Generated power prediction method and device, electronic equipment and storage medium

Abstract

The invention discloses a generation power prediction method and device, electronic equipment and a storage medium, and belongs to the field of artificial intelligence. The method comprises the steps of obtaining background meteorological conditions of weather data prediction; inputting the background meteorological conditions into a trained conditional probability model to obtain conditional probability distribution output by the conditional probability model; disturbing the weather forecast data based on the conditional probability distribution to obtain disturbed weather forecast data; and inputting the disturbed weather forecast data into a preset power prediction model, and determining a power generation power prediction value according to an output result of the power prediction model. According to the embodiment of the invention, the conditional probability distribution of the weather forecast data space-time displacement error is predicted according to the background meteorological condition, the weather forecast data is disturbed based on the prediction result, and the power generation power is predicted according to the disturbed weather forecast data, so that the influence of the weather forecast data space-time displacement error on power generation power prediction is reduced; and the accuracy of power generation power prediction is improved.

Description

Technical Field

[0001] This application belongs to the field of artificial intelligence technology, and in particular relates to a method, device, electronic device and storage medium for predicting power generation. Background Technology

[0002] Accurate forecasting of photovoltaic power generation is of vital importance for grid dispatching, enabling grid operators to formulate reasonable dispatching strategies in advance, optimize grid operation, and ensure the safe and stable operation of the power system.

[0003] Photovoltaic power generation is significantly affected by meteorological factors such as solar irradiance, cloud cover, and temperature, exhibiting marked fluctuations and intermittent characteristics. Therefore, predicting the power generation of photovoltaic systems relies on meteorological input data provided by numerical weather prediction. However, numerical weather prediction results contain inherent and systematic spatiotemporal uncertainties, and directly predicting photovoltaic power generation based on numerical weather prediction is prone to significant errors. Summary of the Invention

[0004] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes a power generation prediction method, apparatus, electronic device, and storage medium to improve the accuracy of power generation prediction.

[0005] In a first aspect, this application provides a method for predicting power generation, including: Obtaining background meteorological conditions from weather forecast data; The background meteorological conditions are input into a trained conditional probability model to obtain a conditional probability distribution output by the conditional probability model; the conditional probability distribution represents the probability that the weather forecast data will have a corresponding spatiotemporal displacement error under the background meteorological conditions. The weather forecast data is perturbed based on the conditional probability distribution to obtain perturbed weather forecast data; The disturbed weather forecast data is input into a preset power prediction model, and the power generation prediction result is determined based on the output of the power prediction model.

[0006] The power generation prediction method provided in this application, by inputting the background meteorological conditions predicted by weather forecast data into a trained conditional probability model, can predict the possible situations and corresponding probabilities of spatiotemporal displacement errors in the weather forecast data; based on the prediction results, the weather forecast data is perturbed, and the perturbed data is input into a preset power prediction model, which can predict power generation according to the possible spatiotemporal displacement errors in the weather forecast data, thereby reducing the impact of spatiotemporal displacement errors in weather forecast data on power generation prediction and improving the accuracy of power generation prediction.

[0007] According to one embodiment of this application, the conditional probability model outputs the conditional probability distribution in the following manner: Calculate the similarity between the background meteorological conditions and historical background meteorological conditions in the historical database; the historical database stores different historical background meteorological conditions and the spatiotemporal displacement errors corresponding to the different historical background meteorological conditions; The k historical background meteorological conditions whose similarity satisfies the preset conditions and the spatiotemporal displacement errors corresponding to the k historical background meteorological conditions are determined as the conditional probability distribution.

[0008] In this embodiment, by calculating the similarity between the current background meteorological conditions and the historical background meteorological conditions in the historical database, the historical background meteorological conditions that meet the preset similarity conditions and the corresponding spatiotemporal displacement errors are selected as the conditional probability distribution. This enables the conditional probability distribution to reflect the actual historical situation under similar background meteorological conditions, thereby improving the accuracy of predicting spatiotemporal displacement errors.

[0009] According to one embodiment of this application, the conditional probability model is trained in the following manner: Obtain historical weather forecast data and the corresponding measured weather data; Calculate the spatiotemporal displacement error between the historical weather forecast data and the actual weather data; The conditional probability model is trained using the spatiotemporal displacement error between different historical weather forecast data and corresponding actual weather data, and the background meteorological conditions predicted by different historical weather forecast data as training datasets.

[0010] In this embodiment, the conditional probability model is trained by using the spatiotemporal displacement error between different historical weather forecast data and corresponding measured weather data, and the background meteorological conditions predicted by different historical weather forecast data as training datasets. This enables the conditional probability model to predict the possible spatiotemporal displacement error of weather forecast data and its corresponding probability based on the current background meteorological conditions.

[0011] According to one embodiment of this application, the spatiotemporal displacement error includes spatial error; The calculation of the spatiotemporal displacement error between the historical weather forecast data and the actual weather data includes: Extract multiple first cloud cluster objects from different times predicted by the historical weather forecast data and multiple second cloud cluster objects from different times measured by the actual weather data; Match the second cloud objects corresponding to each first cloud object at each time point to obtain multiple cloud matching pairs; The spatial error is obtained by calculating the centroid displacement vector between the first cloud object and the second cloud object in each cloud matching pair.

[0012] In this embodiment, by extracting and matching multiple cloud objects from the historical weather forecast data and the actual weather data, and calculating the centroid displacement vectors of the first and second cloud objects in each matching pair, the spatial error between the forecast and actual positions of the same cloud can be obtained as the spatial error between the historical weather forecast data and the actual weather data, thereby improving the accuracy of spatial error calculation.

[0013] According to one embodiment of this application, the spatiotemporal displacement error includes a time error; The calculation of the spatiotemporal displacement error between the historical weather forecast data and the actual weather data includes: Extract the first weather time series predicted from the historical weather forecast data and the second weather time series measured from the actual weather data; The time lag required to calculate the maximum cross-correlation between the first weather time series and the second weather time series is used to obtain the time error.

[0014] In this embodiment, by extracting weather time series from weather data, the weather data can be analyzed in chronological order, thereby matching the historical weather forecast data and the actual weather data according to time, calculating the time error between the historical weather forecast data and the actual weather data, and thus improving the accuracy of event error calculation.

[0015] According to one embodiment of this application, based on the formula:

[0016] The time lag required to achieve the maximum cross-correlation between the first weather time series and the second weather time series is calculated to obtain the time error; in, This indicates the time error. Represents the correlation function. This represents the first weather time series. Indicates the amount of time lag The second weather time series after offset.

[0017] In this embodiment, by calculating the time lag required for the maximum cross-correlation between the first weather time series and the second weather time series, the calculated time error can be made closer to the actual time error between the historical weather forecast data and the actual weather measurement data, thereby improving the accuracy of the time error calculation.

[0018] According to one embodiment of this application, calculating the spatiotemporal displacement error between the historical weather forecast data and the measured weather data includes: Calculate the spatiotemporal displacement vector field between the historical weather forecast data and the actual weather data; the spatiotemporal displacement vector field represents the spatial displacement and time offset at different locations between the historical weather forecast data and the actual weather data; The average spatiotemporal displacement vector of the target region is extracted from the spatiotemporal displacement vector field and determined as the spatiotemporal displacement error.

[0019] In this embodiment, by calculating the spatiotemporal displacement vector field between the historical weather forecast data and the measured weather data, the spatiotemporal displacement of the historical weather forecast data as a whole relative to the measured weather data as a whole at different locations can be obtained. By extracting the average spatiotemporal displacement vector of the target area from the spatiotemporal displacement vector field and determining it as the spatiotemporal displacement error, the spatiotemporal displacement error can reflect the average level of spatiotemporal displacement error within the target area. Using the spatiotemporal displacement error data and the background meteorological conditions predicted by the historical weather forecast data as a training dataset can improve the accuracy of the conditional probability model in predicting the spatiotemporal displacement error of the weather forecast data.

