Power parameter prediction method and device, computer device, readable storage medium and program product

Abstract

The application relates to a power parameter prediction method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: obtaining historical power parameters and first meteorological data of a target power region and related reference power regions in a historical time period, the first meteorological data comprising multi-dimensional initial meteorological characteristics; assigning initial meteorological weights to each power region, and fusing the initial meteorological characteristics based on the initial meteorological weights to obtain target meteorological characteristics; predicting a predicted power parameter of the target power region based on the target meteorological characteristics, and calculating a cost function value according to the predicted power parameter and the historical power parameter; when the cost function value does not satisfy a preset condition, iteratively adjusting the initial meteorological weights until the target meteorological weights are obtained; obtaining second meteorological data of a future time period, and predicting the target power parameter according to the target meteorological weights and the second meteorological data. The method can improve the power parameter prediction efficiency.

Description

Technical Field

[0001] This application relates to the field of power technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for predicting power parameters. Background Technology

[0002] With the development of new energy power generation and inter-regional power transmission technologies, target power parameters are affected by various factors such as load, new energy output, unit status, and meteorological conditions. To improve forecast accuracy, it is usually necessary to combine meteorological data from multiple regions for prediction.

[0003] In traditional techniques, meteorological data from multiple regions are usually directly spliced ​​together, simply averaged, or input into the prediction model after setting fixed weights based on human experience.

[0004] However, the impact of meteorological data on target power parameters varies across different regions and changes with trading hours and weather conditions. Traditional methods, to cover these differences, typically require incorporating a large amount of regional meteorological data and related features, which easily increases the model's input dimensionality, data processing volume, and computational resource consumption, impacting the efficiency of power parameter prediction. Summary of the Invention

[0005] Therefore, it is necessary to provide a power parameter prediction method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can improve the efficiency of power parameter prediction in response to the above-mentioned technical problems.

[0006] In a first aspect, this application provides a method for predicting power parameters, including:

[0007] Historical power parameters and first meteorological data of the target power area and a reference power area within a historical time period are obtained; the reference power area is a power area related to the target power area; the first meteorological data includes multidimensional initial meteorological characteristics.

[0008] Initial meteorological weights are assigned to each of the reference power regions and the target power regions, and each of the initial meteorological features is fused according to the initial meteorological weights to obtain the target meteorological features;

[0009] Based on the target meteorological characteristics, predict the predicted power parameters of the target power area within a historical time period, and calculate the cost function value by combining the predicted power parameters with the historical power parameters.

[0010] When the cost function value does not meet the preset conditions, the initial meteorological weight is adjusted according to the cost function value, and the predicted power parameters of the target power area in the historical time period are re-predicted according to the adjusted initial meteorological weight, until the cost function value meets the preset conditions, and the target meteorological weight is obtained.

[0011] Acquire second meteorological data for a future time period, and predict the target power parameters corresponding to the target power area for the future time period based on the target meteorological weight and the second meteorological data.

[0012] In one embodiment, the method further includes:

[0013] Obtain reference power parameters;

[0014] The prediction of predicted power parameters for the target power area within a historical time period based on the target meteorological characteristics includes:

[0015] Based on the target meteorological characteristics and the reference power parameters, meteorological correlation characteristics are generated; the meteorological correlation characteristics include one or more of the following: time-series correlation characteristics, scenario correlation characteristics, regional correlation characteristics, and power grid operation correlation characteristics.

[0016] Based on the meteorological correlation characteristics, predict the power parameters of the target power area within the historical time period.

[0017] In one embodiment, assigning initial meteorological weights to each of the reference power areas and the target power areas includes at least one of the following:

[0018] Based on the number of reference power areas and target power areas, the initial meteorological weights corresponding to each reference power area and target power area are determined;

[0019] Calculate the historical correlation score between the initial meteorological characteristics of each reference power region and the historical power parameters of the target power region; based on the historical correlation score, determine the initial meteorological weights corresponding to each reference power region and the target power region;

[0020] Calculate the historical frequency of abnormal power parameters in the target power area when extreme weather occurs in the reference power area and the target power area; based on the historical frequency, determine the initial meteorological weight corresponding to each reference power area and the target power area;

[0021] The initial meteorological weights corresponding to each of the reference power regions and the target power regions are randomly generated.

