting high-frequency variables with low-frequency variables and improve prediction accuracy.

Description

Technical Field

[0001] This invention relates to the field of power prediction technology, specifically to photovoltaic power prediction methods, devices, equipment, storage media, and program products. Background Technology

[0002] In the process of forecasting photovoltaic power based on meteorological data and historical power data, the inconsistency of data time resolution is often encountered, which leads to the problem of mixed-frequency data analysis.

[0003] Existing analyses of mixed-frequency data often employ frequency-matching methods, which involve converting low-frequency data into high-frequency data through interpolation, or vice versa, for prediction purposes. This method disrupts the original data structure, preventing the use of the data's true structural information for prediction and thus affecting accuracy. Furthermore, when data is filled through interpolation, the effectiveness of the interpolation is difficult to measure when there are significant differences in temporal resolution.

[0004] To address the challenges of mixed-frequency data analysis, Ghysels et al. (2007) proposed the Mixed-Frequency Data Sampling (MIDAS) model, which effectively solves problems such as information loss, low prediction accuracy, and data processing bias caused by inconsistent data temporal resolution. However, this method is only applicable to the prediction of low-frequency variables based on high-frequency data, and cannot solve the problem of reverse-mixing data prediction of high-frequency power variables, which is often involved in photovoltaic power prediction, resulting in low accuracy of the final prediction results. Summary of the Invention

[0005] In view of this, the present invention provides a photovoltaic power prediction method, apparatus, equipment, storage medium and program product to solve the problem of predicting high-frequency variables from low-frequency variables and to improve prediction accuracy.

[0006] In a first aspect, the present invention provides a photovoltaic power prediction method, which includes: collecting a dataset, the dataset including high-frequency historical photovoltaic power data and low-frequency weather forecast data; using the high-frequency historical photovoltaic power data as a high-frequency dependent variable and the low-frequency weather forecast data as a low-frequency independent variable; constructing a meteorological time effect regression term based on the influence of the current period and lag period of the low-frequency weather forecast data on the current period of the high-frequency historical photovoltaic power data, and constructing a power lag effect regression term based on the influence of the lag period of the high-frequency historical photovoltaic power data on the current period of the high-frequency historical photovoltaic power data; combining the meteorological time effect regression term and the power lag effect regression term to construct a multivariate inverse MIDAS model; training the multivariate inverse MIDAS model using the dataset to obtain an inverse MIDAS photovoltaic power prediction model; and performing photovoltaic power prediction based on the inverse MIDAS photovoltaic power prediction model.

[0007] In this implementation, a reverse MIDAS model based on low-frequency data to predict high-frequency data is constructed, which can solve the problem of reverse mixing data prediction. Specifically, it considers the impact of the current period and lag period of low-frequency weather forecast data on the current period of high-frequency historical photovoltaic power data, and considers the impact of the lag period of high-frequency historical photovoltaic power data on the current period. It can effectively solve the problems of information loss, low prediction accuracy and data processing deviation caused by inconsistent data time resolution. Using the collected dataset to train the reverse MIDAS model can further improve the accuracy of power prediction.

[0008] In one optional implementation, a meteorological time effect regression term is constructed based on the impact of the current and lag periods of low-frequency weather forecast data on the current period of high-frequency historical photovoltaic power data. This includes: fusing low-frequency weather forecast data of multiple meteorological elements, a meteorological time effect weighting function, and a meteorological lag operator to obtain a meteorological time effect regression term; and a power lag effect regression term is constructed based on the impact of the lag period of high-frequency historical photovoltaic power data on the current period of high-frequency historical photovoltaic power data. This includes: fusing high-frequency historical photovoltaic power data and a power lag effect weighting function to obtain a power lag effect regression term.

[0009] In one optional implementation, a multivariate inverse MIDAS model is constructed by combining the meteorological time effect regression term and the power lag effect regression term, including: the structure of the multivariate inverse MIDAS model is as follows: ;in, This is high-frequency historical photovoltaic power data; The total number of meteorological elements. For the first Low-frequency weather forecast data corresponding to each meteorological element; the collection frequency of high-frequency historical photovoltaic power data is the same as the forecast frequency of the low-frequency weather forecast data. times; High-frequency historical photovoltaic power data Compared to low-frequency weather forecast data The number of forward prediction steps; The lag order of high-frequency historical photovoltaic power data; and These are the maximum lag order of low-frequency weather forecast data and high-frequency historical photovoltaic power data, respectively; among them... It is based on the meteorological time effect weighting function The order of sum is s Weather lag operator The polynomial function is composed of the following components: a meteorological time effect weighting function. Non-negative and satisfy Weather lag operator satisfy ;in, Let be a polynomial power lag effect weighting function, and satisfy . ;in, The parameter vector after reducing the space of parameters to be estimated; where, and For regression coefficients, This is the error term.

