Power station power prediction method and device, equipment, storage medium and product
By fusing multi-source data and modeling spatiotemporal features, the topographic and meteorological features of the power station were extracted, which solved the problem of inaccurate power station power prediction, achieved high-precision power station power prediction, and improved the economic benefits of the power station.
Patent Information
- Application Number
- CN202511646213.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-11
- Publication Date
- 2026-02-10
AI Technical Summary
Inaccurate power generation forecasts in existing technologies lead to a decline in grid dispatch and power plant profitability, mainly due to low forecast data resolution, high computational complexity, and susceptibility to interference from the coupling relationship of meteorological factors.
By fusing multi-source data and modeling spatiotemporal features, high-resolution meteorological data is obtained, and the influence characteristics of terrain shading effect and surface attribute differentiation are extracted. Combined with convolutional neural network and long short-term memory network, the future power of the power plant is predicted.
It improves the accuracy and timeliness of power plant power forecasting, adapts to complex terrain and weather changes, reduces information transmission losses, and enhances the profitability of power plants.
Smart Images
Figure CN121507698A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present disclosure relates to the field of new energy technology, and in particular, to a power station power prediction method, device, equipment, storage medium and product. BACKGROUND
[0002] With the increasing demand for renewable energy worldwide, new energy power stations such as wind and solar energy are increasingly important in power supply. However, due to the instability of new energy power generation and the influence of weather factors, the deviation between the actual power generation of the power station and the predicted power generation previously reported to the power grid is too large, affecting the power grid dispatching, reducing the actual income of the power station, and even affecting the efficiency of the power generation equipment. As an important indicator affecting power generation, long-term high power will make the power station generate more and consume more, with limited net power generation, weakening the profitability of the power station.
[0003] In related technologies, the power of the power station is usually predicted based on forecast data output by a global weather model and satellite data, through statistical downscaling, dynamic downscaling, or a physical constraint deep learning model method. However, due to the low resolution of the forecast data, and the dependence of the statistical downscaling prediction method on historical data statistical relationships, the dynamic downscaling prediction method has high computational cost and poor timeliness, resulting in inaccurate power prediction results. The physical constraint deep learning model prediction method has high computational complexity and is difficult to deploy in real time, and is easily disturbed by the coupling relationship of meteorological elements, resulting in inaccurate power prediction. SUMMARY
[0004] To solve the above technical problems, the present disclosure provides a power station power prediction method, device, equipment, storage medium and product, which solves the problem of inaccurate power station power prediction in related technologies.
[0005] In a first aspect, the present disclosure provides a power station power prediction method, comprising: obtaining static ground data, weather forecast data corresponding to a current region, and real-time weather data of a power station in the current region monitored in real time; extracting a feature of a terrain in the current region having a shielding effect on wind speed and irradiance, a differentiated influence feature of a surface property on meteorological elements, and a meteorological change feature of different time scales according to the static ground data, the weather forecast data and the real-time weather data; determining a terrain correction coefficient matrix corresponding to the current region according to the feature having the shielding effect; obtaining high-resolution weather data after downscaling corresponding to the current region according to the static ground data, the weather forecast data, the real-time weather data, the terrain correction coefficient matrix, the differentiated influence feature, and the meteorological change feature; inputting the high-resolution weather data after downscaling into a target prediction model to predict the output power of the power station in the current region in a future time period.
[0006] In an optional implementation, the features of the terrain corresponding to the current region having a shielding effect on wind speed and irradiance, the differentiated influence features of the surface properties on meteorological elements, and the meteorological change features of different time scales are extracted according to the static ground data, the meteorological forecast data, and the real-time meteorological data, including: based on a first algorithm, the static ground data, the meteorological forecast data, and the real-time meteorological data are fused to obtain target high-resolution grid data; the features of the terrain corresponding to the current region having a shielding effect, the differentiated influence features, and the meteorological change features are extracted from the target high-resolution grid data.