[0022] In one embodiment, after assigning initial meteorological weights to each of the reference power regions and the target power region, the method further includes:

[0023] Based on the initial meteorological weights, weight coding information is generated, and based on the weight coding information, an initial population is generated.

[0024] In one embodiment, adjusting the initial meteorological weights based on the cost function value, and adjusting the initial meteorological weights accordingly, includes:

[0025] Based on the cost function value corresponding to each weight encoding information, parent weight encoding information is selected from the initial population;

[0026] The parent generation weight encoding information is cross-processed according to a preset cross-operation to obtain the child generation weight encoding information;

[0027] The offspring weight encoding information is mutated according to a preset mutation operation to obtain an updated population; the updated population includes multiple updated weight encoding information; each updated weight encoding information represents the updated initial meteorological weight.

[0028] In one embodiment, the step of performing cross-processing on the parent generation weight encoding information according to a preset cross-processing operation to obtain the child generation weight encoding information includes:

[0029] Based on the regional attributes of each reference power region and the stability of the meteorological weights corresponding to each reference power region in multiple historical high-quality weight coding information, the gene crossover probability corresponding to each reference power region is determined.

[0030] For each of the reference power regions, based on the corresponding gene crossover probability, crossover is performed on the genes corresponding to the reference power regions in at least two of the parent weight coding information to obtain the offspring genes corresponding to the reference power regions in the offspring weight coding information.

[0031] The offspring weight encoding information is generated based on the offspring genes corresponding to each of the reference power regions.

[0032] In one embodiment, the step of performing mutation processing on the offspring weight encoding information according to a preset mutation operation to obtain an updated population includes:

[0033] Based on at least one of the following: the historical frequency of extreme weather events in each reference power region, the contribution of abnormal power parameters corresponding to each reference power region, the prediction error corresponding to each reference power region, and the stability of the corresponding gene in high-quality weighted coding information over multiple generations, the mutation probability and mutation magnitude of the corresponding gene in each reference power region are determined.

[0034] Based on the mutation probability of the corresponding gene in each of the reference power regions, the gene to be mutated is determined from the offspring weight encoding information;

[0035] Based on the mutation amplitude corresponding to the gene to be mutated, the gene to be mutated is randomly perturbed to obtain the mutated offspring weight encoding information.

[0036] The gene values ​​in the mutated offspring weight encoding information are subjected to non-negative processing and normalization to obtain the updated population.

[0037] In one embodiment, the predicted power parameters include predicted power parameters for a preset scenario; the step of calculating the cost function value by combining the predicted power parameters with the historical power parameters includes:

[0038] Calculate the mean absolute error between the predicted power parameters and the historical power parameters;

[0039] Calculate the preset scenario error value between the predicted power parameters of the preset scenario and the historical power parameters;

[0040] The cost function value is determined by calculating based on the mean absolute error and the preset scenario error value.

[0041] In one embodiment, the preset scenario includes a parameter peak scenario and a line congestion scenario; calculating the preset scenario error value between the predicted power parameters of the preset scenario and the historical power parameters includes:

[0042] Calculate the first sub-error value between the parameter peak scenario and the historical power parameters;

[0043] Calculate the second sub-error value between the line congestion scenario and the historical power parameters;

[0044] The preset scenario error value is obtained based on the first sub-error value, the second sub-error value, and the constraint penalty value; the constraint penalty value is determined based on the current initial meteorological weight and the initial meteorological weight in the previous iteration process.

[0045] In one embodiment, after predicting the target power parameters of the target power region for a future time period, the method further includes:

[0046] Obtain the extreme weather indicators, power grid operation risk indicators, and target weather weights for each of the reference power areas for future time periods;

[0047] If the extreme weather indicator represents that at least one of the reference power areas has experienced extreme weather, and the target weather weight corresponding to the reference power area where the extreme weather occurred is greater than a preset weight threshold, then the future time period is determined to meet the preset correction condition.

[0048] If the preset correction conditions are met in the future time period, the target power parameters, the target meteorological characteristics, the extreme weather indicators, and the power grid operation risk indicators are input into a pre-trained residual correction model to obtain the prediction residuals corresponding to the target power parameters.