[0010] In this implementation, a clear multivariate inverse MIDAS model structure is constructed, in which a weight function is built to avoid the parameter inflation problem caused by the increase of lag order.

[0011] In one optional implementation, the weighting function includes an exponential Almon lag weighting function or a Beta lag weighting function, wherein the exponential Almon lag weighting function is: The Beta lag weighting function is: ;in, Let be the density function of the Beta distribution.

[0012] This implementation provides two weighting functions to choose from, allowing for the selection of a more suitable weighting function based on power prediction in different regions and environments, thereby improving the accuracy of power prediction.

[0013] In one optional implementation, training a multivariate inverse MIDAS model using a dataset to obtain a trained multivariate inverse MIDAS model includes: dividing the dataset into a training set and a test set according to a partitioning ratio; using the training set, performing parameter estimation on the multivariate inverse MIDAS model based on a nonlinear least squares method to obtain the trained multivariate inverse MIDAS model; wherein the objective function for parameter estimation is: The trained multivariate inverse MIDAS model was tested based on the test set to obtain the photovoltaic power prediction model of inverse MIDAS.

[0014] In this implementation, the nonlinear least squares method is used for model training within the framework of minimizing the sum of squared residuals, which can reduce the amount of computation and improve the efficiency of model training. Furthermore, the trained multivariate inverse MIDAS model is tested using a test set to further ensure the accuracy of the model.

[0015] In one optional implementation, the parameters of the multivariate inverse MIDAS model are estimated using a training set based on the nonlinear least squares method to obtain a trained multivariate inverse MIDAS model. This includes: obtaining multiple preset maximum lag orders; training a multivariate inverse MIDAS model corresponding to each preset maximum lag order and each weight function using the training set based on the BFGS algorithm to obtain multiple trained multivariate inverse MIDAS models; and testing the trained multivariate inverse MIDAS models based on a test set to obtain a photovoltaic power prediction model using inverse MIDAS. This includes: calculating the prediction error values ​​of the multiple trained multivariate inverse MIDAS models using the test set; and selecting the multivariate inverse MIDAS model with the smallest prediction error value to obtain the photovoltaic power prediction model using inverse MIDAS.

[0016] In this implementation, the influence of the maximum lag order and weight function on the model is comprehensively considered. The prediction accuracy of multiple models is judged using a test set. The inverse MIDAS model corresponding to the optimal lag order and weight function with the smallest error is selected as the photovoltaic power prediction model. This can obtain a prediction model that is more suitable for the current scenario and further improve the prediction accuracy of the model.

[0017] Secondly, the present invention provides a photovoltaic power prediction device, comprising: a data acquisition module for acquiring a dataset including high-frequency historical photovoltaic power data and low-frequency weather forecast data; a construction module for using high-frequency historical photovoltaic power data as a high-frequency dependent variable and low-frequency weather forecast data as a low-frequency independent variable, constructing a meteorological time effect regression term based on the influence of the current period and lag period of the low-frequency weather forecast data on the current period of the high-frequency historical photovoltaic power data, and constructing a power lag effect regression term based on the influence of the lag period of the high-frequency historical photovoltaic power data on the current period of the high-frequency historical photovoltaic power data; combining the meteorological time effect regression term and the power lag effect regression term to construct a multivariate inverse MIDAS model; a training module for training the multivariate inverse MIDAS model using the dataset to obtain an inverse MIDAS photovoltaic power prediction model; and a prediction module for predicting photovoltaic power based on the inverse MIDAS photovoltaic power prediction model.

[0018] Thirdly, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the photovoltaic power prediction method of the first aspect or any corresponding embodiment described above.

[0019] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the photovoltaic power prediction method of the first aspect or any corresponding embodiment thereof.

[0020] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to execute the photovoltaic power prediction method of the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0021] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0022] Figure 1 This is a flowchart of a photovoltaic power prediction method according to an embodiment of the present invention;

[0023] Figure 2 This is a flowchart of another photovoltaic power prediction method according to an embodiment of the present invention;

[0024] Figure 3 This is a structural block diagram of a photovoltaic power prediction device according to an embodiment of the present invention;

[0025] Figure 4 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] The meteorological data and power data used in photovoltaic power forecasting suffer from inconsistencies in time resolution. Existing methods employ frequency-matching processing, which has the following problems: it destroys the original data structure, failing to fully utilize the true information in the data; and the interpolation effect, achieved through low-frequency to high-frequency conversion, is difficult to measure. These issues may lead to lower final forecast accuracy. Therefore, this application proposes a photovoltaic power forecasting method to address the problem of predicting high-frequency power variables from low-frequency meteorological variables, thereby improving forecast accuracy.