[0020] According to one embodiment of this application, the step of perturbing the weather forecast data based on the conditional probability distribution to obtain perturbed weather forecast data includes: Sampling is performed from the conditional probability distribution to obtain N sets of spatiotemporal displacement error vectors representing different spatiotemporal displacement errors; The weather forecast data is perturbed using N sets of spatiotemporal displacement error vectors to obtain perturbed weather forecast data.

[0021] In this embodiment, by sampling from the conditional probability distribution, the spatiotemporal displacement error of the current weather forecast information based on the current weather forecast data can be obtained by the conditional probability model. By perturbing the weather forecast data with the spatiotemporal displacement error vector, the corresponding accurate weather forecast data can be inferred based on the prediction of the spatiotemporal displacement error for power generation prediction, thereby reducing the impact of the spatiotemporal displacement error of the weather forecast data on power generation prediction and improving the accuracy of power generation prediction.

[0022] According to one embodiment of this application, the step of perturbing the weather forecast data using N sets of the spatiotemporal displacement error vectors to obtain perturbed weather forecast data includes: According to the formula:

[0023] The weather forecast data is perturbed; in, This represents the weather forecast data after the i-th group of disturbances. This represents the weather forecast data prior to the disturbance. t represents spatial location, and t represents time. Indicates background meteorological conditions The probability of the occurrence of the i-th group of spatiotemporal displacement errors. This represents the spatiotemporal displacement error vector of the i-th group. Indicates spatial error, Indicates time error.

[0024] In this embodiment, by subtracting a spatiotemporal displacement error vector from the weather forecast data, accurate weather forecast data can be inferred based on the prediction of the spatiotemporal displacement error.

[0025] According to one embodiment of this application, the disturbed weather forecast data is input into a preset power prediction model, and the power generation prediction result is determined based on the output of the power prediction model, including: When the weather forecast data after the disturbance is a set, the output result includes a power prediction curve representing the power generation at different times, and the output result is determined as the power generation prediction result; When there are N sets of weather forecast data after the disturbance, the output results include N power prediction curves representing power generation at different times. Statistical analysis is performed on the N power prediction curves representing power generation at different times to obtain the power generation prediction results.

[0026] In this embodiment, by statistically analyzing the N power prediction curves representing power generation at different times output by the power prediction model based on each set of weather forecast data when there are N sets of weather forecast data after the disturbance, the power generation for the target future time period comprehensively considers various possible situations of spatiotemporal displacement errors in the weather forecast data, thereby reducing the impact of spatiotemporal displacement errors in the weather forecast data on power generation prediction and improving the accuracy of power generation prediction.

[0027] Secondly, this application provides a power generation prediction device, comprising: The acquisition module is used to acquire the background meteorological conditions for weather forecast data prediction; The input module is used to input the background meteorological conditions into a preset conditional probability model to obtain the conditional probability distribution output by the conditional probability model; the conditional probability distribution represents the probability that the weather forecast data will have a corresponding spatiotemporal displacement error under the background meteorological conditions. The perturbation module is used to perturb the weather forecast data based on the conditional probability distribution to obtain perturbed weather forecast data; The determination module is used to input the disturbed weather forecast data into the preset power prediction model, and determine the power generation prediction result based on the output of the power prediction model.

[0028] According to the power generation prediction device of this application, by inputting the background meteorological conditions predicted by weather forecast data into a trained conditional probability model, it is possible to predict the possible spatiotemporal displacement errors of the weather forecast data and their corresponding probabilities; based on the prediction results, the weather forecast data is perturbed, and the perturbed data is input into a preset power prediction model, which can predict the power generation based on the possible spatiotemporal displacement errors of the weather forecast data, thereby reducing the impact of spatiotemporal displacement errors of the weather forecast data on power generation prediction and improving the accuracy of power generation prediction.

[0029] Thirdly, this application provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the power generation prediction method as described in the first aspect above.

[0030] Fourthly, this application provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the power generation prediction method as described in the first aspect above.

[0031] Fifthly, this application provides a chip including a processor and a communication interface, the communication interface being coupled to the processor, the processor being used to run programs or instructions to implement the power generation prediction method as described in the first aspect above.

[0032] In a sixth aspect, this application provides a computer program product, including a computer program that, when executed by a processor, implements the power generation prediction method as described in the first aspect above.

[0033] The above-described one or more technical solutions in the embodiments of this application have at least one of the following technical effects: According to the power generation prediction method of this application, by inputting the background meteorological conditions predicted by weather forecast data into a trained conditional probability model, it is possible to predict the possible spatiotemporal displacement errors of the weather forecast data and their corresponding probabilities. Based on the prediction results, the weather forecast data is perturbed, and the perturbed data is input into a preset power prediction model. Power generation can be predicted according to the possible spatiotemporal displacement errors of the weather forecast data, thereby reducing the impact of spatiotemporal displacement errors of weather forecast data on power generation prediction and improving the accuracy of power generation prediction.

[0034] In some embodiments, by calculating the similarity between the current background meteorological conditions and the historical background meteorological conditions in the historical database, the historical background meteorological conditions that meet the preset similarity conditions and the corresponding spatiotemporal displacement errors are selected as the conditional probability distribution. This enables the conditional probability distribution to reflect the actual historical situation under similar background meteorological conditions, thereby improving the accuracy of predicting spatiotemporal displacement errors.

[0035] In some embodiments, the conditional probability model is trained by using the spatiotemporal displacement error between different historical weather forecast data and corresponding measured weather data, and the background meteorological conditions predicted by different historical weather forecast data as training datasets, so that the obtained conditional probability model can predict the possible spatiotemporal displacement error of weather forecast data and the corresponding probability based on the current background meteorological conditions.

[0036] In some embodiments, by extracting and matching multiple cloud objects from the historical weather forecast data and the actual weather data, and calculating the centroid displacement vectors of the first and second cloud objects in each matching pair, the spatial error between the forecast and actual locations of the same cloud can be obtained as the spatial error between the historical weather forecast data and the actual weather data, thereby improving the accuracy of spatial error calculation.

[0037] In some embodiments, by extracting weather time series from weather data, weather data can be analyzed in chronological order, thereby matching the historical weather forecast data and the actual weather data according to time, and calculating the time error between the historical weather forecast data and the actual weather data.

[0038] In some embodiments, by calculating the time lag required for the maximum cross-correlation between the first weather time series and the second weather time series, the calculated time error can be made closer to the actual time error between the historical weather forecast data and the actual weather measurement data, thereby improving the accuracy of the time error calculation.

[0039] In some embodiments, by calculating the spatiotemporal displacement vector field between the historical weather forecast data and the measured weather data, the spatiotemporal displacement of the historical weather forecast data as a whole relative to the measured weather data as a whole at different locations can be obtained. By extracting the average spatiotemporal displacement vector of the target area from the spatiotemporal displacement vector field and determining it as the spatiotemporal displacement error, the spatiotemporal displacement error can reflect the average level of spatiotemporal displacement error within the target area. Using the spatiotemporal displacement error data and the background meteorological conditions predicted by the historical weather forecast data as a training dataset can improve the accuracy of the conditional probability model in predicting the spatiotemporal displacement error of the weather forecast data.

[0040] In some embodiments, by sampling from the conditional probability distribution, the spatiotemporal displacement error of the current weather forecast information based on the current weather forecast data can be obtained by the conditional probability model. By perturbing the weather forecast data with the spatiotemporal displacement error vector, the corresponding accurate weather forecast data can be inferred based on the prediction of the spatiotemporal displacement error for power generation prediction, thereby reducing the impact of the spatiotemporal displacement error of the weather forecast data on power generation prediction and improving the accuracy of power generation prediction.