[0049] The target power parameters are corrected based on the predicted residuals to obtain the corrected target power parameters.

[0050] Secondly, this application also provides a power parameter prediction device, comprising:

[0051] The acquisition module is used to acquire historical power parameters and first meteorological data of the target power area and the reference power area within a historical time period; the reference power area is a power area related to the target power area; the first meteorological data includes multidimensional initial meteorological features;

[0052] The feature fusion module is used to assign initial meteorological weights to each of the reference power areas and the target power areas respectively, and to fuse each of the initial meteorological features according to the initial meteorological weights to obtain the target meteorological features;

[0053] The cost calculation module is used to predict the predicted power parameters of the target power area within a historical time period based on the target meteorological characteristics, and calculate the cost function value by combining the predicted power parameters with the historical power parameters.

[0054] An iterative module is used to adjust the initial meteorological weights according to the cost function values ​​when the cost function values ​​do not meet the preset conditions, and to re-predict the predicted power parameters of the target power area within the historical time period according to the adjusted initial meteorological weights, until the cost function values ​​meet the preset conditions, thereby obtaining the target meteorological weights.

[0055] The prediction module is used to acquire second meteorological data for a future time period, and predict the target power parameters corresponding to the target power area in the future time period based on the target meteorological weight and the second meteorological data.

[0056] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method in any of the above embodiments.

[0057] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method in any of the above embodiments.

[0058] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the methods in any of the above embodiments.

[0059] The aforementioned power parameter prediction method, apparatus, computer equipment, computer-readable storage medium, and computer program product acquire historical power parameters and first meteorological data for a target power region and reference power regions within a historical time period. Initial meteorological weights are assigned to the target power region and each reference power region. Based on these initial meteorological weights, meteorological characteristics from multiple regions are fused to obtain target meteorological characteristics. This compresses multidimensional meteorological data from different regions into fused meteorological characteristics that reflect regional differences in impact, reducing data redundancy and processing burden caused by directly inputting meteorological data from multiple regions into the prediction model. Subsequently, the predicted power parameters for the target power region within the historical time period are predicted based on the target meteorological characteristics. A cost function value is calculated based on the predicted power parameters and historical power parameters. The initial meteorological weights are then iteratively adjusted based on the cost function value until the target meteorological weights are obtained. This allows for adaptive determination of the impact degree of meteorological data from each region using historical prediction errors, making subsequent meteorological fusion more consistent with the actual operating patterns of the target power region. Finally, second meteorological data for a future time period are fused based on the target meteorological weights, and the target power parameters are predicted based on the fusion results. In this way, without increasing the input dimensions of the prediction model, the amount of meteorological data processing and computing resource consumption can be reduced, and the processing efficiency of target power parameter prediction can be improved. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 This is a diagram illustrating the application environment of the power parameter prediction method in one embodiment;

[0062] Figure 2 This is a flowchart illustrating a power parameter prediction method in one embodiment;

[0063] Figure 3This is a flowchart illustrating the power parameter prediction step in an exemplary embodiment.

[0064] Figure 4 This is a structural block diagram of a power parameter prediction device in one embodiment;

[0065] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0066] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0067] like Figure 1 As shown, a method for predicting power parameters is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0068] Step 202: Obtain historical power parameters and first meteorological data for the target power area and the reference power area within a historical time period; the reference power area is a power area related to the target power area; the first meteorological data includes multidimensional initial meteorological characteristics.

[0069] The target power region refers to the power region for which power parameter prediction is required. The target power region can be a pre-defined administrative region, power supply region, power dispatch region, or the region where a grid node is located. The reference power region refers to other power regions whose meteorological conditions may affect the power parameters of the target power region. For example, the reference power region may include power regions adjacent to the target power region, power regions connected to the target power region by tie lines, power regions supplying renewable energy to the target power region, or power regions with historical power parameter correlations with the target power region.

[0070] Historical power parameters refer to the actual power operation parameters generated in the target power area within a historical time period. These parameters may include at least one of the following: load, renewable energy output, tie-line power, supply-demand difference, power parameter fluctuations, or other parameters that can characterize the operating status of the target power area. Historical power parameters can be recorded according to a preset time granularity, such as by hour, half hour, or fifteen minutes, and include the actual power parameters corresponding to each historical moment for subsequent evaluation of the prediction results corresponding to meteorological weights.