[0028] According to an embodiment of the present invention, a photovoltaic power prediction method embodiment is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0029] This embodiment provides a photovoltaic power prediction method. Figure 1 This is a flowchart of a photovoltaic power prediction method according to an embodiment of the present invention. It should be noted that if substantially the same result is obtained, this embodiment does not necessarily reflect that result. Figure 1 The illustrated process sequence is limited. For example... Figure 1 As shown, the process includes the following steps:

[0030] Step S101: Collect the dataset.

[0031] The dataset includes high-frequency historical photovoltaic (PV) power data and low-frequency weather forecast data that influences PV power data. In one implementation, the meteorological elements in the low-frequency weather forecast data include solar irradiance, cloud cover, temperature, humidity, air pressure, wind speed, and wind direction. Generally, the time resolution of PV power data and weather forecast data differs; for example, the time resolution of PV power data is fifteen minutes, while the time resolution of the corresponding meteorological element forecast data in the weather forecast product is one hour or three hours, etc.

[0032] Specifically, this application collects high-frequency historical photovoltaic power data containing time series, and collects low-frequency weather forecast data containing time series corresponding to the meteorological elements of low-frequency weather forecast data to form a dataset.

[0033] For example, the time resolution of photovoltaic power data is fifteen minutes, and the time resolution of weather forecast data is one hour. The meteorological elements of the weather forecast data include five variables: temperature data, humidity data, air pressure data, wind speed data, and wind direction data. High-frequency historical photovoltaic power data is collected every fifteen minutes, and temperature data, humidity data, air pressure data, wind speed data, and wind direction data are collected every hour to form a dataset. The dataset contains high-frequency historical photovoltaic power data and its timestamp, as well as low-frequency weather forecast data and its timestamp.

[0034] Step S102: Using high-frequency historical photovoltaic power data as the high-frequency dependent variable and low-frequency weather forecast data as the low-frequency independent variable, a meteorological time effect regression term is constructed based on the impact of the current period and lag period of low-frequency weather forecast data on the current period of high-frequency historical photovoltaic power data. A power lag effect regression term is constructed based on the impact of the lag period of high-frequency historical photovoltaic power data on the current period of high-frequency historical photovoltaic power data. Combining the meteorological time effect regression term and the power lag effect regression term, a multivariate inverse MIDAS model is constructed.

[0035] This application divides high-frequency historical photovoltaic power data into current power and historical photovoltaic power. Specifically, if the high-frequency historical photovoltaic power data at a certain time stamp is determined as the current photovoltaic power, then all high-frequency historical photovoltaic power data prior to that time stamp are considered historical photovoltaic power, and historical photovoltaic power will affect the current photovoltaic power. Similarly, low-frequency weather forecast data prior to that time stamp will also affect the current photovoltaic power.

[0036] High-frequency historical photovoltaic power data is used as the high-frequency dependent variable in a multivariate inverse MIDAS model, while low-frequency weather forecast data is used as the low-frequency independent variable. Due to the different time resolutions and acquisition frequencies of the high-frequency historical photovoltaic power data and the low-frequency weather forecast data, a meteorological time-effect regression term is constructed based on the impact of the current and lag periods of the low-frequency weather forecast data on the current period of the high-frequency historical photovoltaic power data; a power lag effect regression term is constructed based on the impact of the lag period of the high-frequency historical photovoltaic power data on the current period of the high-frequency historical photovoltaic power data. The multivariate inverse MIDAS model is then constructed by combining the meteorological time-effect regression term and the power lag effect regression term.

[0037] Specifically, multiple low-frequency weather forecast data, both current and lagged, are introduced into the multivariate inverse MIDAS model. p The periodic regression term, as a meteorological time effect regression term, represents the time effect of low-frequency meteorological forecast data such as solar irradiance, cloud cover, temperature, humidity, air pressure, wind speed, and wind direction on the prediction of current photovoltaic power data; the inherent lag of high-frequency historical photovoltaic power data is also incorporated. l The regression term for the period serves as the power lag effect regression term, representing the lag effect of historical photovoltaic power data on current photovoltaic power.

[0038] Step S103: Use the dataset to train the multivariate inverse MIDAS model to obtain the photovoltaic power prediction model of inverse MIDAS.

[0039] Using the current photovoltaic power data from the high-frequency historical photovoltaic power data in the dataset as the data label, and the historical photovoltaic power data and low-frequency weather forecast data from the high-frequency historical photovoltaic power data as the data features, a multivariate inverse MIDAS model is trained to obtain the inverse MIDAS photovoltaic power prediction model.