[0041] In some embodiments, by subtracting a spatiotemporal displacement error vector from the weather forecast data, accurate weather forecast data can be inferred based on the prediction of the spatiotemporal displacement error.

[0042] In some embodiments, when there are N sets of weather forecast data after the disturbance, statistical analysis is performed on the N power prediction curves representing power generation at different times output by the power prediction model based on each set of weather forecast data. This allows the power generation for the target future time period to comprehensively consider various possible cases of spatiotemporal displacement errors in the weather forecast data, thereby reducing the impact of spatiotemporal displacement errors in the weather forecast data on power generation prediction and improving the accuracy of power generation prediction.

[0043] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a schematic diagram of the spatial displacement error predicted by NWP according to an embodiment of this application; Figure 2 This is a schematic diagram of the time displacement error of NWP forecast provided in the embodiments of this application; Figure 3 This is a schematic diagram illustrating the power generation prediction and error situation provided in the embodiments of this application; Figure 4 This is a schematic flowchart of the power generation prediction method provided in the embodiments of this application; Figure 5 This is a schematic diagram of the power generation prediction process provided in the embodiments of this application; Figure 6This is a schematic diagram of the power generation prediction device provided in the embodiments of this application; Figure 7 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0046] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0047] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0048] Accurate forecasting of renewable energy generation is crucial for power grid dispatching, enabling grid operators to develop reasonable dispatching strategies in advance, optimize grid operation, and ensure the safe and stable operation of the power system. Renewable energy generation, such as photovoltaic power, is significantly affected by meteorological factors such as solar irradiance, cloud cover, and temperature, exhibiting significant fluctuations and intermittent characteristics. Therefore, forecasting the power generation of renewable energy systems relies on meteorological input data provided by numerical weather prediction.

[0049] NWP (Numerical Weather Prediction) is a weather forecasting system that uses measured meteorological data as initial conditions, models meteorological processes based on atmospheric physics and mathematical equations to simulate atmospheric motion, and solves fluid dynamics and thermodynamic equations by computer to generate future weather predictions.

[0050] Limited by factors such as physical parameterization schemes, resolution, and initial conditions, numerical weather prediction (NWMR) forecasts inherently possess systematic spatiotemporal uncertainties. Taking NWMR cloud cluster location prediction as an example, the predicted cloud cluster location may exhibit displacement errors in both time and space. Spatial displacement error refers to deviations in the predicted cloud cluster movement path. For example, ... Figure 1As shown, the NWP forecast predicted a cloud cluster would cover station A, but the actual cloud cluster passed by station B, 20 kilometers south of station A. Time displacement error refers to the deviation in the NWP forecast of the cloud cluster's movement speed. For example, as... Figure 2 As shown, the NWP forecast predicted that a cloud cluster would arrive at station A at T0+1h, but the actual cloud cluster arrived at T0 earlier.

[0051] Relying directly on numerical weather forecasts to predict renewable energy power generation is prone to significant errors. For example... Figure 3 As shown, taking the prediction of photovoltaic power generation based on the cloud movement path based on NWP forecasts as an example, when the cloud covers the photovoltaic power station and no sunlight shines on the photovoltaic panels, the power generation of the photovoltaic power station is close to 0; when the sky above the photovoltaic power station is clear and cloudless and the sun shines directly on the photovoltaic panels, the photovoltaic power station operates at full load and the power generation is high. If there is a spatiotemporal displacement error in the cloud movement path predicted by NWP, such as NWP predicting that a cloud will arrive at station A at 2 pm, but the actual cloud arrives earlier at 1 pm, then the predicted power may be full load at 1 pm and close to 0 at 2 pm, while the actual power may be close to 0 at 1 pm and full load at 2 pm. The error between the predicted value and the measured value is large, that is, a "present or absent" deviation occurs.

[0052] To address at least one of the aforementioned technical problems, embodiments of this application provide a power generation prediction method, apparatus, electronic device, and storage medium. The power generation prediction method, apparatus, electronic device, and storage medium provided in these embodiments can be applied to any power generation scenario, such as photovoltaic power generation, wind power generation, and power load prediction; these embodiments do not limit the application to these scenarios.

[0053] The following description, in conjunction with the accompanying drawings, details the power generation prediction method, apparatus, electronic equipment, and storage medium provided in this application through specific embodiments and application scenarios.

[0054] The power generation prediction method can be applied to the terminal, specifically executed by the hardware or software in the terminal.

[0055] The terminal includes, but is not limited to, portable communication devices such as mobile phones or tablets with touch-sensitive surfaces (e.g., touchscreen displays and / or touchpads). It should also be understood that, in some embodiments, the terminal may not be a portable communication device, but rather a desktop computer with touch-sensitive surfaces (e.g., touchscreen displays and / or touchpads).

[0056] The following embodiments describe a terminal including a display and a touch-sensitive surface. However, it should be understood that the terminal may include one or more other physical user interface devices such as a physical keyboard, mouse, and joystick.

[0057] The power generation prediction method provided in this application can be executed by an electronic device or a functional module or entity in an electronic device that can implement the power generation prediction method. The electronic devices mentioned in this application include photovoltaic systems, wind power generation systems, servers, etc. The power generation prediction method provided in this application is described below using an electronic device as the execution subject.

[0058] like Figure 4 As shown, the power generation prediction method includes steps 410, 420, 430 and 440.

[0059] Step 410: Obtain the background meteorological conditions predicted by the weather forecast data.

[0060] Background meteorological conditions are a comprehensive state formed by the combination of meteorological elements such as air pressure, temperature, humidity, and wind speed in the atmosphere. Changes in these conditions directly affect the formation and evolution of various weather phenomena.

[0061] Weather forecast data is the prediction information of meteorological element values ​​released by numerical weather prediction systems. It can be the forecast data of meteorological element values ​​such as GHI (Global Horizontal Irradiance), DNI (Direct Normal Irradiance), DHI (Diffuse Horizontal Irradiance), cloud cover, 10-meter U / V (zonal / meridian) wind speed, and sea level pressure released by numerical weather prediction systems such as ECMWF (European Centre for Medium-Range Weather Forecasts) and GFS (Global Forecast System).

[0062] Background meteorological conditions can be formed by selecting one or more meteorological element values ​​from weather forecast data for a specific region and time period published by a numerical weather prediction system. For example, the weather forecast data published by the NWP system includes forecast values ​​of meteorological elements such as GHI, DNI, cloud cover, 10 m U / V wind speed, sea level pressure, and precipitation. The forecast values ​​of three meteorological elements—GHI, cloud cover, and 10 m U / V wind speed—can be selected to form the background meteorological conditions. Of course, other meteorological element values ​​can also be selected to form the background meteorological conditions; this application does not limit this.

[0063] Step 420: Input the background meteorological conditions into the trained conditional probability model to obtain the conditional probability distribution output by the conditional probability model; the conditional probability distribution represents the probability that the weather forecast data will have a corresponding spatiotemporal displacement error under the background meteorological conditions.

[0064] A conditional probability model is a machine learning model used to calculate the probability of one event occurring given that some other events have occurred. In this embodiment, the conditional probability model is pre-trained to output a conditional probability distribution based on input background meteorological conditions. The conditional probability distribution describes the possible scenarios of spatiotemporal displacement error and the probability corresponding to each scenario.

[0065] For example, the background weather conditions published by the NWP system show that the temperature at location (x, y) at time t is T degrees Celsius, denoted as... Background meteorological conditions Input conditional probability model , There is a 25% probability that the predicted spatiotemporal displacement error will result in the location being correct but the time being 2 hours too early; this is denoted as... There is a 50% probability that the time is correct, but the location is offset 2 kilometers to the north, denoted as... There is a 25% probability that the location will shift 1 kilometer to the south, 3 kilometers to the east, and the time will be 3 hours later. From this, we obtain That is, the spatiotemporal displacement error is in probability distribution under certain conditions .

