Photovoltaic power longitudinal error correction method and device

Abstract

The invention relates to a photovoltaic power longitudinal error correction method and device, and the method comprises the steps: determining a to-be-corrected time period of a to-be-corrected predicted power according to historical output data; based on the power predicted value of the to-be-corrected time period in the next sampling period, predicting an error type of the to-be-corrected time period by using a pre-established error probability model; based on the weather related information of the to-be-corrected time period in the next sampling period, a power error prediction value of the to-be-corrected time period in the next sampling period is obtained through prediction by using a pre-established error correction model; and according to the error type corresponding to the to-be-corrected time period, determining a corrected power prediction value of the to-be-corrected time period in the next sampling period by using the power error prediction value of the to-be-corrected time period in the next sampling period. According to the technical scheme provided by the invention, the distribution characteristics of the errors can be more accurately understood, the efficient correction of the power prediction longitudinal errors is realized, and the accuracy and credibility of the power prediction data are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of new energy power generation, and particularly relates to a method and device for correcting the longitudinal error of photovoltaic power. Background Art

[0002] With the deepening and promotion of the dual-carbon policy, the proportion of photovoltaic installations in new energy installations is increasing day by day, and distributed photovoltaic installations are developing particularly rapidly. Due to the influence of weather and climate conditions, the output of photovoltaic power will change at different times and locations, and this instability makes the prediction of new energy power complex. At the same time, photovoltaic power prediction usually relies on mathematical models, which use historical data and weather forecasts to predict future power output. However, these models usually have difficulty accurately simulating the volatility of renewable energy, so there are often longitudinal errors. By correcting the longitudinal deviation of the model to make it more consistent with the actual observed data, the volatility and instability of renewable energy can be better addressed, and the uncertainty of the power system can be reduced.

[0003] Currently, there are many studies on prediction error correction. The main technical routes include: data-driven methods: using the difference between historical observation data and model predictions, and adopting data-driven methods such as regression analysis, machine learning, and time series analysis to correct prediction errors. Model fusion: combining multiple different types of models, such as physical models and statistical models, to obtain more accurate predictions. Deep learning and neural networks: using deep learning and neural network technologies can process a large amount of complex data to improve power prediction.

[0004] However, the above methods have the following disadvantages: (1) Data dependence: Data-driven methods highly depend on the available historical data. If the data is insufficient or contains noise, the accuracy of the model may be affected; (2) Complexity: Model fusion usually requires managing multiple models and their integration methods, increasing the complexity of algorithm implementation; (3) Computational resources: Deep learning models usually require a large amount of computational resources, including high-performance GPUs, for training and real-time prediction; (4) Interpretability: Deep learning models usually lack interpretability and it is difficult to understand their internal working methods; (5) Correction efficiency: The existing technologies correct the photovoltaic power prediction errors for the entire time period, lacking pertinence and having poor timeliness of error correction. Summary of the Invention

[0005] To overcome the problems existing in the above related technologies, the present application provides a method and device for correcting the longitudinal error of photovoltaic power.

[0006] According to the first aspect of the embodiments of the present application, a method for correcting the longitudinal error of photovoltaic power is provided, including:

[0007] Determine a to-be-corrected period for which the predicted power needs to be corrected according to historical output data;

[0008] Based on the power prediction value of the to-be-corrected period in the next sampling period, using the pre-established error probability model, predict the error type of the to-be-corrected period;

[0009] Based on the meteorological-related information of the to-be-corrected period in the next sampling period, using the pre-established error correction model, predict the power error prediction value of the to-be-corrected period in the next sampling period;

[0010] According to the error type corresponding to the to-be-corrected period, using the power error prediction value of the to-be-corrected period in the next sampling period, determine the corrected power prediction value of the to-be-corrected period in the next sampling period;

[0011] The error probability model is constructed using the historical output data, and the error correction model is constructed using the historical output data.

[0012] Preferably, the historical output data includes:

[0013] The predicted power values of each period in the historical sampling period and the measured power values of each period in the historical sampling period.

[0014] Preferably, the establishment process of the error probability model includes:

[0015] Using the measured power values of the to-be-corrected period in the historical sampling period, calculate the mean value and standard deviation of the measured power of the to-be-corrected period in the historical sampling period;

[0016] Based on the mean value and standard deviation of the measured power of the to-be-corrected period in the historical sampling period, establish the error probability model using the probability density function.

[0017] Preferably, the error probability model includes:

[0018]

[0019] In the above formula, f(x,μ,σ) is the probability density function; x is the random variable, that is, the predicted power value of the to-be-corrected period; μ is the mean value of the measured power of the to-be-corrected period in the historical sampling period, and σ is the standard deviation of the measured power of the to-be-corrected period in the historical sampling period.

[0020] Preferably, the establishment process of the error correction model includes:

[0021] Collect the meteorological-related information of each period in the historical sampling period;

[0022] Using the predicted power values for each time period in the historical sampling period and the measured power values for each time period in the historical sampling period, calculate the actual power error values for each time period in the historical sampling period;

[0023] Construct a data set using the meteorological-related information for each time period in the historical sampling period and the actual power error values for each time period in the historical sampling period;

[0024] Divide the data set into a training set and a validation set;

[0025] Train a support vector machine model using the training set to obtain a trained support vector machine model;

[0026] Validate the trained support vector machine model using the validation set. When the validation is successful, the trained support vector machine model is the error correction model.