[0007] In an optional implementation, the meteorological change features include first time scale meteorological change features and second time scale meteorological change features; the features of the terrain corresponding to the current region having a shielding effect, the differentiated influence features, and the meteorological change features are extracted from the target high-resolution grid data, including: the features of the terrain corresponding to the current region having a shielding effect and the differentiated influence features are extracted from the target high-resolution grid data by a convolutional neural network; the first time scale meteorological change features are extracted from the target high-resolution grid data based on a long short-term memory network, and the second time scale meteorological change features are extracted from the target high-resolution grid data based on an attention mechanism.
[0008] In an optional implementation, before the output power of the power station in the current region in the future time period is predicted by inputting the down-scaled high-resolution meteorological data into the target prediction model, the method further includes: comparing the down-scaled high-resolution meteorological data with the real-time meteorological data of the power station in the current region to obtain an error distribution map; based on the error distribution map and the real-time meteorological data of the power station in the current region, the model parameter gradient of the preset prediction model is weighted to obtain the target prediction model.
[0009] In an optional implementation, according to the features having a shielding effect, a terrain correction coefficient matrix corresponding to the current region is determined, including: according to the features having a shielding effect, the meteorological influence weight of each grid point of the high-resolution grid data in the current region is calculated by an attention mechanism; and according to the meteorological influence weight of each grid point in the current region, the terrain correction coefficient matrix corresponding to the current region is output.
[0010] In an optional implementation, the error distribution map comprises a plurality of error values between high-resolution meteorological data corresponding to different time points and different spatial points and real-time meteorological data; and the model parameter gradient of the preset prediction model is calculated based on the error distribution map and the real-time meteorological data of the power plant in the current region to obtain a target prediction model, including: grading a plurality of regions included in the current region according to the size of each error value to obtain a plurality of high-error regions and a plurality of low-error regions; assigning a first model weight to each high-error region and a second model weight to each low-error region to generate a corresponding spatial weight matrix, the first model weight being greater than the second model weight, and the spatial weight matrix corresponding to the high-resolution grid data input into the preset prediction model and the plurality of error values; and calculating the model parameter gradient of the preset prediction model based on the spatial weight matrix and the real-time meteorological data to obtain the target prediction model.
[0011] In a second aspect, the present application provides a power plant power prediction device, the power plant power prediction device comprising: an acquisition module configured to acquire static ground data corresponding to a current region, meteorological forecast data, and real-time meteorological data of a power plant in the current region.
[0012] The extraction module is configured to extract, from the static ground data, the meteorological forecast data, and the real-time meteorological data, a feature of a terrain in the current region having a shielding effect on wind speed and irradiance, a feature of a differentiated influence of a surface property on meteorological elements, and a feature of meteorological changes in different time scales.
[0013] The processing module is configured to determine a terrain correction coefficient matrix corresponding to the current region based on the feature having the shielding effect; obtain high-resolution meteorological data corresponding to the current region after being downscaled based on the static ground data, the meteorological forecast data, the real-time meteorological data, the terrain correction coefficient matrix, the feature of the differentiated influence, and the feature of the meteorological changes; and input the high-resolution meteorological data after being downscaled into a target prediction model to predict output power of the power plant in the current region in a future time period.
[0014] In a third aspect, the present application provides a computer device, comprising: a memory and a processor, which are communicatively connected to each other, and the memory stores computer instructions; and the processor executes the computer instructions to perform the power plant power prediction method of the first aspect or any of the corresponding embodiments thereof.
[0015] In a fourth aspect, the present application provides a computer readable storage medium, which stores computer instructions for causing a computer to perform the power plant power prediction method of the first aspect or any of the corresponding embodiments thereof.
[0016] In a fifth aspect, the present application provides a computer program product comprising computer instructions for causing a computer to perform the power station power prediction method of the first aspect or any of its possible implementation forms.
[0017] The technical solution provided by the embodiments of the present disclosure has the following advantages compared with the prior art: the multi-source data (static ground data, weather forecast data and real-time weather data) can be acquired and deeply fused, thereby breaking through the limitation of a single data source and improving the accuracy of power prediction; in addition, by extracting features with shielding effect, differentiated influence features and weather change features of different time scales, the influence of local wind speed and irradiance change under complex terrain on power station power is comprehensively considered, and the prediction accuracy in complex regions is improved; the down-scaled high-resolution weather data adapt to the target prediction model of wind / light power, reduces information transmission loss, and further improves the prediction accuracy. BRIEF DESCRIPTION OF DRAWINGS
[0018] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and serve to explain the principles of the present disclosure together with the specification.