[0040] Step S104: Photovoltaic power prediction is performed based on the photovoltaic power prediction model of reverse MIDAS.

[0041] The system acquires historical photovoltaic power data and weather forecast data for the current moment, inputs them into the photovoltaic power prediction model of the reverse MIDAS, and predicts the photovoltaic power for future moments.

[0042] The photovoltaic power prediction method provided in this embodiment constructs a reverse MIDAS model based on low-frequency data to predict high-frequency data, which can solve the problem of reverse mixing data prediction. Specifically, it considers the impact of the current period and lag period of low-frequency weather forecast data on the current period of high-frequency historical photovoltaic power data, and considers the impact of the lag period of high-frequency historical photovoltaic power data on the current period. It can effectively solve the problems of information loss, low prediction accuracy and data processing deviation caused by inconsistent data time resolution. By using the collected dataset to train the reverse MIDAS model, the accuracy of power prediction can be further improved.

[0043] This embodiment provides a photovoltaic power prediction method. Figure 2 This is a flowchart of another photovoltaic power prediction method according to an embodiment of the present invention. It should be noted that if substantially the same result is obtained, this embodiment does not necessarily use it as the standard. Figure 2 The illustrated process sequence is limited. For example... Figure 2 As shown, the process includes the following steps:

[0044] Step S201: Collect the dataset.

[0045] The dataset includes high-frequency historical photovoltaic power data and low-frequency weather forecast data.

[0046] Since high-frequency historical photovoltaic power data and low-frequency weather forecast data have different time resolutions, a corresponding acquisition frequency is determined. Specifically, the acquisition frequency for high-frequency historical photovoltaic power data is determined as the first acquisition frequency, and the forecast frequency for low-frequency weather forecast data is determined as the first forecast frequency. The first forecast frequency is less than the first acquisition frequency.

[0047] Specifically, step S201 includes:

[0048] Step S2011: Collect high-frequency historical photovoltaic power data according to the first acquisition frequency.

[0049] Specifically, photovoltaic power data is acquired in real time through a photovoltaic power station monitoring system, and high-frequency historical photovoltaic power data is collected from the photovoltaic power data according to the first acquisition frequency.

[0050] Step S2012: Based on the WRF model, the meteorological forecast product predicts low-frequency meteorological forecast data corresponding to multiple meteorological elements according to the first forecast frequency.

[0051] Specifically, the meteorological forecast products of the WRF model simulate atmospheric motion, turbulence and physical processes by solving atmospheric dynamics and thermodynamic equations, predict and generate meteorological forecast data with the same data variables as those in historical meteorological data, and predict low-frequency meteorological forecast data corresponding to multiple meteorological elements according to the first forecast frequency.

[0052] Furthermore, the collected dataset undergoes data preprocessing, such as missing value interpolation, outlier identification and handling, and data normalization.

[0053] Specifically, through clustering algorithms, 3- The dataset is analyzed using principles and methods to identify outliers; missing and outlier values ​​are imputed and replaced using statistical methods; and all variable sequence data are normalized. Clustering algorithms include K-means and DBSCAN, and negative historical photovoltaic power values ​​are identified as outliers. The statistical method used for missing value imputation and outlier handling is the spline method.

[0054] In this implementation, the collection frequency and forecast frequency of high-frequency historical photovoltaic power data and low-frequency weather forecast data are determined respectively to obtain uniformly distributed data, which facilitates subsequent model construction and training. At the same time, collecting low-frequency weather forecast data can take into account the impact of predicted meteorological data on power data, thereby improving the accuracy of subsequent model training.

[0055] Step S202: Using high-frequency historical photovoltaic power data as the high-frequency dependent variable and low-frequency weather forecast data as the low-frequency independent variable, a meteorological time effect regression term is constructed based on the impact of the current period and lag period of low-frequency weather forecast data on the current period of high-frequency historical photovoltaic power data. A power lag effect regression term is constructed based on the impact of the lag period of high-frequency historical photovoltaic power data on the current period of high-frequency historical photovoltaic power data. Combining the meteorological time effect regression term and the power lag effect regression term, a multivariate inverse MIDAS model is constructed.

[0056] By fusing low-frequency weather forecast data, weather time effect weighting functions, and weather lag operators from multiple meteorological elements, a weather time effect regression term is obtained; by fusing high-frequency historical photovoltaic power data and power lag effect weighting functions, a power lag effect regression term is obtained.

[0057] Specifically, the structure of the multivariate inverse MIDAS model is as follows:

[0058] .