[0027] Preferably, the training of the support vector machine model using the training set to obtain a trained support vector machine model includes:

[0028] Use the meteorological-related information for each time period in the historical sampling period in the training set as the input layer training samples of the support vector machine model, and use the actual power error values for each time period in the historical sampling period in the training set as the output layer training samples of the support vector machine model to train the support vector machine model, obtaining a trained support vector machine model.

[0029] Preferably, the validation of the trained support vector machine model using the validation set includes:

[0030] Use the meteorological-related information for each time period in the historical sampling period in the validation set as the input of the trained support vector machine model, and output the predicted power error values for each time period in the historical sampling period;

[0031] Based on the predicted power error values for each time period in the historical sampling period and the actual power error values for each time period in the historical sampling period in the validation set, determine the accuracy rate of the trained support vector machine model;

[0032] If the accuracy rate of the trained support vector machine model is greater than or equal to the accuracy rate threshold, the validation is successful, and the trained support vector machine model is the error correction model; if the accuracy rate of the trained support vector machine model is less than the accuracy rate threshold, the validation fails, adjust the parameters of the support vector machine model, and retrain the support vector machine model until the validation is successful.

[0033] Preferably, the determination of the time periods to be corrected for the predicted power according to the historical output data includes:

[0034] Using the predicted power values for each time period in the historical sampling period and the measured power values for each time period in the historical sampling period, calculate the actual power error values for each time period in the historical sampling period;

[0035] According to the actual power error values for each time period in the historical sampling period, calculate the average error probability for each time period in the historical sampling period;

[0036] Let the time period corresponding to the average error probability greater than the first threshold be the time period to be corrected.

[0037] Preferably, based on the power prediction value for the time period to be corrected in the next sampling period, using a pre-established error probability model, predict the error type of the time period to be corrected, including:

[0038] Taking the power prediction value for the time period to be corrected in the next sampling period as the input of the error probability model, determine the probability density function value corresponding to the power prediction value for the time period to be corrected in the next sampling period;

[0039] If the probability density function value is greater than or equal to the second threshold, the error type of the time period to be corrected is a positive error; if the probability density function value is less than the second threshold, the error type of the time period to be corrected is a negative error.

[0040] Preferably, the positive error is: the power prediction value is greater than the power measured value;

[0041] The negative error is: the power prediction value is less than the power measured value.

[0042] Preferably, based on the meteorological-related information for the time period to be corrected in the next sampling period, using a pre-established error correction model, predict the power error prediction value for the time period to be corrected in the next sampling period, including:

[0043] Taking the meteorological-related information for the time period to be corrected in the next sampling period as the input of the error correction model, output the power error prediction value for the time period to be corrected in the next sampling period.

[0044] Preferably, according to the error type corresponding to the time period to be corrected, using the power error prediction value for the time period to be corrected in the next sampling period, determine the corrected power prediction value for the time period to be corrected in the next sampling period, including:

[0045] When the error type corresponding to the period to be corrected is a positive error, the power prediction value of the period to be corrected in the next sampling period minus the power error prediction value of the period to be corrected in the next sampling period, to obtain the corrected power prediction value of the period to be corrected in the next sampling period;

[0046] When the error type corresponding to the period to be corrected is a negative error, the power prediction value of the period to be corrected in the next sampling period plus the power error prediction value of the period to be corrected in the next sampling period, to obtain the corrected power prediction value of the period to be corrected in the next sampling period.

[0047] Preferably, before obtaining the power error prediction value of the period to be corrected in the next sampling period by using the pre-established error correction model based on the meteorological-related information of the period to be corrected in the next sampling period, further comprising:

[0048] Collect the meteorological-related information of the period to be corrected in the next sampling period.

[0049] Preferably, the meteorological-related information includes: season data and meteorological data.

[0050] According to a second aspect of the embodiments of the present application, there is provided a photovoltaic power longitudinal error correction device, comprising:

[0051] A determination unit, configured to determine a period to be corrected for the predicted power to be corrected according to historical output data;

[0052] A first prediction unit, configured to predict the error type of the period to be corrected by using a pre-established error probability model based on the power prediction value of the period to be corrected in the next sampling period;

[0053] A second prediction unit, configured to predict the power error prediction value of the period to be corrected in the next sampling period by using a pre-established error correction model based on the meteorological-related information of the period to be corrected in the next sampling period;

[0054] A correction unit, configured to determine the corrected power prediction value of the period to be corrected in the next sampling period according to the error type corresponding to the period to be corrected and by using the power error prediction value of the period to be corrected in the next sampling period;

[0055] The error probability model is constructed by using the historical output data, and the error correction model is constructed by using the historical output data.

[0056] According to a third aspect of the embodiments of the present application, there is provided a computer device, comprising: one or more processors;

[0057] The processor is used to store one or more programs;

[0058] When the one or more programs are executed by the one or more processors, the photovoltaic power longitudinal error correction method described above is implemented.

[0059] According to the fourth aspect of the embodiments of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed, the photovoltaic power longitudinal error correction method described above is implemented.