[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure or the prior art, the accompanying drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative labor.
[0020] Figure 1 A schematic diagram of a power station power prediction device provided by the embodiments of the present application; Figure 2 A flowchart of a power station power prediction method according to the embodiments of the present application; Figure 3 A structure schematic diagram of a terrain attention network structure provided by the embodiments of the present application; Figure 4 A flowchart of another power station power prediction method according to the embodiments of the present application; Figure 5 A flowchart of dynamically adjusting model weights provided by the embodiments of the present application; Figure 6 A device structure diagram of a power station power prediction device provided by the embodiments of the present application; Figure 7 A structure schematic diagram of a computer device provided by the optional embodiments of the present application. DETAILED DESCRIPTION
[0021] In order to enable a more clear understanding of the above-mentioned objects, features and advantages of the present disclosure, the schemes of the present disclosure will be further described below. It should be noted that the embodiments of the present disclosure and the features in the embodiments can be combined with each other without conflict.
[0022] In the following description, a large number of specific details are set forth in order to facilitate a thorough understanding of the present disclosure, but the present disclosure can also be implemented in other manners different from those described herein; obviously, the embodiments described in the specification are only a part of the embodiments of the present disclosure, and not all the embodiments.
[0023] The embodiments of the present application are applied to the scene of power prediction of power stations.
[0024] In the related art, the power of the power station is usually predicted by statistical downscaling, dynamic downscaling or a physical constraint deep learning model method based on the forecast data output by the global weather model and satellite data. However, due to the low resolution of the forecast data, which is usually 10-50 kilometers, it is difficult to depict the local wind speed and irradiance changes under complex terrain; and the prediction method of statistical downscaling relies on the statistical relationship of historical data, and cannot effectively model the nonlinear spatiotemporal characteristics, the prediction method of dynamic downscaling has high calculation cost and poor timeliness, and is difficult to support real-time power prediction, resulting in inaccurate power prediction results; the prediction method of the physical constraint deep learning model has high computational complexity, is difficult to deploy in real time, and is easily disturbed by the coupling relationship of meteorological elements, and has insufficient adaptability to extreme weather, resulting in inaccurate power prediction. And the traditional prediction method is not optimized for micro-weather sensitive parameters (such as turbulence intensity and cloud layer mutation) of wind / solar power stations.
[0025] In order to solve the above technical problems, the present application provides a power station power prediction method, which converts low-resolution weather forecast data into high-resolution weather data by multi-source data fusion and spatiotemporal feature modeling for downscaling, thereby improving the power prediction accuracy of new energy power stations.
[0026] The power station power prediction method of the embodiments of the present application is executed by a power station power prediction device. The power station power prediction device can be any device with computing and communication functions. For example, the power station power prediction device can be a server, a cloud server. As shown in Figure 1 , Figure 1 is a schematic diagram of a power station power prediction device provided by the embodiments of the present application; in Figure 1In specific embodiments, the power station power prediction device can include a data input module, a three-stage processing module, and a result output module. The data input module is configured to obtain static ground data corresponding to a current region, weather forecast data, and real-time weather data of a power station in the current region. The three-stage processing module is configured to perform multi-source fusion prediction preprocessing (spatiotemporal difference gridding), spatiotemporal feature enhancement modeling (dynamic adaptive optimization and spatial feature extraction), and time feature extraction (multi-scale time series network) on the obtained data. The result output module is configured to output high-resolution weather data and new energy power prediction of the power station.
[0027] In specific embodiments, a power station power prediction method is provided, which can be used in the power station power prediction device described above, Figure 2 A flowchart of a power station power prediction method according to an embodiment of the present application is shown in FIG. 2. The flowchart includes the following steps: Figure 2 S201, obtaining static ground data corresponding to a current region, weather forecast data, and real-time weather data of a power station in the current region. S201, obtaining static ground data corresponding to a current region, weather forecast data, and real-time weather data of a power station in the current region.
[0028] The static ground data includes high-resolution static data, including terrain data, surface roughness data, and power station distribution data.