[0066] Conditional probability models can be implemented using GMM (Gaussian Mixture Model) framework, cGAN (Conditional Generative Adversarial Network) framework, or AnEn (Analog Ensemble) framework. Alternatively, simpler XGBoost (Extreme Gradient Boosting) structures or more complex Transformer encoder structures can be used. This application does not limit the specific implementation of such conditional probability models.

[0067] In some embodiments, the conditional probability model is trained as follows: Obtain historical weather forecast data and corresponding measured weather data; Calculate the spatiotemporal displacement error between historical weather forecast data and actual weather measurement data; The spatiotemporal displacement error between different historical weather forecast data and corresponding actual weather data, and the background meteorological conditions predicted by different historical weather forecast data are used as training datasets to train the conditional probability model.

[0068] Observed weather data are records of meteorological elements directly measured by meteorological stations, satellites, radar, and other equipment. These data may include actual observed values ​​of temperature, humidity, GHI, DNI, DHI, cloud cover, 10 m U / V wind speed, and sea level pressure. Observed weather data can be obtained from geostationary satellite systems such as the Fengyun Meteorological Satellite Series and the Himawari Geostationary Meteorological Satellite (GMS). Spatial and temporal resampling processes can be used to process the raw observed weather data to obtain observed weather data that corresponds to the NWP forecast data in both time and space.

[0069] After obtaining historical weather forecast data and corresponding measured weather data, the spatiotemporal displacement error between the historical weather forecast data and the measured weather data is calculated. For example, for GHI forecast data and measured data within a certain time period and a certain area, data records whose GHI values ​​fall within a user-defined interval can be searched in order from south to north, from west to east, and from old to new, and numbered sequentially. For example, the search result in the forecast data is... The search results in the actual test data are For each data point with the same number, the predicted data can be subtracted from the actual measured data, and the result can be used as the spatiotemporal displacement error value of that data point, such as... .

[0070] For each data point It can be obtained from meteorological databases. Location, The forecast values ​​for cloud cover and 10-meter U / V wind speed at specific times are used as... Background meteorological conditions Then a sample data is obtained. Collect all historical data to form a dataset. Used for training conditional probability models .

[0071] The training task is to input background meteorological conditions. The conditional probability distribution of the spatiotemporal displacement error is output. ,in ,Right now Under the condition that the spatiotemporal displacement error is The probability, .

[0072] In each round of training, the input... Calculate based on the loss function The output result and the actual error value The loss value; where the mean squared error, absolute average absolute error, etc., can be used as the loss function. For example, for conditional probability distributions... You can choose one of the probabilities The largest As prediction error value ,calculate and The mean squared error value is used as the loss value. Based on the loss value, the value is updated... The parameters in the code are used to start a new round of training until the prediction error value is reached. Error value The training will tend to be consistent or reach the number of training rounds preset by the user.

[0073] in, It can be a machine learning model based on the GMM framework, or it can be a machine learning model based on frameworks such as cGAN, AnEn, the simpler XGBoost, or the more complex Transformer. This application does not limit the specific implementation of the model.

[0074] In this embodiment, the conditional probability model is trained by using the spatiotemporal displacement error between different historical weather forecast data and corresponding measured weather data, and the background meteorological conditions predicted by different historical weather forecast data as training datasets. This enables the conditional probability model to predict the possible spatiotemporal displacement error of weather forecast data and its corresponding probability based on the current background meteorological conditions.

[0075] In some embodiments, spatiotemporal displacement error includes spatial error; Calculate the spatiotemporal displacement error between historical weather forecast data and actual weather data, including: Extract multiple first cloud cluster objects from historical weather forecast data at different times and multiple second cloud cluster objects from actual weather measurement data at different times; Match the second cloud objects corresponding to each first cloud object at each time point to obtain multiple cloud matching pairs; Calculate the centroid displacement vector between the first and second cloud objects in each cloud pair to obtain the spatial error.

[0076] Weather forecasts / observed data released by the NWP system / satellite system can be stored as meteorological element fields in the form of multidimensional digital matrices and visualized as multiple color images. For example, the NWP system releases a forecast GHI field for a certain area on day D-1, which is stored in the computer as a three-dimensional array. The value at (x, y, t) represents the predicted GHI value at (x, y) time t. Similarly, the measured GHI field over day D within the same region recorded by the satellite system can be stored as a three-dimensional array. .

[0077] First, extract multiple first cloud cluster objects from different times predicted by historical weather forecast data and multiple second cloud cluster objects from different times measured by actual weather data.

[0078] For each time t, an image segmentation algorithm can be applied to segment the meteorological element field data at time t into multiple cloud cluster objects. Image segmentation is a computer vision technique used to divide an image into multiple meaningful regions. Since a planar image is essentially a two-dimensional array, image segmentation algorithms can also be used to segment any dataset stored in the form of a two-dimensional array. Threshold-based algorithms such as Otsu's thresholding method, clustering-based algorithms such as K-Means, or deep learning-based image segmentation algorithms such as U-Net (U-Net Convolutional Neural Network) can be used to segment the meteorological element field data at time t; this application does not limit the specific methods used.

[0079] For example, for the GHI field at time t, a user can pre-define a GHI interval, identifying areas with GHI values ​​within this interval as regions with cloud cover that could affect photovoltaic power generation, thus obtaining a series of data points. The K-Means clustering algorithm can then be used to cluster these data points according to their location, resulting in multiple categories. Data points within each category are close together, while data points between categories are relatively far apart. Each category of data points forms a cloud cluster object.

[0080] After obtaining the first cloud cluster object and the second cloud cluster object, match the second cloud cluster object corresponding to each first cloud cluster object at each time point to obtain multiple cloud cluster matching pairs.

[0081] For each data point within each category, we can analyze the common characteristics of these data points, which can then be used as the feature vector of the cloud object. For example, for the cloud object... It can be calculated The average value of the positions of the data points is used as the centroid position of cloud object i. ;calculate The average GHI intensity of the data points in the middle is used as the average GHI intensity of cloud object i. Thus, the feature vector of cloud object i is obtained. .

[0082] The forecast meteorological element field data at each time point is segmented to obtain the cloud cluster object, which is the first cloud cluster object; the measured meteorological element field data at each time point is segmented to obtain the cloud cluster object, which is the second cloud cluster object.

[0083] Match the second cloud object corresponding to each first cloud object at each time step. For example, for the GHI field data at time t, segment the NWP forecast field to obtain N first cloud objects. M second cloud objects were obtained by segmenting the NWP test field. The nearest neighbor matching method can be used for each first cloud object. ,calculate With each second cloud object Choose the Euclidean distance from the centroid to the nearest centroid. Forming matching pairs. Of course, other target tracking algorithms such as overlap matching, Hungarian matching, and continuous optical flow can also be used for matching, but this application does not limit this.

[0084] After obtaining the cloud cluster matching pairs, the centroid displacement vector between the first and second cloud cluster objects in each pair is calculated to obtain the spatial error. For example, for cloud cluster matching pairs... ,calculate As spatial error.

[0085] In this embodiment, by extracting and matching multiple cloud objects from historical weather forecast data and actual weather data, and calculating the centroid displacement vectors of the first and second cloud objects in each matching pair, the spatial error between the forecast and actual positions of the same cloud can be obtained as the spatial error between historical weather forecast data and actual weather data, thereby improving the accuracy of spatial error calculation.

[0086] In some embodiments, spatiotemporal displacement error includes time error; Calculate the spatiotemporal displacement error between historical weather forecast data and actual weather data, including: Extract the first weather time series predicted from historical weather forecast data and the second weather time series from actual weather measurement data; The time lag required to calculate the maximum cross-correlation between the first and second weather time series is used to obtain the time error.