[0071] The first meteorological data refers to the meteorological data collected from the target power area and each reference power area within a historical time period. The first meteorological data for each power area includes multiple meteorological feature dimensions, which may include at least one of temperature, humidity, wind speed, wind direction, cloud cover, precipitation, irradiance, and air pressure. Therefore, the first meteorological data can be represented as multidimensional data composed of time dimension, regional dimension, and meteorological feature dimension.

[0072] Optionally, historical power parameters of the target power area can be obtained from the power dispatching system, power operation monitoring system, or power data platform, and first meteorological data of the target power area and each reference power area can be obtained from the meteorological observation station, meteorological service platform, or meteorological forecasting system.

[0073] In this embodiment, since the power parameters of the target power area are affected not only by local meteorological conditions but also by load changes in surrounding areas, changes in renewable energy output, and cross-regional power transmission, acquiring only the meteorological data of the target power area itself may not fully reflect the meteorological information affecting its operational status. By simultaneously acquiring the first meteorological data of the reference power area, the impact of meteorological changes in other power areas on the power parameters of the target power area through load, renewable energy output, or cross-regional power transmission can be further reflected, providing a data foundation for subsequently determining the meteorological weights corresponding to different power areas.

[0074] Step 204: Assign initial meteorological weights to each reference power area and the target power area, and fuse each initial meteorological feature according to the initial meteorological weights to obtain the target meteorological feature.

[0075] Since the meteorological conditions in different power areas have varying degrees of impact on the target power area, it is necessary to assign initial meteorological weights to each power area separately. Furthermore, for each meteorological feature dimension, the initial meteorological features of each power area under the same dimension are weighted and fused to obtain the target meteorological features. These target meteorological features may include fused temperature, fused humidity, fused wind speed, and fused irradiance, among others.

[0076] Optionally, the initial meteorological weights can be configured according to preset rules. For example, the same initial meteorological weight can be assigned to each power area based on the number of power areas; alternatively, the corresponding initial meteorological weights can be determined based on the correlation between the initial meteorological characteristics of each power area and historical power parameters.

[0077] Optionally, the initial meteorological weights can also be used as gene values ​​for weight encoding information in a genetic algorithm. Multiple sets of initial meteorological weights can form multiple weight encoding information and together constitute an initial population, so that each initial meteorological weight can be iteratively optimized based on the prediction error in subsequent iterations.

[0078] Step 206: Based on the target meteorological characteristics, predict the predicted power parameters of the target power area within the historical time period, and calculate the cost function value by combining the predicted power parameters with the historical power parameters.

[0079] The cost function value is a numerical value used to characterize the degree of difference between the predicted result and the actual result corresponding to the current initial meteorological weights. The smaller the cost function value, the more reasonable the meteorological fusion method corresponding to the current initial meteorological weights.

[0080] Optionally, after obtaining the target meteorological characteristics, the target meteorological characteristics can be input together with other power operation characteristics such as load, new energy output, and unit status into a pre-trained power parameter prediction model to obtain the predicted power parameters of the target power area within a historical time period.

[0081] Since historical power parameters are actual power parameters recorded within a historical time period, the cost function value corresponding to the current initial meteorological weight can be calculated based on the difference between the predicted power parameters and the corresponding historical power parameters.

[0082] Optionally, the absolute difference between the predicted power parameters and the actual power parameters can be calculated, and the average of the absolute differences corresponding to multiple historical times can be determined as the cost function value, that is, the cost function value can be calculated using the mean absolute error.

[0083] Alternatively, the squared difference between the predicted power parameters and the actual power parameters can be calculated, and the squared difference can be used as the cost function value.

[0084] Step 208: When the cost function value does not meet the preset conditions, adjust the initial meteorological weights according to the cost function value, and re-predict the predicted power parameters of the target power area within the historical time period according to the adjusted initial meteorological weights, until the cost function value meets the preset conditions and the target meteorological weights are obtained.