[0060] The technical solution provided by the present invention has the following beneficial effects:

[0061] The present invention provides a photovoltaic power longitudinal error correction method and device, including: determining a to-be-corrected period for correcting the predicted power according to historical output data; predicting the error type of the to-be-corrected period by using a pre-established error probability model based on the power prediction value of the to-be-corrected period in the next sampling period; predicting the power error prediction value of the to-be-corrected period in the next sampling period by using a pre-established error correction model based on the meteorological-related information of the to-be-corrected period in the next sampling period; and determining the corrected power prediction value of the to-be-corrected period in the next sampling period according to the error type corresponding to the to-be-corrected period and by using the power error prediction value of the to-be-corrected period in the next sampling period. By selecting the to-be-corrected period for correcting the predicted power, the present invention can more effectively cope with the changes faced by photovoltaic in different periods, improve the pertinence and accuracy of power prediction, and thus eliminate or reduce the possible systematic and random longitudinal errors in the prediction data; by determining the corrected power prediction value of the to-be-corrected period in the next sampling period according to the error type corresponding to the to-be-corrected period and by using the power error prediction value of the to-be-corrected period in the next sampling period, the distribution characteristics of the errors can be more accurately understood, the efficient correction of the longitudinal error of the power prediction is realized, the accuracy and credibility of the power prediction data are significantly improved. At the same time, this customized correction method makes this technical route more applicable and can better meet the needs of specific application fields for high-quality power prediction data. Description of the Drawings

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0063] Figure 1 It is a flowchart of a photovoltaic power longitudinal error correction method provided by an embodiment of the present invention;

[0064] Figure 2 It is a schematic diagram of the error during the noon peak period provided by an embodiment of the present invention;

[0065] Figure 3 It is a schematic diagram of the longitudinal error correction effect provided by an embodiment of the present invention;

[0066] Figure 4 It is a structural block diagram of a photovoltaic power longitudinal error correction device provided by an embodiment of the present invention. Detailed implementation manners

[0067] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Apparently, the following embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention fall within the scope of protection of the present invention.

[0068] Embodiment 1

[0069] The present invention provides a method for correcting the longitudinal error of photovoltaic power, as Figure 1 shown, including the following steps:

[0070] Step 101: Determine the period to be corrected for the predicted power to be corrected according to historical output data;

[0071] Step 102: Based on the power prediction value of the period to be corrected in the next sampling period, use the pre-established error probability model to predict the error type of the period to be corrected;

[0072] Step 103: Based on the meteorological-related information of the period to be corrected in the next sampling period, use the pre-established error correction model to predict the power error prediction value of the period to be corrected in the next sampling period;

[0073] Step 104: According to the error type corresponding to the period to be corrected, use the power error prediction value of the period to be corrected in the next sampling period to determine the corrected power prediction value of the period to be corrected in the next sampling period;

[0074] The error probability model is constructed using historical output data, and the error correction model is constructed using historical output data.

[0075] Further, before step 103, it further includes:

[0076] Collect the meteorological-related information of the period to be corrected in the next sampling period.

[0077] It is understandable that the sampling period can be, but is not limited to, 15 minutes, and there are 96 sampling periods in a day. The present invention can correct the power prediction value for the next sampling period, and by the same method, the power prediction values for multiple future sampling periods can also be corrected.

[0078] The present invention significantly improves the accuracy and reliability of power prediction data by performing vertical error correction for specific key periods (i.e., the periods to be corrected for predicted power). The present invention selects key periods for photovoltaic power prediction and systematically analyzes, models, and corrects the errors in the power prediction data to eliminate or reduce the possible systematic and random vertical errors in the prediction data.

[0079] The present invention focuses on specific power prediction requirements. By constructing and classifying an error data set, it specifically models and corrects for vertical error types. This customized correction method makes this technical route more applicable and can better meet the needs of specific application fields for high-quality power prediction data.

[0080] Furthermore, the historical output data includes:

[0081] The predicted power values for each period in the historical sampling period and the measured power values for each period in the historical sampling period.

[0082] Error probability modeling is an important part of vertical error correction. By establishing a probability distribution model of the error, it is possible to more comprehensively understand and describe the nature of the prediction error. The modeling of the probability density function (PDF) is to calculate the parameters of the probability distribution model using the existing power prediction error data. Furthermore, the process of establishing the error probability model includes:

[0083] Using the measured power values for the periods to be corrected in the historical sampling period, calculate the mean measured power and the standard deviation of the measured power for the periods to be corrected in the historical sampling period;

[0084] Based on the mean measured power and the standard deviation of the measured power for the periods to be corrected in the historical sampling period, establish an error probability model using the probability density function.

[0085] Specifically, the error probability model includes:

[0086]

[0087] In the above formula, f(x,μ,σ) is the probability density function; x is the random variable, that is, the predicted power value for the period to be corrected; μ is the mean measured power for the period to be corrected in the historical sampling period, and σ is the standard deviation of the measured power for the period to be corrected in the historical sampling period.

[0088] It should be noted that the probability density function expresses the relative probability of the random variable x occurring under the given mean and standard deviation. The normal distribution is a symmetric distribution, and its shape is jointly determined by the mean and the standard deviation. Therefore, it is most appropriate to use the probability density function of the normal distribution (Gaussian distribution) to represent the probability distribution of the longitudinal error in the period to be corrected.

[0089] In the present invention, by comparing the probability density function values at different x values, the relative likelihood of different values under the given distribution is calculated, that is, the error probabilities of different error types (positive and negative) are calculated. In the normal distribution, the maximum value of the probability density function appears at the mean μ, and as x moves farther away from the mean, the probability density gradually decreases.

[0090] By establishing an error probability model in the present invention, the system can more accurately understand the distribution characteristics of the errors, and realizes the efficient correction of the longitudinal error of power prediction. This makes the method not only improve the accuracy of power prediction, but also demonstrates the full application of advanced technologies.