[0059] in, This refers to high-frequency historical photovoltaic power data, that is, correspondingly, This is the current photovoltaic power data. This is historical photovoltaic power data. The total number of meteorological elements. For the first Low-frequency weather forecast data corresponding to each meteorological element; the collection frequency of high-frequency historical photovoltaic power data is the same as the forecast frequency of the low-frequency weather forecast data. times; High-frequency historical photovoltaic power data Compared to low-frequency weather forecast data The number of forward prediction steps; The lag order of high-frequency historical photovoltaic power data; and These represent the maximum lag order for low-frequency weather forecast data and high-frequency historical photovoltaic power data, respectively.

[0060] To avoid parameter inflation caused by increasing the lag order, a small number of parameters are used. Constructing weight functions The control parameter space has different parameters for each low-frequency weather forecast data. For the first Low-frequency weather forecast data High-frequency historical photovoltaic power data Time effect, It is based on the meteorological time effect weighting function The order of sum is s Weather lag operator The polynomial function is composed of the following components: a meteorological time effect weighting function. Non-negative and satisfy Weather lag operator satisfy .

[0061] Among them, the power lag effect weighting function using polynomials is used. The parameters to be estimated for the regression terms of high-frequency historical photovoltaic power data that are lagged by multiple periods. High-frequency historical photovoltaic power data High-frequency historical photovoltaic power data The lag effect, and satisfying ,in, This refers to the parameter vector after reducing the space of the parameters to be estimated.

[0062] in, and For regression coefficients, This is the error term.

[0063] In one implementation, the weighting function is an exponential Almon lag weighting function. Specifically, the exponential Almon lag weighting function is: .

[0064] In another implementation, the weighting function is a beta-lag weighting function. Specifically, the beta-lag weighting function is: ;in, Let be the density function of the Beta distribution.

[0065] This implementation constructs a clear multivariate inverse MIDAS model structure. The weighting function avoids parameter inflation caused by increasing lag order. Two weighting functions are provided for selection, allowing for the choice of a more suitable weighting function based on different regions and environments, thus improving power prediction accuracy.

[0066] Step S203: Use the dataset to train the multivariate inverse MIDAS model to obtain the photovoltaic power prediction model of inverse MIDAS.

[0067] Specifically, step S203 includes:

[0068] Step S2031: Divide the dataset into training set and test set according to the partition ratio.

[0069] Specifically, the dataset is divided into a training set and a test set in a 7:3 ratio.

[0070] Step S2032: Parameter estimation of the multivariate inverse MIDAS model is performed based on the nonlinear least squares method to obtain the trained multivariate inverse MIDAS model.

[0071] Specifically, based on the framework of minimizing the sum of squared errors, parameter estimation is performed using the nonlinear least squares method, where the objective function for parameter estimation is:

[0072] .

[0073] That is, within the framework of minimizing the sum of squared residuals, the initial values ​​of each parameter in the weight function are set, and the nonlinear least squares method is used for iteration to determine the estimated values ​​of each parameter in the weight function, thus obtaining the trained multivariate inverse MIDAS model.

[0074] In one implementation, multiple multivariate inverse MIDAS models are trained based on various weighting functions and various selectable lag periods. Specifically, step S2032 includes:

[0075] Step a1: Obtain multiple preset maximum lag orders.

[0076] For example, the maximum lag period is preset to be in the range of 1 to 5.

[0077] Step a2: Based on the BFGS algorithm, train a multivariate inverse MIDAS model corresponding to each preset maximum lag order and each weight function using the training set to obtain multiple trained multivariate inverse MIDAS models.

[0078] Among them, the BFGS (Broyden-Fletcher-Goldfarb-Shanno) algorithm is a quasi-Newton method for solving unconstrained optimization problems. It iteratively updates a positive definite matrix. (Approximately the inverse of the Hessian matrix), curvature information is constructed using gradient information. In each iteration, the change in the current parameters is used. and gradient change Update To make it satisfy the quasi-Newtonian condition The Hessian inverse approximation matrix update formula is as follows:

[0079] .

[0080] Multiple objective functions were constructed using different preset maximum lag orders and different weight functions, and multivariate inverse MIDAS models were trained using the BFGS algorithm on the training set for each function. This resulted in multiple corresponding trained multivariate inverse MIDAS models.

[0081] Step S2033: Test the trained multivariate inverse MIDAS model based on the test set to obtain the photovoltaic power prediction model of inverse MIDAS.

[0082] By comparing the impact of different weight functions and different lag orders on the model's fit on the test set, the optimal weight function and optimal lag order are determined. Specifically, step S2033 includes:

[0083] Step b1: Calculate the prediction error values ​​of multiple trained multivariate inverse MIDAS models using the test set.