[0085] When the cost function value does not meet the preset conditions, it indicates that the meteorological fusion method corresponding to the current initial meteorological weights has not yet achieved the expected results. In this case, the initial meteorological weights are adjusted, and the first meteorological data is re-fused based on the adjusted initial meteorological weights to obtain updated target meteorological features. Then, based on the updated target meteorological features, the predicted power parameters for the target power area within the historical time period are re-predicted, and the corresponding cost function value is recalculated. If the recalculated cost function value still does not meet the preset conditions, the meteorological weights are adjusted again until the cost function value meets the preset conditions, and the corresponding meteorological weight at this point is determined as the target meteorological weight.

[0086] Optionally, the initial meteorological weights can be adjusted according to a preset adjustment range. For example, if the preset adjustment range is 0.1, the meteorological weights corresponding to at least one power area can be increased or decreased by 0.1 in each adjustment process, and the adjusted meteorological weights can be normalized.

[0087] Optionally, if a genetic algorithm is used to adjust the meteorological weights, parent weight encoding information can be selected from the current population based on the cost function values ​​corresponding to each set of meteorological weights. This parent weight encoding information can then be cross-crossed and mutated to obtain multiple sets of updated meteorological weights. Subsequently, the cost function values ​​are recalculated based on each set of updated meteorological weights. Selection, cross-crossing, and mutation continue until the preset iteration termination condition is met. The set of updated meteorological weights whose cost function values ​​satisfy the preset optimization condition is determined as the target meteorological weights.

[0088] Step 210: Obtain the second meteorological data for the future time period, and predict the target power parameters corresponding to the target power area in the future time period based on the target meteorological weight and the second meteorological data.

[0089] The second meteorological data refers to the predicted meteorological data for the target power area and each reference power area in the future time period. The second meteorological data can be provided by a meteorological forecasting system, a meteorological service platform, or a numerical weather prediction system, and can include multi-dimensional meteorological characteristics such as temperature, humidity, wind speed, wind direction, cloud cover, precipitation, and irradiance. Therefore, the second meteorological data and the first meteorological data can contain the same dimensions of meteorological characteristics. The difference lies in that the first meteorological data is the actual meteorological data collected within a historical time period, while the second meteorological data is the predicted meteorological data for the future time period.

[0090] Optionally, based on the target meteorological weights, the second meteorological data of the target power region and each reference power region under the same meteorological feature dimension are weighted and fused to obtain the fused meteorological features corresponding to the future time period. Then, the fused meteorological features, along with load forecast data, renewable energy output forecast data, and other power operation features such as unit status corresponding to the future time period, are input into a pre-trained power parameter prediction model to obtain the target power parameters of the target power region corresponding to the future time period.

[0091] In the above embodiments, historical power parameters and first meteorological data for the target power region and reference power regions within a historical time period are obtained. Initial meteorological weights are assigned to the target power region and each reference power region. Based on these initial meteorological weights, the meteorological characteristics of multiple regions are fused to obtain the target meteorological characteristics. This compresses multidimensional meteorological data from different regions into fused meteorological characteristics that reflect regional differences in impact, reducing data redundancy and processing burden caused by directly inputting meteorological data from multiple regions into the prediction model. Then, based on the target meteorological characteristics, the predicted power parameters for the target power region within the historical time period are predicted. A cost function value is calculated based on the predicted power parameters and historical power parameters. The initial meteorological weights are then iteratively adjusted based on the cost function value until the target meteorological weights are obtained. This allows for the adaptive determination of the impact degree of meteorological data in each region using historical prediction errors, making subsequent meteorological fusion more consistent with the actual operating patterns of the target power region. Finally, second meteorological data for future time periods are fused based on the target meteorological weights, and the target power parameters are predicted based on the fusion results. Thus, without increasing the input dimensions of the prediction model, the amount of meteorological data processing and computational resource consumption can be reduced, improving the processing efficiency of target power parameter prediction.

[0092] In one embodiment, the method further includes: obtaining reference power parameters; predicting the predicted power parameters of the target power area within a historical time period based on the target meteorological characteristics, including: generating meteorological correlation features based on the target meteorological characteristics and the reference power parameters; the meteorological correlation features include one or more of time-series correlation features, scenario correlation features, regional correlation features, and power grid operation correlation features; and predicting the predicted power parameters of the target power area within a historical time period based on the meteorological correlation features.