[0029] The weather forecast data includes low-resolution temperature data, wind speed data, and irradiance data.
[0030] The real-time weather data includes data monitored by a power station weather station, data monitored by an irradiance instrument, and Supervisory Control and Data Acquisition (SCADA) data.
[0031] S202, extracting features of the current region corresponding to the terrain having a shielding effect on wind speed and irradiance, the differentiated influence of surface properties on meteorological elements, and the meteorological change characteristics of different time scales from the static ground data, the weather forecast data, and the real-time weather data.
[0032] The target high resolution can be any kind of high-precision resolution. For example, the target high resolution can be 1 km x 1 km.
[0033] The surface properties can include vegetation, water area, and buildings.
[0034] The meteorological change characteristics of different time scales can include meteorological change characteristics of a first time scale and meteorological change characteristics of a second time scale. For example, the meteorological change characteristics include hourly meteorological evolution trends and minute-level sudden change characteristics (such as rapid movement of clouds).
[0035] In an example, the power station power prediction device fuses static ground data, weather forecast data, and real-time weather data based on a first algorithm to obtain target high-resolution grid data; and extracts features with shielding effects, differentiated influence features, and weather change features corresponding to the current region from the target high-resolution grid data.
[0036] The first algorithm is a space-time difference gridding algorithm.
[0037] Optionally, the power station power prediction device can introduce ground type embedding coding to distinguish the differentiated influence of vegetation, water area, building, and other ground properties on meteorological elements.
[0038] It can be understood that uniformly mapping multiple source data (static ground data, weather forecast data, and real-time weather data) to target high-resolution grid data can solve the inconsistency of space-time benchmarks.
[0039] In some optional embodiments, the power station power prediction device extracts features with shielding effects and differentiated influence features corresponding to the current region from the target high-resolution grid data through a convolutional neural network; extracts weather change features of a first time scale from the target high-resolution grid data based on a long short-term memory network, and extracts weather change features of a second time scale from the target high-resolution grid data based on an attention mechanism.
[0040] S203, determining a terrain correction coefficient matrix corresponding to the current region according to the features with shielding effects.
[0041] In some optional embodiments, the power station power prediction device calculates the meteorological influence weight of each grid point of the high-resolution grid data in the current region through an attention mechanism according to the features with shielding effects; and outputs the terrain correction coefficient matrix corresponding to the current region according to the meteorological influence weight of each grid point in the current region.
[0042] The terrain attention network structure corresponding to the attention mechanism includes an input layer, a processing layer, and an output layer. As shown in Figure 3 Figure 3 The embodiment of the present application provides a structure diagram of a terrain attention network structure; wherein the input layer is used for inputting low-resolution weather forecast data and high-resolution terrain data. The processing layer is used for processing the input data through a convolutional layer, and generating a terrain correction coefficient matrix corresponding to the current region according to the meteorological influence weight (terrain feature weight matrix) of each grid point in the current region. The output layer is used for outputting a high-resolution corrected meteorological field.
[0043] It can be understood that the shielding effect is one of the important factors affecting the power of the power station, for example, cloud cover will affect the power generation of the photovoltaic power station, and then affect the power of the power station. The attention mechanism can dynamically calculate the meteorological influence weight of each grid point in the current region according to the high-resolution grid data. In this way, the device can more accurately capture the influence of meteorological factors on the power of the power station, thereby improving the accuracy of power prediction and reducing the increase in cost caused by prediction error.
[0044] In S204, the high-resolution meteorological data corresponding to the current region after the scale reduction is obtained according to the static ground data, the meteorological forecast data, the real-time meteorological data, the terrain correction coefficient matrix, the differentiated influence feature, and the meteorological change feature.
[0045] In an example, taking the wind speed scale reduction of a mountain wind farm as an example, in this scenario, the terrain correction coefficient matrix (1km resolution wind speed correction field) can be output by a terrain attention network (through an attention mechanism) according to the static ground data (30m resolution digital elevation model data), the meteorological forecast data (9km resolution wind speed forecast), and the real-time meteorological data (20 wind measurement towers in the wind farm). At the same time, the wind measurement tower observation data (real-time wind speed measured on site) is introduced as a reference, the mean square error is used as a loss function, the mean square error between the predicted value of the 1km wind speed correction field and the real value of the wind measurement tower is calculated to measure the accuracy of the corrected wind speed data, and the model parameters are continuously optimized based on the error feedback.