[0084] Specifically, the prediction error of the trained multivariate inverse MIDAS model under different lag orders and weight functions is calculated using the test set, and the BIC value is calculated.

[0085] In one implementation, the prediction error is calculated by calculating the sum of squared residuals.

[0086] Step b2: Select the multivariate inverse MIDAS model with the smallest prediction error to obtain the photovoltaic power prediction model of inverse MIDAS.

[0087] Specifically, the lag order and weighting function corresponding to the minimum BIC value are selected to determine the corresponding multivariate reverse MIDAS model, thus obtaining the optimal reverse MIDAS photovoltaic power prediction model.

[0088] In this implementation, nonlinear least squares method is used for model training within the framework of minimizing the sum of squared residuals, which reduces computation and improves training efficiency. Furthermore, the trained multivariate inverse MIDAS model is tested using a test set to further ensure its accuracy. Considering the combined impact of the maximum lag order and weight function on the model, the prediction accuracy of multiple models is evaluated using the test set. The inverse MIDAS model with the minimum BIC (Best Imaging Capacity) corresponding to the optimal lag order and weight function is selected as the photovoltaic power prediction model. This yields a prediction model more suitable for the current scenario, further improving prediction accuracy.

[0089] Step S204: Perform photovoltaic power prediction based on the reverse MIDAS photovoltaic power prediction model.

[0090] The system acquires historical photovoltaic power data and weather forecast data for the current moment, inputs them into the photovoltaic power prediction model of the reverse MIDAS, and predicts the photovoltaic power for future moments.

[0091] The photovoltaic power prediction method provided in this embodiment constructs a reverse MIDAS model based on low-frequency data to predict high-frequency data. This effectively solves the problems of information loss, low prediction accuracy, and data processing deviation caused by inconsistent data time resolution, while also solving the problem of reverse mixing data prediction. By constructing a weight function, it can avoid the situation where the estimated parameters increase significantly due to the increase in the frequency of high-frequency variables and the increase in the lag order of the predicted variables. After training the reverse MIDAS model, further testing can be conducted to select the maximum lag order and weight function, thereby improving the accuracy of power prediction.

[0092] This embodiment also provides a photovoltaic power prediction device, which is used to implement the above embodiments and preferred embodiments, and will not be repeated as already described. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0093] This embodiment provides a photovoltaic power prediction device. Figure 3 This is a structural block diagram of a photovoltaic power prediction device according to an embodiment of the present invention, such as... Figure 3 As shown, the photovoltaic power prediction device includes:

[0094] The acquisition module 301 is used to acquire datasets, which include high-frequency historical photovoltaic power data and low-frequency weather forecast data.

[0095] Module 302 is used to construct a meteorological time effect regression term based on the impact of the current period and lag period of the low-frequency weather forecast data on the current period of the high-frequency historical photovoltaic power data, and a power lag effect regression term based on the impact of the lag period of the high-frequency historical photovoltaic power data on the current period of the high-frequency historical photovoltaic power data. Combining the meteorological time effect regression term and the power lag effect regression term, a multivariate inverse MIDAS model is constructed.

[0096] Training module 303 is used to train the multivariate inverse MIDAS model using the dataset to obtain the photovoltaic power prediction model of inverse MIDAS.

[0097] Prediction module 304 is used to predict photovoltaic power based on the reverse MIDAS photovoltaic power prediction model.

[0098] In some alternative implementations, the construction module 302 includes:

[0099] The first building unit is used to fuse low-frequency weather forecast data of multiple meteorological elements, meteorological time effect weighting functions, and meteorological lag operators to obtain meteorological time effect regression terms.

[0100] The second building block is used to fuse high-frequency historical photovoltaic power data and the power lag effect weighting function to obtain the power lag effect regression term.

[0101] In some alternative implementations, the construction module 302 includes:

[0102] Building blocks are used to construct multivariate inverse MIDAS models. The structure of a multivariate inverse MIDAS model is as follows: ;in, This is high-frequency historical photovoltaic power data; The total number of meteorological elements. For the first Low-frequency weather forecast data corresponding to each meteorological element; the collection frequency of high-frequency historical photovoltaic power data is the same as the forecast frequency of the low-frequency weather forecast data. times; High-frequency historical photovoltaic power data Compared to low-frequency weather forecast data The number of forward prediction steps; The lag order of high-frequency historical photovoltaic power data; and These are the maximum lag order of low-frequency weather forecast data and high-frequency historical photovoltaic power data, respectively; among them... It is based on the meteorological time effect weighting function The order of sum is s Weather lag operator The polynomial function is composed of the following components: a meteorological time effect weighting function. Non-negative and satisfy Weather lag operator satisfy ;in, Let be a polynomial power lag effect weighting function, and satisfy . ;in, The parameter vector after reducing the space of parameters to be estimated; where, and For regression coefficients, This is the error term.