[0091] Furthermore, the establishment process of the error correction model includes:

[0092] Collect the meteorological-related information of each period in the historical sampling period;

[0093] Using the predicted power values of each period in the historical sampling period and the measured power values of each period in the historical sampling period, calculate the actual power error values of each period in the historical sampling period;

[0094] Construct a data set using the meteorological-related information of each period in the historical sampling period and the actual power error values of each period in the historical sampling period;

[0095] Divide the data set into a training set and a validation set;

[0096] Use the training set to train the support vector machine model to obtain the trained support vector machine model;

[0097] Use the validation set to validate the trained support vector machine model. When the validation is successful, the trained support vector machine model is the error correction model.

[0098] Furthermore, the meteorological-related information includes: season data and meteorological data.

[0099] In some embodiments, the season data may but is not limited to include: seasons (i.e., spring, summer, autumn, winter), the twenty-four solar terms, etc.; the meteorological data may but is not limited to include: temperature, air pressure, humidity, precipitation, evaporation, wind direction, wind speed, sunshine, etc.

[0100] It should be noted that based on the established error probability model, the present invention uses a Support Vector Machine (SVM) correction algorithm to perform longitudinal error correction on power prediction data. The correction process can effectively correct the prediction data accurately according to different error types, improving the accuracy and reliability of the data.

[0101] SVM is a supervised learning algorithm that is widely used in regression and classification problems. In the longitudinal error correction of power prediction, the power prediction error can be used as the target variable, and meteorological conditions, time, seasonal information, etc. can be used as feature variables, and SVM is used to learn the relationship of systematic errors. SVM separates different categories of samples by finding an optimal hyperplane, thereby realizing the correction of systematic errors and making the systematic errors best fit in the feature space.

[0102] Based on the results of error probability modeling analysis, select the features for training SVM. These features include meteorological data, time, seasonal information, etc., which specifically depend on the nature of the data and the requirements of the application. The selection of features should take into account their impact on systematic errors.

[0103] Since photovoltaic power prediction has strong seasonal variation characteristics, that is, the midday peak and the sunrise and sunset times are greatly affected by seasonal changes, select seasonal information, meteorological data (temperature, humidity, air pressure) and other conventional meteorological information as features for the construction of the SVM model training.

[0104] The mathematical form of SVM is:

[0105]

[0106] Satisfy the constraint condition: y i (w·x i +b)≥1-ξ i

[0107] Among them, i∈[1,n], n is the total number of sample points, w is the weight vector (customized adjustment), b is the bias term (customized adjustment), C is the regularization parameter (constant value), ξ i is the slack variable of the i-th sample point (the degree of deviation allowed for the sample point), x i is the feature vector of the i-th sample point, y i is the corresponding class label of the i-th sample point (customized adjustment).

[0108] Use the existing power prediction error data to train the SVM model. By adjusting the weight vector w and the bias term b, and selecting an appropriate regularization parameter C, the systematic error can be reduced as much as possible, and the final parameter indicators can be obtained.

[0109] Prediction of systematic error: The trained SVM model can be used to predict new power prediction errors. For the features at each timestamp (i.e., the sampling period, which can be but is not limited to 15 minutes), the model outputs an estimated value of the systematic error.

[0110] Longitudinal error correction: Correct the power prediction error. Add the estimated value of the systematic error output by the SVM to the original power prediction value to obtain the corrected power prediction value. Corrected predicted power = original predicted power + estimated value of the systematic error output by the SVM.

[0111] Model evaluation and tuning: To evaluate the performance of the SVM model, methods such as cross-validation can be used. According to the actual results, the model is tuned, which may involve adjusting the hyperparameter C, selecting different kernel functions, etc.

[0112] Through features such as meteorological data (wind speed, temperature), time, and season data, by constructing an SVM model, the present invention can obtain an estimated value of the systematic error for the period to be corrected (i.e., if the model outputs an estimated value of the systematic error of 3%, the corrected power prediction value is the original prediction value plus a 3% correction). This process can learn the complex relationship of the systematic error through the SVM, thereby making a more accurate longitudinal correction of the power prediction error.

[0113] The present invention establishes a complete error data set, and pays particular attention to the diversity of power prediction errors during the construction stage of the error data set, by carefully classifying the errors under different conditions. This helps to construct a more complete and specific error data set, providing a more sufficient basis for subsequent probability modeling and correction.

[0114] Furthermore, use the training set to train the support vector machine model to obtain the trained support vector machine model, including:

[0115] Use the meteorological-related information of each period in the historical sampling period of the training set as the training samples of the input layer of the support vector machine model, and use the actual power error values of each period in the historical sampling period of the training set as the training samples of the output layer of the support vector machine model to train the support vector machine model to obtain the trained support vector machine model.

[0116] Furthermore, use the validation set to validate the trained support vector machine model, including:

[0117] Use the meteorological-related information of each period in the historical sampling period of the validation set as the input of the trained support vector machine model, and output the predicted values of the power error for each period in the historical sampling period;

[0118] Based on the predicted power error values for each time period in the historical sampling period and the actual power error values for each time period in the historical sampling period in the validation set, determine the accuracy rate of the trained support vector machine model;

[0119] If the accuracy rate of the trained support vector machine model is greater than or equal to the accuracy rate threshold, the verification is successful, and the trained support vector machine model is an error correction model; if the accuracy rate of the trained support vector machine model is less than the accuracy rate threshold, the verification fails, adjust the parameters of the support vector machine model, and retrain the support vector machine model until the verification is successful.