[0046] In another example, taking the irradiance scale reduction of a desert photovoltaic power station as an example, in this scenario, the high-resolution ground albedo map, the meteorological forecast data (15km resolution irradiance forecast), and the real-time meteorological data (photovoltaic string-level current monitoring data) can be used as input to the long short-term memory model to capture the daily cycle change, and the Transformer layer is used to model the rapid shielding event of dust. When a sandstorm is detected, the sandstorm mode sub-model is enabled, and the aerosol optical depth correction is introduced to obtain the high-resolution meteorological data corresponding to the current region after the scale reduction.
[0047] Understandably, by integrating static ground data, dynamic meteorological data, and localized correction coefficients, macroscopic meteorological data is downscaled to meter- or hundred-meter resolution, transforming meteorological data from covering a broad area to precisely matching the location of each photovoltaic panel and wind turbine in the power station. The downscaled, high-resolution illumination data can accurately capture the impact of localized cloud cover and mountain shadow movement on different photovoltaic arrays (for example, when the array on the east side of the power station is blocked by a mountain, the array on the west side still receives normal sunlight), avoiding prediction bias caused by a uniform illumination assumption for the entire station and reducing power prediction errors. High-resolution wind speed and direction data can match the location of each wind turbine, accurately calculating the actual wind energy captured at different locations, avoiding inaccurate power predictions due to misjudgments of average wind speed.
[0048] S205 inputs downscaled high-resolution meteorological data into the target prediction model to predict the output power of power plants in the current region over a future period.
[0049] The target prediction model can be used to predict the output power of power plants in the current region over a future time period. Optionally, the target prediction model can also be used to calculate prediction uncertainty and output power curves.
[0050] The future time period can be set according to actual needs and is not limited. For example, the future time period can be 72 hours.
[0051] For example, a power plant power prediction device inputs downscaled high-resolution meteorological data into a target prediction model to predict the power output of power plants in the current region over a period of 72 hours.
[0052] based on Figure 2 The method described involves a power plant power prediction device acquiring static ground data, weather forecast data, and real-time weather data of the power plant in the current region. Based on the static ground data, weather forecast data, and real-time weather data, the device extracts features of the terrain's shading effect on wind speed and irradiance, the differential influence of surface attributes on meteorological elements, and meteorological change characteristics at different time scales. Based on the shading effect features, a terrain correction coefficient matrix is determined for the current region. Using the static ground data, weather forecast data, real-time weather data, terrain correction coefficient matrix, differential influence characteristics, and meteorological change characteristics, downscaled high-resolution meteorological data for the current region is obtained. This downscaled high-resolution meteorological data is then input into the target prediction model to predict the power output of the power plant in the current region over future time periods.
[0053] Since the multi-source data (static ground data, weather forecast data and real-time weather data) can be acquired and deeply fused, the limitation of a single data source is broken through, and the accuracy of power prediction is improved. In addition, by extracting features with shielding effect, differentiated impact features, and weather change features of different time scales, the influence of local wind speed and irradiance change under complex terrain on power station power is comprehensively considered, and the prediction accuracy in complex regions is improved. The high-resolution weather data after downscaling is adapted to the target prediction model of wind / light power, reduces the information transmission loss, and further improves the prediction accuracy.
[0054] In an optional example, on the basis of the foregoing embodiment, as introduced in the foregoing, before the high-resolution weather data after downscaling is input into the target prediction model to predict the output power of the power station in the current region in the future time period, the power station power prediction device can further perform the following steps, as shown in Figure 4 Figure 4 For a flowchart of another power station power prediction method according to an embodiment of the application, comprising: S401, comparing the high-resolution weather data after downscaling with the real-time weather data of the power station in the current region to obtain an error distribution map.
[0055] The error distribution map includes a plurality of error values between the high-resolution weather data and the real-time weather data corresponding to different time points and different spatial points.