[0103] In some alternative implementations, the building unit includes:

[0104] Construct sub-units to build weight functions, which include either exponential Almon lag weight functions or Beta lag weight functions. The exponential Almon lag weight function is as follows: The Beta lag weighting function is: ;in, Let be the density function of the Beta distribution.

[0105] In some alternative implementations, training module 303 includes:

[0106] The partitioning unit is used to divide the dataset into training and test sets according to the partitioning ratio.

[0107] The training unit is used to perform parameter estimation on a multivariate inverse MIDAS model using the training set and based on the nonlinear least squares method, to obtain the trained multivariate inverse MIDAS model; the objective function for parameter estimation is: .

[0108] The test unit is used to test the trained multivariate inverse MIDAS model based on the test set, so as to obtain the photovoltaic power prediction model of inverse MIDAS.

[0109] In some alternative implementations, the training unit includes:

[0110] Get sub-units, used to obtain multiple preset maximum lag periods.

[0111] The training subunit is used to train a multivariate inverse MIDAS model corresponding to each preset maximum lag order and each weight function based on the BFGS algorithm, thereby obtaining multiple trained multivariate inverse MIDAS models.

[0112] In some alternative implementations, the test unit includes:

[0113] The test subunit is used to calculate the prediction error values ​​of multiple trained multivariate inverse MIDAS models using the test set.

[0114] The sub-unit is selected to choose the multivariate inverse MIDAS model with the smallest prediction error, thus obtaining the inverse MIDAS photovoltaic power prediction model.

[0115] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0116] In this embodiment, the photovoltaic power prediction device is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0117] This invention also provides a computer device having the above-described features. Figure 4 The photovoltaic power prediction device shown.

[0118] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 4As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 4 Take a processor 10 as an example.

[0119] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0120] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0121] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0122] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0123] The computer device also includes an input device 30 and an output device 40. The processor 10, memory 20, input device 30, and output device 40 can be connected via a bus or other means. Figure 4 Taking the example of a connection between China and Israel via a bus.

[0124] Input device 30 can receive input numerical or character information, and generate key signal inputs related to user settings and function control of the computer device, such as a touchscreen, keypad, mouse, trackpad, touchpad, joystick, one or more mouse buttons, trackball, joystick, etc. Output device 40 may include display devices, auxiliary lighting devices (e.g., LEDs), and haptic feedback devices (e.g., vibration motors). The aforementioned display devices include, but are not limited to, liquid crystal displays, light-emitting diodes, displays, and plasma displays. In some alternative embodiments, the display device may be a touchscreen.

[0125] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0126] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0127] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A photovoltaic power prediction method, characterized in that, The method includes: Collect datasets, which include high-frequency historical photovoltaic power data and low-frequency weather forecast data; Using the high-frequency historical photovoltaic power data as the high-frequency dependent variable and the low-frequency weather forecast data as the low-frequency independent variable, a meteorological time effect regression term is constructed based on the impact of the current period and lag period of the low-frequency weather forecast data on the current period of the high-frequency historical photovoltaic power data. A power lag effect regression term is constructed based on the impact of the lag period of the high-frequency historical photovoltaic power data on the current period of the high-frequency historical photovoltaic power data. Combining the meteorological time effect regression term and the power lag effect regression term, a multivariate inverse MIDAS model is constructed. The construction of a meteorological time effect regression term based on the impact of the current period and lag period of the low-frequency meteorological forecast data on the current period of the high-frequency historical photovoltaic power data includes: fusing the low-frequency meteorological forecast data of multiple meteorological elements, the meteorological time effect weight function, and the meteorological lag operator to obtain the meteorological time effect regression term; The construction of a power lag effect regression term based on the impact of the lag period of the high-frequency historical photovoltaic power data on the current period of the high-frequency historical photovoltaic power data includes: fusing the high-frequency historical photovoltaic power data and the power lag effect weighting function to obtain the power lag effect regression term; The multivariate inverse MIDAS model is trained using the dataset to obtain the photovoltaic power prediction model of inverse MIDAS. Photovoltaic power prediction is performed based on the photovoltaic power prediction model of the reverse MIDAS.