[0120] Further, step 101 includes:

[0121] Step 1011: Use the predicted power values for each time period in the historical sampling period and the measured power values for each time period in the historical sampling period to calculate the actual power error values for each time period in the historical sampling period;

[0122] Step 1012: According to the actual power error values for each time period in the historical sampling period, calculate the average error probability for each time period in the historical sampling period;

[0123] Step 1013: Let the time periods corresponding to the average error probability greater than the first threshold be the time periods to be corrected.

[0124] It should be noted that the embodiments of the present invention do not limit the "first threshold", which can be selected by those skilled in the art according to engineering needs or experimental data, etc.

[0125] In some embodiments, it is possible but not limited to perform mathematical analysis on historical output data to identify time periods with significant power changes or special events. Through trend analysis, determine the possible systematic errors within a specific time period. By analyzing the power prediction scenarios of photovoltaic, analyze historical data, and find that there are significant changes in solar irradiance in the morning (8:00 - 10:00) and evening (16:00 - 19:00), and the probability of power prediction errors is relatively large during the midday peak period (11:00 - 14:00).

[0126] Calculate and statistically analyze the actual power error values for each time period in the past 6 months, as shown in Table 1. The photovoltaic power prediction error during the midday period is greater than that during the morning and evening periods, and the need for error correction is greater.

[0127] Table 1 Actual power error values for each time period in the past 6 months

[0128]

[0129] Determine the first threshold according to the assessment requirements for the prediction accuracy of photovoltaic new energy. According to the determined first threshold, the period to be corrected for the predicted power can be determined as the midday peak period. Therefore, select the midday peak period (11:00 - 14:00) as the period to be corrected for error correction.

[0130] In the present invention, through the selection of the period to be corrected for power prediction, this process combines various key factors, including meteorological conditions, time, etc., making the selected period more accurate and targeted. Compared with traditional methods, this precise selection of the period to be corrected can more effectively cope with the changes faced by photovoltaics at different times, improving the pertinence and accuracy of power prediction.

[0131] Furthermore, based on the power prediction value of the period to be corrected in the next sampling period, using the pre-established error probability model, predict the error type of the period to be corrected, including:

[0132] Take the power prediction value of the period to be corrected in the next sampling period as the input of the error probability model, and determine the probability density function value corresponding to the power prediction value of the period to be corrected in the next sampling period;

[0133] If the probability density function value is greater than or equal to the second threshold, the error type of the period to be corrected is a positive error; if the probability density function value is less than the second threshold, the error type of the period to be corrected is a negative error.

[0134] Furthermore, a positive error means: the power prediction value is greater than the measured power value;

[0135] A negative error means: the power prediction value is less than the measured power value.

[0136] It should be noted that the embodiments of the present invention do not limit the "second threshold", and it can be selected by those skilled in the art according to engineering needs or experimental data, etc.

[0137] For example, select the midday peak period (11:00 - 14:00) as the period to be corrected for error correction. During the selected period to be corrected (11:00 - 14:00), collect and organize the error data related to the monitoring data, and classify the error data according to its nature, that is, classify the errors into positive errors and negative errors. This helps to better understand the impact of different errors on the prediction data and provides detailed classification information for the subsequent longitudinal error correction steps.

[0138] A positive error means that the midday predicted peak exceeds the midday measured peak, that is, positive error = midday predicted peak - midday measured peak > 0; a negative error means that the midday predicted peak is lower than the midday measured peak, that is, negative error = midday predicted peak - midday measured peak < 0.

[0139] Further, based on the meteorological-related information in the next sampling period for the period to be corrected, using the pre-established error correction model, the predicted power error value for the period to be corrected in the next sampling period is obtained, including:

[0140] Taking the meteorological-related information in the next sampling period for the period to be corrected as the input of the error correction model, and outputting the predicted power error value for the period to be corrected in the next sampling period.

[0141] Further, according to the error type corresponding to the period to be corrected, using the predicted power error value for the period to be corrected in the next sampling period, the corrected power prediction value for the period to be corrected in the next sampling period is determined, including:

[0142] When the error type corresponding to the period to be corrected is a positive error, subtract the predicted power error value for the period to be corrected in the next sampling period from the power prediction value for the period to be corrected in the next sampling period to obtain the corrected power prediction value for the period to be corrected in the next sampling period;

[0143] When the error type corresponding to the period to be corrected is a negative error, add the predicted power error value for the period to be corrected in the next sampling period to the power prediction value for the period to be corrected in the next sampling period to obtain the corrected power prediction value for the period to be corrected in the next sampling period.

[0144] To further illustrate the above-mentioned longitudinal error correction method for photovoltaic power, the present invention provides a specific example as follows:

[0145] Select the measured and short-term predicted output data of the whole society's photovoltaic power from June 19, 2023 to September 18, 2023, for 90 days as sample training data. Conduct error analysis and modeling on the output and prediction data of the whole society's photovoltaic new energy, and calculate the error (measured output - predicted output) during the midday peak period (11:00 - 14:00). During the sample period, there are 19 days with negative errors (prediction greater than measurement) and the proportion is 21%, and there are 71 days with positive errors (measurement greater than prediction) and the proportion is 79%. The key period with concentrated recent errors is the midday peak period, as Figure 2 shown.