[0056] S402, based on the error distribution map and the real-time weather data of the power station in the current region, performing weighted calculation on the model parameter gradient of the preset prediction model to obtain a target prediction model.
[0057] In some optional embodiments, the power station power prediction device grades a plurality of regions included in the current region according to the size of each error value to obtain a plurality of high-error regions and a plurality of low-error regions; assigns a first model weight to each high-error region and a second model weight to each low-error region to generate a corresponding spatial weight matrix; and performs weighted calculation on the model parameter gradient of the preset prediction model according to the spatial weight matrix and the real-time weather data to obtain a target prediction model.
[0058] The first model weight is greater than the second model weight, and the spatial weight matrix corresponds to the high-resolution grid data input into the preset prediction model and the plurality of error values.
[0059] It can be understood that the meteorological conditions and terrain and other factors in different regions are complex and changeable, and the error distribution map can reflect the error characteristics of different regions in the current region. According to the error value, the region is classified and different model weights are allocated to generate a spatial weight matrix, which can make the model better consider the influence of regional differences and complex environment. For high error areas, a larger weight is given, which means that the model will pay more attention to the meteorological data and error situation of these areas, so as to better adapt to complex terrain, climate and other conditions and improve the prediction performance of the model in different environments.
[0060] Optionally, the power station power prediction device can dynamically adjust the model weight through the online meta-learning controller. As shown in Figure 5 , Figure 5 a flowchart for dynamically adjusting the model weight provided by the embodiment of the application; in Figure 5 , the pre-trained model predicts the meteorological or energy data based on its own parameters, and the difference between the predicted value and the real-time observed true value is obtained to obtain the error value. Based on the error value, the error distribution map is calculated and generated. According to the result of the error distribution map, the gradient of the model parameter is calculated through the meta-learning controller, and different weights are allocated to different regions to generate a weight matrix. Subsequently, the weight matrix is combined with the input model meteorological field data to realize the targeted adjustment of the model parameters; after error analysis and weight adjustment, the new model parameters are fed back to the pre-trained cross-scale model to complete a model update. After that, the updated model will enter the cycle again, and the above process will be repeated based on the new field observation data to iteratively optimize in time, ensuring that the model can always adapt to the changes of the actual scene and continuously improve the prediction accuracy.
[0061] Based on the method shown in Figure 4 , by comparing the reduced high-resolution meteorological data with the real-time meteorological data, the error distribution map is obtained, which can clearly understand the distribution of the difference between the predicted data and the actual data in time and space. Based on this error information, the parameter gradient of the preset prediction model is weighted calculated, which can adjust the model parameters in a targeted manner, so that the model can better fit the actual situation, thereby improving the accuracy of the prediction model. For example, in the area and time point with large error, the model parameters are adjusted by a larger weight, which can make the model pay more attention to these key parts and reduce the prediction error.
[0062] In this embodiment, a power station power prediction device is also provided, which is used to implement the above-mentioned embodiments and preferred embodiments, and will not be described again. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware or a combination of software and hardware is also possible and contemplated.
[0063] The embodiment provides a power station power prediction device, which comprises Figure 6 as shown in the figure, Figure 6 A device structure diagram of the power station power prediction device provided by the embodiment is shown in the figure. The device comprises: An acquisition module 601 is configured to acquire static ground data corresponding to a current region, meteorological forecast data, and real-time meteorological data of a power station in the current region.
[0064] An extraction module 602 is configured to extract, according to the static ground data, the meteorological forecast data, and the real-time meteorological data, a feature of a terrain corresponding to the current region having a shielding effect on wind speed and irradiance, a differentiated influence feature of a surface property on a meteorological element, and a meteorological change feature of different time scales.
[0065] A processing module 603 is configured to determine a terrain correction coefficient matrix corresponding to the current region according to the feature having the shielding effect; obtain high-resolution meteorological data of the current region after being reduced in scale according to the static ground data, the meteorological forecast data, the real-time meteorological data, the terrain correction coefficient matrix, the differentiated influence feature, and the meteorological change feature; and input the high-resolution meteorological data after being reduced in scale into a target prediction model to predict output power of the power station in the current region in a future time period.