[0087] A weather time series is a collection of weather data points arranged in chronological order, where each weather data point records the meteorological element values ​​at a specific location at a specific time.

[0088] You can search by location in the weather forecast / measured data released by the NWP system / satellite system to obtain a series of forecast / measured meteorological element data at different times for the search location. Arrange them in chronological order and use the resulting sequence as the first / second weather time series.

[0089] For example, the NWP system publishes GHI forecast data for a certain area over a certain period of time. Search location From the GHI forecast data at that location, a series of data points were obtained. ,in express At what time, the NWP system predicts The GHI value at that location Arrange these data points in chronological order to obtain a time series. ,in That is Location at GHI forecast value at time 10:00 This is the first weather time series. Similarly, in the GHI measured data released by the satellite system... The search location is the search location. The second weather time series was obtained from the measured GHI data at the location. .

[0090] Obtain the first weather time series Second weather time series Then, calculate and The time lag required for the maximum cross-correlation is used to obtain the time error between historical weather forecast data and actual weather data.

[0091] For example, a user can set a GHI interval, which is respectively in and Search for data points whose GHI values ​​fall within a preset range, record their time values, and number them sequentially. For example, in... The search results in ,exist The search results in Locations with the same number can be used as the locations of maximum cross-correlation between historical weather forecast data and observed weather data, for example... The position of maximum cross-correlation is The time value of the measured data can be subtracted from the time value of the forecast data, and the result can be used as the time lag for that data point. Calculate the average time lag of all data points with the largest cross-correlation, and use this as the time error between historical weather forecast data and observed weather data. .

[0092] In this embodiment, by extracting weather time series from weather data, the weather data can be analyzed in chronological order, thereby matching historical weather forecast data and actual weather data according to time and calculating the time error between historical weather forecast data and actual weather data.

[0093] In some embodiments, according to the formula:

[0094] The time lag required to calculate the maximum cross-correlation between the first and second weather time series is used to obtain the time error. in, Indicates time error. Represents the correlation function. This represents the first weather time series. Indicates the amount of time lag The offset second weather time series.

[0095] For example, users can adjust time lag. Set a search range For example, setting it to (-120, 120, 15) means letting In each search, the time increments are changed from 120 minutes behind to 120 minutes ahead, in 15-minute increments.

[0096] Then, for each ,Will Move on the timeline ,get For example, for , yes Measured GHI value at time 10:00 That is The measured GHI value 15 minutes later.

[0097] Calculate using correlation function and The correlation function is a mathematical tool used to measure the degree of linear correlation between two variables. Formulas such as the Pearson correlation coefficient formula and the cross-correlation function can be used as correlation functions. .

[0098] Results of correlation function calculation It is a value between -1 and 1, the closer to 1, the higher the value. and The stronger the positive correlation and the more similar the pairs, the better. Therefore, for each pair... After completing the calculation, select to make The largest value Soon Move on the timeline The obtained GHI measured value sequence It is most similar to the GHI forecast value series. This time lag value is used as the time error between historical weather forecast data and actual weather observation data.

[0099] In this embodiment, by calculating the time lag required for the maximum cross-correlation between the first weather time series and the second weather time series, the calculated time error can be made closer to the actual time error between historical weather forecast data and actual weather measurement data, thereby improving the accuracy of time error calculation.

[0100] Step 430: Perturb the weather forecast data based on the conditional probability distribution to obtain the perturbed weather forecast data.

[0101] In this embodiment, the conditional probability distribution describes the possible scenarios of spatiotemporal displacement errors and the probability corresponding to each scenario. Weather forecast data can be perturbed based on the spatiotemporal displacement errors in the conditional probability distribution. For example, the spatiotemporal displacement error with the highest probability in the conditional probability distribution can be selected, and then the weather forecast data can be perturbed using this spatiotemporal displacement error to obtain perturbed weather forecast data.

[0102] In some embodiments, weather forecast data is perturbed based on a conditional probability distribution to obtain perturbed weather forecast data, including: Sampling is performed from the conditional probability distribution to obtain N sets of spatiotemporal displacement error vectors representing different spatiotemporal displacement errors; The weather forecast data is perturbed by using N sets of spatiotemporal displacement error vectors to obtain the perturbed weather forecast data.

[0103] Sampling refers to generating random samples based on a known conditional probability distribution. Sampling from the conditional probability distribution output by the conditional probability model involves generating one or more samples of spatiotemporal displacement errors based on the possible spatiotemporal displacement errors of weather forecast data predicted by the conditional probability model according to background meteorological conditions and the probability corresponding to each condition. The statistical distribution of these spatiotemporal displacement error samples is consistent with the conditional probability distribution.

[0104] For example, conditional probability models Based on the input background meteorological conditions Output conditional probability distribution .

[0105] First, construct a cumulative probability distribution based on the conditional probability distribution, that is, map the probability value of each case in the conditional probability distribution to the corresponding subinterval on the interval [0, 1). For example, The interval is [0, 0.2). The interval is [0.2, 0.5). The interval is [0.5, 1).

[0106] In each sampling, a random number generator can be used to generate a random number r uniformly distributed in the range [0, 1). This r is then compared to the boundary of the cumulative probability interval to determine which case to select as the generated sample. For example, when... Then choose ;when Then choose ;when Then choose .

[0107] Repeating the single sampling process N times yields N spatiotemporal displacement error vectors, which serve as samples of the spatiotemporal displacement error. When N is sufficiently large, the distribution of these samples will infinitely approximate the conditional probability distribution. For example... That is, in the sample The frequency of occurrence will approach 20%. The frequency of occurrence will approach 30%. The frequency of occurrence will approach 50%.

[0108] Of course, in the actual sampling process, N can be a small value such as 1 or 5, or a large value such as 100 or 1000. This application embodiment does not limit this.

[0109] After obtaining N sets of spatiotemporal displacement error vectors, the weather forecast data is perturbed using the N sets of spatiotemporal displacement error vectors to obtain N sets of perturbed weather forecast data.

[0110] In this embodiment, by sampling from the conditional probability distribution, the spatiotemporal displacement error of the current weather forecast information based on the current weather forecast data can be obtained by the conditional probability model. By perturbing the weather forecast data with the spatiotemporal displacement error vector, the corresponding accurate weather forecast data can be inferred based on the prediction of the spatiotemporal displacement error to predict the power generation, thereby reducing the impact of the spatiotemporal displacement error of the weather forecast data on the power generation prediction and improving the accuracy of the power generation prediction.

[0111] In some embodiments, N sets of spatiotemporal displacement error vectors are used to perturb the weather forecast data to obtain perturbed weather forecast data, including: According to the formula:

[0112] Perturbation of weather forecast data; in, This represents the weather forecast data after the i-th group of disturbances. This represents the weather forecast data prior to the disturbance. t represents spatial location, and t represents time. Indicates background meteorological conditions The probability of the occurrence of the i-th group of spatiotemporal displacement errors. This represents the spatiotemporal displacement error vector of the i-th group. Indicates spatial error, Indicates time error.

[0113] For example, the GHI value at (20, 30, 10) in the original NWP data is 1000 W / m². The first set of spatiotemporal displacement error vectors is (-5, 10, 1), indicating that the forecast value is 5 km westward, 10 km northward, and 1 hour later than the actual value. Therefore, the forecast value needs to be moved 5 km eastward, 10 km southward, and 1 hour earlier, i.e., (20-(-5), 30-10, 10-1) to obtain the accurate forecast value (25, 20, 9).

[0114] In this embodiment, by subtracting the spatiotemporal displacement error vector from the weather forecast data, accurate weather forecast data can be inferred based on the prediction of the spatiotemporal displacement error.