[0146] Select the data from July 10th to July 16th as the data to be corrected, and use the method of the present invention for longitudinal error correction and prediction back-calculation. The effect is as Figure 3 shown.

[0147] Table 2 Comparison results of the prediction accuracy of the original system and the prediction accuracy after error correction

[0148]

[0149] As shown in Table 2, for the positive and negative errors of the sample data, the prediction accuracy after the longitudinal error correction in the key period (noon peak: 11:00 - 14:00) of the present invention is increased by at least 3 percentage points compared with the prediction accuracy of the original system.

[0150] The present invention provides a method for longitudinal error correction of photovoltaic power. By selecting the period to be corrected for the predicted power to be corrected, it can more effectively cope with the changes faced by photovoltaic power at different times, improve the pertinence and accuracy of power prediction, and thus eliminate or reduce the possible systematic and random longitudinal errors in the prediction data. By determining the corrected power prediction value of the period to be corrected in the next sampling period according to the error type corresponding to the period to be corrected and using the predicted power error value of the period to be corrected in the next sampling period, it can more accurately understand the distribution characteristics of the errors, achieve efficient correction of the longitudinal error of power prediction, significantly improve the accuracy and credibility of power prediction data. At the same time, this customized correction method makes this technical route more applicable and can better meet the needs of specific application fields for high-quality power prediction data.

[0151] Embodiment 2

[0152] The present invention also provides a device for longitudinal error correction of photovoltaic power, as Figure 4 shown, including:

[0153] A determination unit, configured to determine the period to be corrected for the predicted power to be corrected according to historical output data;

[0154] A first prediction unit, configured to predict the error type of the period to be corrected by using a pre-established error probability model based on the predicted power value of the period to be corrected in the next sampling period;

[0155] A second prediction unit, configured to predict the predicted power error value of the period to be corrected in the next sampling period by using a pre-established error correction model based on the meteorological-related information of the period to be corrected in the next sampling period;

[0156] A correction unit, configured to determine the corrected power prediction value of the period to be corrected in the next sampling period according to the error type corresponding to the period to be corrected and using the predicted power error value of the period to be corrected in the next sampling period;

[0157] The error probability model is constructed by using historical output data, and the error correction model is constructed by using historical output data.

[0158] Furthermore, the historical output data includes:

[0159] The predicted power values of each period in the historical sampling period and the measured power values of each period in the historical sampling period.

[0160] Further, the device further includes: a first establishment unit configured to establish an error probability model;

[0161] The first establishment unit includes:

[0162] A first calculation module configured to calculate the measured power mean value and the measured power standard deviation of the to-be-corrected period in the historical sampling period by using the measured power values of the to-be-corrected period in the historical sampling period;

[0163] An establishment module configured to establish an error probability model by using the probability density function based on the measured power mean value and the standard deviation of the to-be-corrected period in the historical sampling period.

[0164] Further, the error probability model includes:

[0165]

[0166] In the above formula, f(x, μ, σ) is the probability density function; x is a random variable, that is, the predicted power value of the to-be-corrected period; μ is the measured power mean value of the to-be-corrected period in the historical sampling period, and σ is the measured power standard deviation of the to-be-corrected period in the historical sampling period.

[0167] Further, the device further includes: a second establishment unit configured to establish an error correction model;

[0168] The second establishment unit includes:

[0169] An acquisition module configured to acquire the meteorological-related information of each period in the historical sampling period;

[0170] A second calculation module configured to calculate the actual power error value of each period in the historical sampling period by using the predicted power value of each period in the historical sampling period and the measured power value of each period in the historical sampling period;

[0171] A construction module configured to construct a data set by using the meteorological-related information of each period in the historical sampling period and the actual power error value of each period in the historical sampling period;

[0172] A division module configured to divide the data set into a training set and a validation set;

[0173] A training module configured to train a support vector machine model by using the training set to obtain a trained support vector machine model;

[0174] A validation module configured to validate the trained support vector machine model by using the validation set. When the validation is successful, the trained support vector machine model is the error correction model.

[0175] Further, the training module is specifically configured to:

[0176] Using the meteorological-related information of each time period in the historical sampling periods in the training set as the training samples for the input layer of the support vector machine model, and using the actual power error values of each time period in the historical sampling periods in the training set as the training samples for the output layer of the support vector machine model to train the support vector machine model, a trained support vector machine model is obtained.

[0177] Further, a verification module, specifically used for:

[0178] Using the meteorological-related information of each time period in the historical sampling periods in the verification set as the input of the trained support vector machine model, and outputting the predicted power error values of each time period in the historical sampling periods;

[0179] Based on the predicted power error values of each time period in the historical sampling periods and the actual power error values of each time period in the historical sampling periods in the verification set, determining the accuracy rate of the trained support vector machine model;

[0180] If the accuracy rate of the trained support vector machine model is greater than or equal to the accuracy rate threshold, the verification is successful, and the trained support vector machine model is an error correction model; if the accuracy rate of the trained support vector machine model is less than the accuracy rate threshold, the verification fails, the parameters of the support vector machine model are adjusted, and the support vector machine model is retrained until the verification is successful.

[0181] Further, a determination unit, specifically used for:

[0182] Using the predicted power values of each time period in the historical sampling periods and the measured power values of each time period in the historical sampling periods, calculating the actual power error values of each time period in the historical sampling periods;

[0183] According to the actual power error values of each time period in the historical sampling periods, calculating the average error probability of each time period in the historical sampling periods;

[0184] Making the time periods corresponding to the average error probability greater than the first threshold be the time periods to be corrected.