[0066] In some optional embodiments, the extraction module 602 is specifically configured to perform fusion processing on the static ground data, the meteorological forecast data, and the real-time meteorological data based on a first algorithm to obtain target high-resolution grid data; and extract, from the target high-resolution grid data, the feature having the shielding effect, the differentiated influence feature, and the meteorological change feature corresponding to the current region.
[0067] In some optional embodiments, the meteorological change feature comprises a meteorological change feature of a first time scale and a meteorological change feature of a second time scale; the extraction module 602 is specifically configured to extract, from the target high-resolution grid data, the feature having the shielding effect and the differentiated influence feature corresponding to the current region by using a convolutional neural network; extract the meteorological change feature of the first time scale from the target high-resolution grid data based on a long short-term memory network, and extract the meteorological change feature of the second time scale from the target high-resolution grid data based on an attention mechanism.
[0068] In some optional implementations, before inputting the downscaled high-resolution meteorological data into the target prediction model to predict the output power of the power station in the current region in the future time period, the processing module 603 is also used to compare the downscaled high-resolution meteorological data with the real-time meteorological data of the power station in the current region to obtain an error distribution map; based on the error distribution map and the real-time meteorological data of the power station in the current region, the gradient of the model parameters of the preset prediction model is weighted and calculated to obtain the target prediction model.
[0069] In some optional implementations, the processing module 603 is specifically used to calculate the meteorological influence weight of each grid point in the current region based on the feature with shading effect through an attention mechanism; and to output the terrain correction coefficient matrix corresponding to the current region based on the meteorological influence weight of each grid point in the current region.
[0070] In some optional implementations, the error distribution map includes multiple error values between high-resolution meteorological data and real-time meteorological data corresponding to different time points and different spatial points; the processing module 603 is specifically used to classify the multiple regions included in the current region according to the magnitude of each error value, to obtain multiple high-error regions and multiple low-error regions; to assign a first model weight to each high-error region and a second model weight to each low-error region, generating a corresponding spatial weight matrix, wherein the first model weight is greater than the second model weight, and the spatial weight matrix corresponds to the high-resolution grid data and multiple error values input to the preset prediction model; and to perform weighted calculation of the model parameter gradient of the preset prediction model according to the spatial weight matrix and the real-time meteorological data, to obtain the target prediction model.
[0071] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.
[0072] In this embodiment, the power plant power prediction device is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.
[0073] This invention also provides a computer device having the above-described features. Figure 6 The device shown is a power prediction device for a power plant.
[0074] Please see Figure 7 , Figure 7 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 7As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 7 Take a processor 10 as an example.
[0075] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.
[0076] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.
[0077] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.
[0078] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.
[0079] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.
[0080] This invention provides a computer program product, which includes computer instructions for causing a computer to execute the method of any embodiment of this invention.
[0081] This invention can be implemented in many ways, including as: a process; an apparatus; a system; a composition of matter; a computer program product embodied on a computer-readable storage medium; and / or a processor, such as a processor configured to execute instructions stored in and / or provided by memory coupled to the processor. In this specification, these embodiments or any other form in which the invention may take may be referred to as technology. Generally, the order of steps of the disclosed process can be varied within the scope of this invention. Unless otherwise stated, components described as configured to perform a task (such as a processor or memory) can be implemented as general components temporarily configured to perform a task at a given time, or manufactured as specific components to perform a task. As used herein, the term 'processor' refers to one or more devices, circuits, and / or processing cores configured to process data (such as computer program instructions).
[0082] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, 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 said element.
[0083] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for predicting power generation of a power plant, characterized in that, The method includes: Acquire static ground data, weather forecast data, and real-time weather data of power plants in the current region; Based on the static ground data, the weather forecast data, and the real-time weather data, the characteristics of the shading effect of the terrain on wind speed and irradiance corresponding to the current region, the characteristics of the differential influence of surface attributes on meteorological elements, and the characteristics of meteorological changes at different time scales are extracted. Based on the characteristics of the shading effect, determine the terrain correction coefficient matrix corresponding to the current region; Based on the static ground data, the weather forecast data, the real-time weather data, the terrain correction coefficient matrix, the differential impact characteristics, and the weather change characteristics, the downscaled high-resolution weather data corresponding to the current region is obtained. The downscaled high-resolution meteorological data is input into the target prediction model to predict the output power of the power plant in the current region in the future time period.