[0093] Reference power parameters refer to other power parameters, besides those to be predicted, that can reflect the operating status of the target power area or affect changes in the power parameters to be predicted. Reference power parameters may include one or more of the following: load data, renewable energy output data, tie-line plans, transmission capacity, generator maintenance status, and grid operating status.

[0094] When predicting the power parameters of a target power area within a historical time period based on the target meteorological characteristics, the target meteorological characteristics and reference power parameters can be further combined to generate meteorological correlation features that can reflect the relationship between meteorological changes and power operation status, so as to improve the ability of the prediction model to represent different operating scenarios.

[0095] Among them, the temporal correlation and features include the lagging features of the target meteorological features in the current time period and the previous time period, as well as the rate of change, rolling mean, rolling maximum value and duration features of the target meteorological features.

[0096] Among them, scenario-related features refer to the features used to characterize the current meteorological operating scenario of the target power area. Scenario-related features may include one or more of the following: high temperature and high load scenario, cold wave load change scenario, midday high output of new energy scenario, high output of wind power with low parameter value scenario, line congestion scenario, parameter peak scenario, and negative parameter value scenario.

[0097] Among them, grid operation correlation characteristics refer to features used to reflect the interaction between target meteorological characteristics and grid operation parameters. For example, temperature characteristics can be combined with predicted load to obtain temperature-load correlation characteristics; irradiance characteristics can be combined with photovoltaic predicted output to obtain irradiance-photovoltaic output correlation characteristics; wind speed characteristics can be combined with wind power predicted output to obtain wind speed-wind power output correlation characteristics; and rainstorm, strong wind, or freezing indicators can be combined with transmission capacity, tie line plans, or unit maintenance status to obtain transmission constraint correlation characteristics.

[0098] Among them, grid operation correlation characteristics refer to features constructed based on the correspondence between target meteorological characteristics and reference power parameters, used to reflect the combined effect of meteorological changes and grid operation status. Optionally, the target meteorological characteristics can be multiplied, subtracted, or concatenated with the corresponding reference power parameters to obtain grid operation correlation characteristics. For example, multiplying the fused temperature characteristics with load data yields temperature-load correlation characteristics; multiplying the fused irradiance characteristics with photovoltaic output data yields irradiance-photovoltaic correlation characteristics; multiplying the fused wind speed characteristics with wind power output data yields wind speed-wind power correlation characteristics; and multiplying rainstorm, strong wind, or freezing indicators with transmission capacity or tie-line plans yields meteorological transmission correlation characteristics.

[0099] Among them, regional correlation characteristics refer to features used to characterize the degree of correlation between each reference power region and the target power region. Optionally, the regional correlation value corresponding to each reference power region can be determined based on the historical correlation of power parameters, electrical distance, tie-line transmission relationship, renewable energy transmission volume or load contribution between each reference power region and the target power region. The regional correlation value is then combined with the initial meteorological characteristics of the corresponding reference power region to obtain the regional correlation characteristics. For example, the historical correlation coefficient between a reference power region and the target power region can be multiplied by the temperature, wind speed or irradiance of the reference power region to obtain the regional correlation characteristics corresponding to that reference power region.

[0100] In the above embodiments, by constructing time-series correlation features, scene correlation features, regional correlation features, and power grid operation correlation features, the impact of meteorological changes on target power parameters at different times, in different regions, and under different power grid operation states can be further characterized, avoiding the direct input of target meteorological features into the prediction model, thereby improving the accuracy of the prediction.

[0101] In one embodiment, the above-mentioned allocation of initial meteorological weights to each reference power area and target power area includes at least one of the following: determining the initial meteorological weights corresponding to each reference power area and target power area based on the number of reference power areas and target power areas; calculating the historical correlation score between the initial meteorological characteristics of each reference power area and the historical power parameters of the target power area; determining the initial meteorological weights corresponding to each reference power area and target power area based on the historical correlation score; calculating the historical frequency of abnormal power parameters occurring in the target power area when extreme weather occurs in the reference power area and target power area; determining the initial meteorological weights corresponding to each reference power area and target power area based on the historical frequency; and randomly generating the initial meteorological weights corresponding to each reference power area and target power area.