[0185] Further, a first prediction unit, specifically used for:

[0186] Using the predicted power value of the time period to be corrected in the next sampling period as the input of the error probability model, determining the probability density function value corresponding to the predicted power value of the time period to be corrected in the next sampling period;

[0187] If the probability density function value is greater than or equal to the second threshold, the error type of the time period to be corrected is a positive error; if the probability density function value is less than the second threshold, the error type of the time period to be corrected is a negative error.

[0188] Further, a positive error is: the predicted power value is greater than the measured power value;

[0189] The negative error is that the predicted power value is less than the measured power value.

[0190] Further, the second prediction unit is specifically configured to:

[0191] Use the meteorological information related to the next sampling period of the period to be corrected as the input of the error correction model, and output the predicted power error value of the period to be corrected in the next sampling period.

[0192] Further, the correction unit is specifically configured to:

[0193] When the error type corresponding to the period to be corrected is a positive error, subtract the predicted power error value of the period to be corrected in the next sampling period from the predicted power value of the period to be corrected in the next sampling period to obtain the corrected predicted power value of the period to be corrected in the next sampling period;

[0194] When the error type corresponding to the period to be corrected is a negative error, add the predicted power error value of the period to be corrected in the next sampling period to the predicted power value of the period to be corrected in the next sampling period to obtain the corrected predicted power value of the period to be corrected in the next sampling period.

[0195] Further, before the second prediction unit, there is also included:

[0196] A collection unit for collecting meteorological information related to the next sampling period of the period to be corrected.

[0197] Further, the meteorological information related includes: season data and meteorological data.

[0198] It can be understood that the device embodiments provided above correspond to the above method embodiments, and the corresponding specific contents can be referred to each other, and will not be elaborated here.

[0199] It can be understood that the same or similar parts in the above embodiments can be referred to each other, and the content not detailed in some embodiments can be seen in the same or similar content of other embodiments.

[0200] Embodiment III

[0201] Based on the same inventive concept, the present invention further provides a computer device, which includes a processor and a memory. The memory is used to store a computer program, and the computer program includes program instructions. The processor is used to execute the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, and is suitable for implementing one or more instructions. Specifically, it is suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function, so as to implement the steps of a photovoltaic power longitudinal error correction method and device in the above embodiments.

[0202] Embodiment 4

[0203] Based on the same inventive concept, the present invention further provides a storage medium, specifically a computer-readable storage medium (Memory). The computer-readable storage medium is a memory device in a computer device and is used to store programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and, of course, the extended storage medium supported by the computer device. The computer-readable storage medium provides a storage space, and this storage space stores the operating system of the terminal. And, one or more instructions suitable for being loaded and executed by the processor are also stored in this storage space. These instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The one or more instructions stored in the computer-readable storage medium can be loaded and executed by the processor to implement the steps of a photovoltaic power longitudinal error correction method and device in the above embodiments.

[0204] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0205] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for realizing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0206] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means realizes the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0207] These computer program instructions can also be loaded onto a computer or other programmable data processing devices, so that a series of operation steps are executed on the computer or other programmable devices to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable devices provide steps for realizing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0208] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: still can modify the specific implementation manners of the present invention or make equivalent replacements, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for correcting the longitudinal error of photovoltaic power, characterized in that, Including: Determine the period to be corrected for the predicted power that needs to be corrected according to the historical output data; Based on the power prediction value of the period to be corrected in the next sampling period, use the pre-established error probability model to predict the error type of the period to be corrected; Based on the meteorological-related information of the period to be corrected in the next sampling period, use the pre-established error correction model to predict the power error prediction value of the period to be corrected in the next sampling period; According to the error type corresponding to the period to be corrected, use the power error prediction value of the period to be corrected in the next sampling period to determine the corrected power prediction value of the period to be corrected in the next sampling period; The error probability model is constructed using the historical output data, and the error correction model is constructed using the historical output data.

2. The photovoltaic power longitudinal error correction method according to claim 1, characterized in that The historical output data includes: The predicted power values of each period in the historical sampling period and the measured power values of each period in the historical sampling period.

3. The photovoltaic power longitudinal error correction method according to claim 1, wherein The establishment process of the error probability model includes: Use the measured power value of the period to be corrected in the historical sampling period to calculate the mean measured power value and the standard deviation of the measured power of the period to be corrected in the historical sampling period; Based on the mean measured power value and the standard deviation of the period to be corrected in the historical sampling period, use the probability density function to establish the error probability model.

4. The photovoltaic power longitudinal error correction method according to claim 3, characterized in that The error probability model includes: In the above formula, f(x,μ,σ) is the probability density function; x is the random variable, that is, the predicted power value of the period to be corrected; μ is the mean measured power value of the period to be corrected in the historical sampling period, and σ is the standard deviation of the measured power of the period to be corrected in the historical sampling period.

5. The photovoltaic power longitudinal error correction method according to claim 2, characterized in that The establishment process of the error correction model includes: Collect the meteorological-related information of each period in the historical sampling period; Use the predicted power values of each period in the historical sampling period and the measured power values of each period in the historical sampling period to calculate the actual power error values of each period in the historical sampling period; Use the meteorological-related information of each period in the historical sampling period and the actual power error values of each period in the historical sampling period to construct a data set; Divide the data set into a training set and a validation set; Use the training set to train the support vector machine model to obtain the trained support vector machine model; Use the validation set to validate the trained support vector machine model. When the validation is successful, the trained support vector machine model is the error correction model.