[0115] Step 440: Input the disturbed weather forecast data into the preset power prediction model, and determine the power generation prediction result based on the output of the power prediction model.

[0116] A power prediction model is a machine learning model that outputs a predicted power generation capacity based on input weather forecast data. The power prediction model can be a physical model, or a machine learning model with an XGBoost framework, LSTM (Long Short-Term Memory) framework, or other structures; this application does not limit the specific implementation of such models.

[0117] A power prediction model can be trained using historical weather forecast data and corresponding measured power generation data published by the NWP system. For example, an LSTM-based machine learning model can be established as the initial photovoltaic power base model. , The training dataset is The training task is to train based on the input NWP forecast data. The predicted value of PV (Photovoltaic) power generation output. In each round of training, the input is... The predicted output power is calculated based on the loss function. Compared with actual power generation The loss value is updated based on the loss value. The parameters in the [training program] are updated. After the update, a new round of training begins until the predicted power generation value is [achieved / reset]. The value of PV tends to be consistent with the actual value or reaches the number of training rounds preset by the user.

[0118] In some embodiments, the disturbed weather forecast data is input into a preset power prediction model, and the power generation prediction result is determined based on the output of the power prediction model, including: When the weather forecast data after the disturbance is a set, the output includes a power prediction curve representing the power generation at different times, and the output is determined as the power generation prediction result. Given N sets of weather forecast data after the disturbance, the output includes N power prediction curves representing power generation at different times. Statistical analysis is performed on the N power prediction curves representing power generation at different times to obtain the power generation prediction results.

[0119] When the weather forecast data after the disturbance is a set, the output includes a power prediction curve representing the power generation at different times, and the output is determined as the power generation prediction result.

[0120] For example, the weather forecast data obtained from the disturbance is ,Will enter The output is a power prediction curve. Then This result was determined to be the predicted power generation output.

[0121] Given N sets of weather forecast data after the disturbance, the output includes N power prediction curves representing the power generation at different times.

[0122] For example, the disturbance yields N sets of weather forecast data. Enter each set of weather forecast data one by one. , For each set of data Output a power prediction curve A total of N power prediction curves were obtained. .

[0123] Statistical analysis was performed on N power prediction curves representing power generation at different times to obtain power generation prediction results.

[0124] The output can be a probabilistic prediction, meaning the output is a set of prediction intervals, for example... This indicates a 95% probability that the power generation capacity will be sufficient. There is a 5% probability that There is a 90% probability that it is in and between.

[0125] The quantiles of the power prediction values ​​for each power prediction curve at each time point can be calculated to form a prediction interval. For example, for each time point t, first calculate the quantiles of each power prediction curve. Power prediction at time t The obtained N power values ​​are sorted in descending order. Then, quantiles are calculated, which means finding the value at a specific position in the sorted sequence. For example, Q95 (the 95th quantile) is the value located at the [missing quantile]th position in the sorted sequence. The power values ​​at (rounded up) positions represent the power values ​​of 95% of the members in all possible predictions that are lower than this value; similarly, Q5 is calculated to form the prediction interval for time t. Finally, based on the prediction intervals of the power values ​​at each time t, the power prediction interval for the entire time period is synthesized.

[0126] The output can also be a deterministic prediction, meaning the output is a definite power value, for example... This indicates that the predicted power generation value is PV. The mean or median of the predicted power values ​​for each time point and each power prediction curve can be calculated as a definite predicted value. For example, for each time point t, the mean or median of the predicted power values ​​for each power prediction curve can be calculated. Power prediction at time t The obtained N power values ​​are sorted from largest to smallest, and the power value at the 50% position is selected as the power prediction value.

[0127] In this embodiment, by statistically analyzing the N power prediction curves representing power generation at different times output by the power prediction model based on each set of weather forecast data when there are N sets of weather forecast data after disturbance, the power generation for the target future time period comprehensively considers various possible situations of spatiotemporal displacement error of weather forecast data, thereby reducing the impact of spatiotemporal displacement error of weather forecast data on power generation prediction and improving the accuracy of power generation prediction.

[0128] The power generation prediction method provided in this application, by inputting the background meteorological conditions predicted by weather forecast data into a trained conditional probability model, can predict the possible situations and corresponding probabilities of spatiotemporal displacement errors in weather forecast data; based on the prediction results, the weather forecast data is perturbed, and the perturbed data is input into a preset power prediction model, which can predict power generation according to the possible spatiotemporal displacement errors in weather forecast data, thereby reducing the impact of spatiotemporal displacement errors in weather forecast data on power generation prediction and improving the accuracy of power generation prediction.

[0129] In some embodiments, the conditional probability model outputs a conditional probability distribution in the following manner: Calculate the similarity between the background meteorological conditions and the historical background meteorological conditions in the historical database; the historical database stores different historical background meteorological conditions and the spatiotemporal displacement errors corresponding to different historical background meteorological conditions; The k historical background meteorological conditions whose similarity satisfies the preset conditions and the spatiotemporal displacement errors corresponding to the k historical background meteorological conditions are determined as conditional probability distributions.

[0130] The historical database stores meteorological conditions under different historical backgrounds and the corresponding spatiotemporal displacement errors. Historical weather forecast data and historical background meteorological condition data can be collected from numerical weather prediction systems and meteorological databases, and spatiotemporal displacement errors can be calculated to form the historical database.

[0131] Input background meteorological conditions ,calculate Similarity to historical weather conditions in the historical database. This can be done for each data point in the historical database. Calculate using correlation function and similarity .

[0132] After all calculations are completed, select k similarity values ​​that meet the preset conditions. Count k The distribution of the data is analyzed, and the results are determined as a conditional probability distribution. Preset conditions are defined by the user, for example, selecting the N items with the highest similarity. To perform statistics. Statistics can be compiled. Several kinds appeared in Each The frequency of occurrence will The frequency of occurrence as a condition in the probability distribution The corresponding probability.

[0133] In this embodiment, by calculating the similarity between the current background meteorological conditions and the historical background meteorological conditions in the historical database, the historical background meteorological conditions that meet the preset similarity conditions and the corresponding spatiotemporal displacement errors are selected as the conditional probability distribution. This enables the conditional probability distribution to reflect the actual historical situation under similar background meteorological conditions, thereby improving the accuracy of predicting spatiotemporal displacement errors.

[0134] In some embodiments, calculating the spatiotemporal displacement error between historical weather forecast data and actual weather data includes: Calculate the spatiotemporal displacement vector field between historical weather forecast data and actual weather data; the spatiotemporal displacement vector field represents the spatial displacement and time offset at different locations between historical weather forecast data and actual weather data; The average spatiotemporal displacement vector of the target region is extracted from the spatiotemporal displacement vector field and determined as the spatiotemporal displacement error.

[0135] The spatiotemporal displacement vector field can be a three-dimensional array E(x, y, t), along with historical weather forecast data. Actual weather data Each point in the vector corresponds one-to-one, and the value stored at E(x, y, t) is a spatiotemporal displacement vector. , which represents the temporal and spatial error of the weather forecast data at time t and position (x, y) relative to the actual weather data.

[0136] Optical flow can be used to calculate the spatiotemporal displacement vector field between historical weather forecast data and actual weather data. Optical flow is a target tracking algorithm used to estimate the motion vector of each object point in a continuous series of images. It assumes that the brightness or color intensity of the same object point remains constant over very short time intervals. Therefore, matching pixels with similar brightness in adjacent images should indicate the same object point, and calculating the position difference between these two pixels yields the motion vector of that object point. Weather data over a period of time is essentially the same as a series of continuous images; both can be viewed as a three-dimensional array. The only difference is that the array of continuous image data stores color and brightness values, while the array of weather data stores meteorological element values ​​such as GHI and cloud cover. Therefore, optical flow is also suitable for calculating the spatiotemporal displacement vector field between historical weather forecast data and actual weather data.