2. The method according to claim 1, characterized in that, The step of extracting, based on the static ground data, the weather forecast data, and the real-time weather data, the characteristics of the shading effect of the terrain on wind speed and irradiance in the current region, the characteristics of the differentiated influence of surface attributes on meteorological elements, and the characteristics of meteorological changes at different time scales, includes: Based on the first algorithm, the static ground data, the weather forecast data, and the real-time weather data are fused to obtain target high-resolution grid data. Extract the features with shading effect, the differential impact features, and the meteorological change features corresponding to the current region from the target high-resolution grid data.
3. The method according to claim 2, characterized in that, The meteorological change features include meteorological change features at a first time scale and meteorological change features at a second time scale; the extraction of the shading effect features, the differential impact features, and the meteorological change features corresponding to the current region from the target high-resolution grid data includes: The occlusion effect features and differential influence features corresponding to the current region are extracted from the target high-resolution grid data using a convolutional neural network. Based on a long short-term memory network, meteorological change features at the first time scale are extracted from the target high-resolution grid data, and meteorological change features at the second time scale are extracted from the target high-resolution grid data based on an attention mechanism.
4. The method according to any one of claims 1-3, characterized in that, Before inputting the downscaled high-resolution meteorological data into the target prediction model to predict the output power of the power plants in the current region over a future time period, the method further includes: The downscaled high-resolution meteorological data is compared with the real-time meteorological data of the power station in the current region to obtain an error distribution map; Based on the error distribution map and real-time meteorological data of the power station in the current region, the gradient of the model parameters of the preset prediction model is weighted and calculated to obtain the target prediction model.
5. The method according to claim 4, characterized in that, The step of determining the terrain correction coefficient matrix corresponding to the current region based on the shading effect features includes: Based on the shading effect, the meteorological influence weight of each grid point corresponding to the high-resolution grid data in the current region is calculated through an attention mechanism. Based on the meteorological influence weight of each grid point in the current region, the terrain correction coefficient matrix corresponding to the current region is output.
6. The method according to claim 4, characterized in that, The error distribution map includes multiple error values between the high-resolution meteorological data and the real-time meteorological data corresponding to different time points and different spatial points; the step of weighted calculation of the model parameter gradient of the preset prediction model based on the error distribution map and the real-time meteorological data of the power station in the current region to obtain the target prediction model includes: The current region is classified into multiple regions based on the magnitude of each error value, resulting in multiple high-error regions and multiple low-error regions. A first model weight is assigned to each high-error region, and a second model weight is assigned to each low-error region to generate a corresponding spatial weight matrix. The first model weight is greater than the second model weight, and the spatial weight matrix corresponds to the high-resolution grid data and multiple error values input to the preset prediction model. Based on the spatial weight matrix and the real-time meteorological data, the gradient of the model parameters of the preset prediction model is calculated by weighting to obtain the target prediction model.
7. A power plant power prediction device, characterized in that, The power prediction device for the power plant includes: The acquisition module is used to acquire static ground data, weather forecast data, and real-time weather data of power plants in the current region. The extraction module is used to extract, based on the static ground data, the weather forecast data, and the real-time weather data, the characteristics of the shading effect of the terrain on wind speed and irradiance corresponding to the current region, the characteristics of the differential influence of surface attributes on meteorological elements, and the characteristics of meteorological changes at different time scales. The processing module is used to determine the terrain correction coefficient matrix corresponding to the current region based on the characteristics with the shading effect; to obtain downscaled high-resolution meteorological data corresponding to the current region based on the static ground data, the meteorological forecast data, the real-time meteorological data, the terrain correction coefficient matrix, the differential influence characteristics, and the meteorological change characteristics; and to input the downscaled high-resolution meteorological data into the target prediction model to predict the output power of the power station in the current region in the future time period.
8. A computer device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the power prediction method for a power plant as described in any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, it implements the steps of the power prediction method for a power plant as described in any one of claims 1 to 6.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the power prediction method for a power plant as described in any one of claims 1 to 6.