6. The photovoltaic power longitudinal error correction method according to claim 5, wherein The step of using the training set to train the support vector machine model to obtain the trained support vector machine model includes: Use the meteorological-related information of each period in the historical sampling period in the training set as the input layer training samples of the support vector machine model, and use the actual power error values of each period in the historical sampling period in the training set as the output layer training samples of the support vector machine model to train the support vector machine model to obtain the trained support vector machine model.

7. The photovoltaic power longitudinal error correction method according to claim 5, characterized in that The step of using the validation set to validate the trained support vector machine model includes: Using the meteorological-related information for each time period in the historical sampling periods in the validation set as the input to the trained support vector machine model, output the predicted power error values for each time period in the historical sampling periods; Based on the predicted power error values for each time period in the historical sampling periods and the actual power error values for each time period in the historical sampling periods in the validation set, determine the accuracy rate of the trained support vector machine model; If the accuracy rate of the trained support vector machine model is greater than or equal to the accuracy rate threshold, the verification is successful, and the trained support vector machine model is the error correction model; if the accuracy rate of the trained support vector machine model is less than the accuracy rate threshold, the verification fails, adjust the parameters of the support vector machine model, and retrain the support vector machine model until the verification is successful.

8. The photovoltaic power longitudinal error correction method according to claim 2, wherein The determining the time periods to be corrected for the predicted power according to the historical output data includes: Using the predicted power values for each time period in the historical sampling periods and the measured power values for each time period in the historical sampling periods, calculate the actual power error values for each time period in the historical sampling periods; According to the actual power error values for each time period in the historical sampling periods, calculate the average error probability for each time period in the historical sampling periods; Let the time periods corresponding to the average error probability greater than the first threshold be the time periods to be corrected.

9. The photovoltaic power longitudinal error correction method according to claim 1, wherein The predicting the error type of the time period to be corrected based on the power prediction value of the time period to be corrected in the next sampling period using the pre-established error probability model includes: Using the power prediction value of the time period to be corrected in the next sampling period as the input to the error probability model, determine the probability density function value corresponding to the power prediction value of the time period to be corrected in the next sampling period; If the probability density function value is greater than or equal to the second threshold, the error type of the time period to be corrected is a positive error; if the probability density function value is less than the second threshold, the error type of the time period to be corrected is a negative error.

10. The photovoltaic power longitudinal error correction method according to claim 9, characterized in that The positive error is: the power prediction value is greater than the measured power value; The negative error is: the power prediction value is less than the measured power value.

11. The photovoltaic power longitudinal error correction method according to claim 1, wherein The predicting the power error prediction value of the time period to be corrected in the next sampling period based on the meteorological-related information of the time period to be corrected in the next sampling period using the pre-established error correction model includes: Using the meteorological-related information of the time period to be corrected in the next sampling period as the input to the error correction model, output the power error prediction value of the time period to be corrected in the next sampling period.

12. The photovoltaic power longitudinal error correction method according to claim 1, wherein The determining the corrected power prediction value of the time period to be corrected in the next sampling period according to the error type corresponding to the time period to be corrected using the power error prediction value of the time period to be corrected in the next sampling period includes: When the error type corresponding to the time period to be corrected is a positive error, subtract the power error prediction value of the time period to be corrected in the next sampling period from the power prediction value of the time period to be corrected in the next sampling period to obtain the corrected power prediction value of the time period to be corrected in the next sampling period; When the error type corresponding to the to-be-corrected period is a negative error, the power prediction value of the to-be-corrected period in the next sampling period plus the power error prediction value of the to-be-corrected period in the next sampling period gives the corrected power prediction value of the to-be-corrected period in the next sampling period.

13. The photovoltaic power longitudinal error correction method according to claim 1, characterized in that Before obtaining the power error prediction value of the to-be-corrected period in the next sampling period by using the pre-established error correction model based on the meteorological-related information of the to-be-corrected period in the next sampling period, it further includes: Collecting the meteorological-related information of the to-be-corrected period in the next sampling period.

14. The photovoltaic power longitudinal error correction method according to claim 1, wherein The meteorological-related information includes: season data and meteorological data.

15. A longitudinal error correction device for photovoltaic power, characterized in that, It includes: A determination unit for determining the to-be-corrected period whose predicted power needs to be corrected according to the historical output data; A first prediction unit for predicting the error type of the to-be-corrected period by using the pre-established error probability model based on the power prediction value of the to-be-corrected period in the next sampling period; A second prediction unit for predicting the power error prediction value of the to-be-corrected period in the next sampling period by using the pre-established error correction model based on the meteorological-related information of the to-be-corrected period in the next sampling period; A correction unit for determining the corrected power prediction value of the to-be-corrected period in the next sampling period according to the error type corresponding to the to-be-corrected period and using the power error prediction value of the to-be-corrected period in the next sampling period; The error probability model is constructed by using the historical output data, and the error correction model is constructed by using the historical output data.

16. A computer device, characterized in that, It includes: One or more processors; The processor is used to store one or more programs; When the one or more programs are executed by the one or more processors, the photovoltaic power longitudinal error correction method described in any one of claims 1 to 14 is implemented.

17. A computer-readable storage medium, characterized in that, There is a computer program stored thereon, and when the computer program is executed, the photovoltaic power longitudinal error correction method described in any one of claims 1 to 14 is implemented.

Images