2. The photovoltaic power prediction method according to claim 1, characterized in that, The multivariate inverse MIDAS model is constructed by combining the meteorological time effect regression term and the power lag effect regression term, including: The structure of the multivariate inverse MIDAS model is as follows: ； in, The high-frequency historical photovoltaic power data; The total number of meteorological elements. For the first Low-frequency weather forecast data corresponding to the aforementioned meteorological elements; the acquisition frequency of the high-frequency historical photovoltaic power data is the forecast frequency of the low-frequency weather forecast data. times; The high-frequency historical photovoltaic power data Compared to the low-frequency weather forecast data The number of forward prediction steps; The lag order of the high-frequency historical photovoltaic power data; and These are the maximum lag orders of the low-frequency weather forecast data and the high-frequency historical photovoltaic power data, respectively. in, It is determined by the meteorological time effect weighting function The order of sum is The meteorological lag operator The polynomial function composed of; wherein, the meteorological time effect weighting function Non-negative and satisfy The meteorological lag operator satisfy ; in, Let be the power lag effect weighting function of the polynomial, and satisfy . ;in, The parameter vector after reducing the space of parameters to be estimated; in, and For regression coefficients, This is the error term.

3. The photovoltaic power prediction method according to claim 2, characterized in that, The weighting function includes either the exponential Almon lag weighting function or the Beta lag weighting function, wherein... The Almon lag weighting function for the exponent is: ； The Beta lag weighting function is: ； in, Let be the density function of the Beta distribution.

4. The photovoltaic power prediction method according to claim 2, characterized in that, The step of training the multivariate inverse MIDAS model using the dataset to obtain the trained multivariate inverse MIDAS model includes: The dataset is divided into a training set and a test set according to the partitioning ratio; Using the training set, the parameters of the multivariate inverse MIDAS model are estimated based on the nonlinear least squares method to obtain the trained multivariate inverse MIDAS model; wherein, the objective function for parameter estimation is: ； The trained multivariate inverse MIDAS model was tested based on the test set to obtain the photovoltaic power prediction model of inverse MIDAS.

5. The photovoltaic power prediction method according to claim 4, characterized in that, The step of using the training set to estimate the parameters of the multivariate inverse MIDAS model based on the nonlinear least squares method to obtain the trained multivariate inverse MIDAS model includes: Obtain multiple preset maximum hysteresis orders; Based on the BFGS algorithm, the training set is used to train the multivariate inverse MIDAS model corresponding to each preset maximum lag order and each weight function, resulting in multiple trained multivariate inverse MIDAS models. The step of testing the trained multivariate inverse MIDAS model based on the test set to obtain an inverse MIDAS photovoltaic power prediction model includes: The prediction error values ​​of the multiple trained multivariate inverse MIDAS models are calculated using the test set. The multivariate inverse MIDAS model with the smallest prediction error is selected to obtain the photovoltaic power prediction model of the inverse MIDAS.

6. A photovoltaic power prediction device, characterized in that, The device includes: The data acquisition module is used to acquire datasets, which include high-frequency historical photovoltaic power data and low-frequency weather forecast data. The construction module is used to take the high-frequency historical photovoltaic power data as the high-frequency dependent variable and the low-frequency weather forecast data as the low-frequency independent variable, construct a meteorological time effect regression term based on the impact of the current period and lag period of the low-frequency weather forecast data on the current period of the high-frequency historical photovoltaic power data, and construct a power lag effect regression term based on the impact of the lag period of the high-frequency historical photovoltaic power data on the current period of the high-frequency historical photovoltaic power data; and combine the meteorological time effect regression term and the power lag effect regression term to construct a multivariate inverse MIDAS model. The construction of a meteorological time effect regression term based on the impact of the current period and lag period of the low-frequency meteorological forecast data on the current period of the high-frequency historical photovoltaic power data includes: fusing the low-frequency meteorological forecast data of multiple meteorological elements, the meteorological time effect weight function, and the meteorological lag operator to obtain the meteorological time effect regression term; The construction of a power lag effect regression term based on the impact of the lag period of the high-frequency historical photovoltaic power data on the current period of the high-frequency historical photovoltaic power data includes: fusing the high-frequency historical photovoltaic power data and the power lag effect weighting function to obtain the power lag effect regression term; The training module is used to train the multivariate inverse MIDAS model using the dataset to obtain the photovoltaic power prediction model of inverse MIDAS. The prediction module is used to predict photovoltaic power based on the photovoltaic power prediction model of the reverse MIDAS.

7. A computer device, characterized in that, include: A memory and a processor are interconnected, the memory storing computer instructions, and the processor executing the computer instructions to perform the photovoltaic power prediction method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the photovoltaic power prediction method according to any one of claims 1 to 5.

9. A computer program product, characterized in that, It includes computer instructions for causing a computer to execute the photovoltaic power prediction method according to any one of claims 1 to 5.

Images