[0137] Variational optical flow methods or deep learning-based optical flow models such as RAFT (Recurrent All-Pairs Field Transforms) can be used to calculate the spatiotemporal displacement vector field between the NWP forecast field and the actual satellite field.

[0138] After calculating the spatiotemporal displacement vector field E(x, y, t), the average displacement vector of the target region is extracted as... For example, for the target area It can calculate the average value of all spatiotemporal displacement vectors within the target area, which is then used as the average displacement vector. Where N is the target region The number of data points within the dataset. The calculated... As the target area Spatiotemporal displacement error between historical weather forecast data and actual weather measurement data .

[0139] It is important to note that the spatiotemporal displacement error in this embodiment is obtained by calculating the spatiotemporal displacement vector field of the entire target region using a target tracking algorithm, and then calculating the average displacement vector. In contrast, the spatiotemporal displacement error in the aforementioned embodiment was obtained by dividing the target region into multiple cloud objects and calculating the error for each cloud object using a target tracking algorithm. By applying a target tracking algorithm to the entire target region, the resulting spatiotemporal displacement vector field can comprehensively reflect the spatiotemporal displacement error at various locations within the target region.

[0140] In this embodiment, by calculating the spatiotemporal displacement vector field between historical weather forecast data and actual weather data, the spatiotemporal displacement of the overall historical weather forecast data relative to the overall actual weather data at different locations can be obtained. By extracting the average spatiotemporal displacement vector of the target area from the spatiotemporal displacement vector field and determining it as the spatiotemporal displacement error, the spatiotemporal displacement error can reflect the average level of spatiotemporal displacement error within the target area. Using the spatiotemporal displacement error data and the background meteorological conditions predicted by different historical weather forecast data as a training dataset can improve the accuracy of the conditional probability model in predicting the spatiotemporal displacement error of weather forecast data.

[0141] The following scenario example illustrates the power generation prediction method provided in this application. For example... Figure 5 As shown in the example, the first step is to prepare the data. Historical weather forecast data, measured weather data, and the latest weather forecast data for the power plant area over a past period are obtained from numerical weather prediction systems and meteorological databases. Historical power generation data is also obtained from the power plant's historical operating data. Background meteorological conditions data are extracted from the weather forecast data. The spatiotemporal displacement error of the historical weather forecast data is calculated. For each moment in the historical weather forecast data, cloud clusters are segmented, and the centroid displacement vector of the cloud clusters at different times is calculated to obtain the spatial error. The weather time series of the historical weather forecast data and measured weather data at each moment are extracted. The time lag required for the forecast and actual weather time series to achieve maximum cross-correlation is calculated to obtain the time error.

[0142] After acquiring the data, the model is trained offline. A conditional probability model is trained using the spatiotemporal displacement error between different historical weather forecast data and corresponding actual weather data, along with the background meteorological conditions predicted by different historical weather forecast data as the training dataset. A power prediction model is then trained using different historical weather forecast data and corresponding actual power generation data.

[0143] After obtaining the conditional probability model and power prediction model, online prediction is performed. First, the latest weather forecast data is input into the conditional probability model to predict the conditional probability distribution of spatiotemporal displacement error. Then, based on the conditional probability distribution of spatiotemporal displacement error, the weather forecast data is perturbed to obtain N perturbed weather forecast data. Finally, the N perturbed weather forecast data are input into the power prediction model to obtain N power prediction curves. Statistical analysis is performed on these N power prediction curves to obtain the power generation prediction result.

[0144] The power generation prediction method provided in this application can be executed by a power generation prediction device. This application uses the example of a power generation prediction device executing the power generation prediction method to illustrate the power generation prediction device provided in this application.

[0145] This application also provides a power generation prediction device.

[0146] like Figure 6 As shown, the power generation prediction device includes: The acquisition module 610 is used to acquire the background meteorological conditions for weather forecast data prediction; The input module 620 is used to input background meteorological conditions into a preset conditional probability model to obtain the conditional probability distribution output by the conditional probability model; the conditional probability distribution represents the probability that the weather forecast data will have a corresponding spatiotemporal displacement error under the background meteorological conditions. The perturbation module 630 is used to perturb the weather forecast data based on the conditional probability distribution to obtain the perturbed weather forecast data; The determination module 640 is used to input the disturbed weather forecast data into the preset power prediction model and determine the power generation prediction result based on the output of the power prediction model.

[0147] According to the power generation prediction device of this application, by inputting the background meteorological conditions predicted by weather forecast data into a trained conditional probability model, it is possible to predict the possible spatiotemporal displacement errors of weather forecast data and their corresponding probabilities; based on the prediction results, the weather forecast data is perturbed, and the perturbed data is input into a preset power prediction model, which can predict power generation based on the possible spatiotemporal displacement errors of weather forecast data, thereby reducing the impact of spatiotemporal displacement errors of weather forecast data on power generation prediction and improving the accuracy of power generation prediction.

[0148] In some embodiments, the input module 620 is further configured to: Output the conditional probability distribution in the following manner: Calculate the similarity between the background meteorological conditions and the historical background meteorological conditions in the historical database; the historical database stores different historical background meteorological conditions and the spatiotemporal displacement errors corresponding to different historical background meteorological conditions; The k historical background meteorological conditions whose similarity satisfies the preset conditions and the spatiotemporal displacement errors corresponding to the k historical background meteorological conditions are determined as conditional probability distributions.

[0149] In some embodiments, the input module 620 is further configured to: The conditional probability model is trained using the following method: Obtain historical weather forecast data and corresponding measured weather data; Calculate the spatiotemporal displacement error between historical weather forecast data and actual weather measurement data; The spatiotemporal displacement error between different historical weather forecast data and corresponding actual weather data, and the background meteorological conditions predicted by different historical weather forecast data are used as training datasets to train the conditional probability model.

[0150] In some embodiments, the input module 620 is further configured to: Calculate the spatiotemporal displacement error between historical weather forecast data and actual weather data, including: Extract multiple first cloud cluster objects from historical weather forecast data at different times and multiple second cloud cluster objects from actual weather measurement data at different times; Match the second cloud objects corresponding to each first cloud object at each time point to obtain multiple cloud matching pairs; Calculate the centroid displacement vector between the first and second cloud objects in each cloud pair to obtain the spatial error.

[0151] In some embodiments, the input module 620 is further configured to: Calculate the spatiotemporal displacement error between historical weather forecast data and actual weather data, including: Extract the first weather time series predicted from historical weather forecast data and the second weather time series from actual weather measurement data; The time lag required to calculate the maximum cross-correlation between the first and second weather time series is used to obtain the time error.

[0152] In some embodiments, the input module 620 is further configured to: According to the formula:

[0153] The time lag required to calculate the maximum cross-correlation between the first and second weather time series is used to obtain the time error. in, Indicates time error. Represents the correlation function. This represents the first weather time series. Indicates the amount of time lag The second weather time series after offset.

[0154] In some embodiments, the input module 620 is further configured to: Calculate the spatiotemporal displacement error between historical weather forecast data and actual weather data, including: Calculate the spatiotemporal displacement vector field between historical weather forecast data and actual weather data; the spatiotemporal displacement vector field represents the spatial displacement and time offset at different locations between historical weather forecast data and actual weather data; The average spatiotemporal displacement vector of the target region is extracted from the spatiotemporal displacement vector field and determined as the spatiotemporal displacement error.

[0155] In some embodiments, the disturbance module 630 is further configured to: The weather forecast data is perturbed based on the conditional probability distribution to obtain perturbed weather forecast data, including: Sampling is performed from the conditional probability distribution to obtain N sets of spatiotemporal displacement error vectors representing different spatiotemporal displacement errors; The weather forecast data is perturbed by using N sets of spatiotemporal displacement error vectors to obtain the perturbed weather forecast data.

[0156] In some embodiments, the disturbance module 630 is further configured to: The weather forecast data is perturbed using N sets of spatiotemporal displacement error vectors to obtain the perturbed weather forecast data, including: According to the formula:

[0157] Perturbation of weather forecast data; in, This represents the weather forecast data after the i-th group of disturbances. This represents the weather forecast data prior to the disturbance. t represents spatial location, and t represents time. Indicates background meteorological conditions The probability of the occurrence of the i-th group of spatiotemporal displacement errors. This represents the spatiotemporal displacement error vector of the i-th group. Indicates spatial error, Indicates time error.

[0158] In some embodiments, the determining module 640 is further configured to: The disturbed weather forecast data is input into a pre-set power prediction model, and the power generation prediction result is determined based on the output of the power prediction model, including: When the weather forecast data after the disturbance is a set, the output includes a power prediction curve representing the power generation at different times, and the output is determined as the power generation prediction result. Given N sets of weather forecast data after the disturbance, the output includes N power prediction curves representing power generation at different times. Statistical analysis is performed on the N power prediction curves representing power generation at different times to obtain the power generation prediction results.

[0159] The power generation prediction device in this application embodiment can be an electronic device or a component within an electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal or other devices besides a terminal. For example, the electronic device can be a photovoltaic system, a wind power generation system, a server, etc., and this application embodiment does not specifically limit it.

[0160] The power generation prediction device in this application embodiment can be a device with an operating system. This operating system can be Microsoft (Windows), Android, iOS, or other possible operating systems; this application embodiment does not specifically limit it.

[0161] In some embodiments, such as Figure 7 As shown, this application embodiment also provides an electronic device 700, including a processor 701, a memory 702, and a computer program stored in the memory 702 and executable on the processor 701. When the program is executed by the processor 701, it implements the various processes of the above-described power generation prediction method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0162] It should be noted that the electronic devices in the embodiments of this application include the aforementioned mobile electronic devices and non-mobile electronic devices.

[0163] This application also provides a non-transitory computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the various processes of the above-described power generation prediction method embodiments and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0164] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0165] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described power generation prediction method.

[0166] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0167] This application also provides a chip, which includes a processor and a communication interface. The communication interface and the processor are coupled. The processor is used to run programs or instructions to implement the various processes of the above-described power generation prediction method embodiments and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0168] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0169] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0170] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0171] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

[0172] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0173] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for predicting power generation, characterized in that, include: Obtaining background meteorological conditions from weather forecast data; The background meteorological conditions are input into a trained conditional probability model to obtain a conditional probability distribution output by the conditional probability model; the conditional probability distribution represents the probability that the weather forecast data will have a corresponding spatiotemporal displacement error under the background meteorological conditions. The weather forecast data is perturbed based on the conditional probability distribution to obtain perturbed weather forecast data; The disturbed weather forecast data is input into a preset power prediction model, and the power generation prediction result is determined based on the output of the power prediction model.

2. The method according to claim 1, characterized in that, The conditional probability model outputs the conditional probability distribution in the following manner: Calculate the similarity between the background meteorological conditions and historical background meteorological conditions in the historical database; the historical database stores different historical background meteorological conditions and the spatiotemporal displacement errors corresponding to the different historical background meteorological conditions; The k historical background meteorological conditions whose similarity satisfies the preset conditions and the spatiotemporal displacement errors corresponding to the k historical background meteorological conditions are determined as the conditional probability distribution.

3. The method according to claim 1, characterized in that, The conditional probability model is trained in the following manner: Obtain historical weather forecast data and the corresponding measured weather data; Calculate the spatiotemporal displacement error between the historical weather forecast data and the actual weather data; The conditional probability model is trained using the spatiotemporal displacement error between different historical weather forecast data and corresponding actual weather data, and the background meteorological conditions predicted by different historical weather forecast data as training datasets.

4. The method according to claim 3, characterized in that, The spatiotemporal displacement error includes spatial error; The calculation of the spatiotemporal displacement error between the historical weather forecast data and the actual weather data includes: Extract multiple first cloud cluster objects from different times predicted by the historical weather forecast data and multiple second cloud cluster objects from different times measured by the actual weather data; Match the second cloud objects corresponding to each first cloud object at each time point to obtain multiple cloud matching pairs; The spatial error is obtained by calculating the centroid displacement vector between the first cloud object and the second cloud object in each cloud matching pair.

5. The method according to claim 3, characterized in that, The spatiotemporal displacement error includes time error; The calculation of the spatiotemporal displacement error between the historical weather forecast data and the actual weather data includes: Extract the first weather time series predicted from the historical weather forecast data and the second weather time series measured from the actual weather data; The time lag required to calculate the maximum cross-correlation between the first weather time series and the second weather time series is used to obtain the time error.

6. The method according to claim 5, characterized in that, According to the formula: The time lag required to achieve the maximum cross-correlation between the first weather time series and the second weather time series is calculated to obtain the time error; in, This indicates the time error. Represents the correlation function. This represents the first weather time series. Indicates the amount of time lag The offset second weather time series.

7. The method according to claim 3, characterized in that, The calculation of the spatiotemporal displacement error between the historical weather forecast data and the actual weather data includes: Calculate the spatiotemporal displacement vector field between the historical weather forecast data and the actual weather data; the spatiotemporal displacement vector field represents the spatial displacement and time offset at different locations between the historical weather forecast data and the actual weather data; The average spatiotemporal displacement vector of the target region is extracted from the spatiotemporal displacement vector field and determined as the spatiotemporal displacement error.

8. The method according to claim 1, characterized in that, The process of perturbing the weather forecast data based on the conditional probability distribution to obtain perturbed weather forecast data includes: Sampling is performed from the conditional probability distribution to obtain N sets of spatiotemporal displacement error vectors representing different spatiotemporal displacement errors; The weather forecast data is perturbed using N sets of spatiotemporal displacement error vectors to obtain perturbed weather forecast data.

9. The method according to claim 8, characterized in that, The process of perturbing the weather forecast data using N sets of spatiotemporal displacement error vectors to obtain perturbed weather forecast data includes: According to the formula: The weather forecast data is perturbed; in, This represents the weather forecast data after the i-th group of disturbances. This represents the weather forecast data prior to the disturbance. t represents spatial location, and t represents time. Indicates background meteorological conditions The probability of the occurrence of the i-th group of spatiotemporal displacement errors. This represents the spatiotemporal displacement error vector of the i-th group. Indicates spatial error, Indicates time error.

10. The method according to claim 1, characterized in that, The disturbed weather forecast data is input into a preset power prediction model, and the power generation prediction result is determined based on the output of the power prediction model, including: When the weather forecast data after the disturbance is a set, the output result includes a power prediction curve representing the power generation at different times, and the output result is determined as the power generation prediction result; When there are N sets of weather forecast data after the disturbance, the output results include N power prediction curves representing power generation at different times. Statistical analysis is performed on the N power prediction curves representing power generation at different times to obtain the power generation prediction results.

11. A power generation prediction device, characterized in that, include: The acquisition module is used to acquire the background meteorological conditions for weather forecast data prediction; The input module is used to input the background meteorological conditions into a preset conditional probability model to obtain the conditional probability distribution output by the conditional probability model; the conditional probability distribution represents the probability that the weather forecast data will have a corresponding spatiotemporal displacement error under the background meteorological conditions. The perturbation module is used to perturb the weather forecast data based on the conditional probability distribution to obtain perturbed weather forecast data; The determination module is used to input the disturbed weather forecast data into the preset power prediction model, and determine the power generation prediction result based on the output of the power prediction model.

12. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1-10.

13. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-10.

Images