[0102] Optionally, the initial meteorological weights can be determined based on the total number of target power areas and reference power areas. For example, when there are four power areas in total, the initial meteorological weights for each power area can be set to 0.25 to generate weighted coding information corresponding to uniform weights.

[0103] Optionally, the historical correlation score between the initial meteorological characteristics of each power region and the historical power parameters of the target power region can be calculated. Specifically, the correlation coefficients between the initial meteorological characteristics such as temperature, humidity, wind speed, and irradiance of each power region and the historical power parameters can be calculated separately. Based on the absolute value, average value, or weighted sum of each correlation coefficient, the historical correlation score of the corresponding power region can be determined. Then, the historical correlation scores of each power region are processed to achieve non-negativity and normalization to obtain the corresponding initial meteorological weights. It should be noted that since the sum of the weights of the target power region and each reference power region is 1, the initial meteorological weight of the target power region can be determined after the weights of each power region have been determined.

[0104] Optionally, the historical frequency of abnormal power parameters in the target power region during extreme weather events in each power region can be statistically analyzed, and the historical frequencies for each power region can be normalized to obtain the corresponding initial meteorological weights. A higher historical frequency indicates a stronger correlation between the extreme weather in that power region and the abnormal power parameters in the target power region, and the corresponding initial meteorological weight can be larger.

[0105] Optionally, non-negative values ​​corresponding to each power zone can be randomly generated, and the randomly generated values ​​can be normalized to obtain the randomly generated initial meteorological weights.

[0106] Furthermore, in one embodiment, after allocating initial meteorological weights to each reference power region and the target power region respectively, the method further includes: generating weight coding information based on the initial meteorological weights, and generating an initial population based on the weight coding information.

[0107] After determining the initial meteorological weights for each power region, these weights can be sequentially encoded according to a pre-defined arrangement, resulting in a weight encoding information, or a chromosome. Each chromosome corresponds to a complete set of meteorological weight combinations. Based on multiple sets of initial meteorological weights obtained through different methods, multiple chromosomes can be generated, and these chromosomes can form an initial population.

[0108] For example, assuming the total number of target power regions and reference power regions is three, the different types of chromosomes can be represented as follows:

[0109] Type I chromosomes: 0.33, 0.33, 0.34;

[0110] Type II chromosomes: 0.20, 0.50, 0.30;

[0111] Type III chromosomes: 0.10, 0.70, 0.20;

[0112] Type IV chromosomes: 0.45, 0.15, 0.40.

[0113] The first type of chromosome is a uniformly weighted chromosome, and typically only one needs to be generated to serve as the baseline weighting scheme. The second and third types of chromosomes can be generated in multiples based on different historical time windows, different combinations of meteorological characteristics, or different extreme weather types. For example, correlation scores can be calculated based on historical data from the past month, three months, and one year, thereby generating multiple second-type chromosomes. The fourth type of chromosome can be randomly generated in multiples to supplement the preset population size.

[0114] Optionally, the initial population size can be preset, and the number of chromosomes of each type can be configured according to a preset ratio. For example, when the initial population size is 64, it can include one type I chromosome, several type II chromosomes, several type III chromosomes, and the remaining number of type IV chromosomes.

[0115] For example, each set of meteorological weight vectors (W) is encoded as a chromosome. A chromosome is represented as: Here, gi represents the gene value corresponding to the i-th region. To satisfy the non-negativity constraint and the constraint that the weight sum is 1, the gene values ​​can be converted into meteorological weights using normalization: .in, When all gene values ​​are 0, the chromosome can be reinitialized, or the weights of each region can be set to uniform weights.

[0116] In the above embodiments, by setting four types of chromosomes simultaneously in the initial population, the initial weights can be avoided from being too singular. The initial population can simultaneously include the baseline weight scheme, the weight scheme that conforms to historical patterns, and the random weight scheme, thereby expanding the search range of meteorological weights and increasing the possibility of obtaining better meteorological weights.

[0117] In one embodiment, the above-mentioned adjustment of the initial meteorological weights based on the cost function value, and the adjustment of the initial meteorological weights, includes: selecting parent weight encoding information from the initial population based on the cost function value corresponding to each weight encoding information; performing crossover processing on the parent weight encoding information according to a preset crossover operation to obtain offspring weight encoding information; performing mutation processing on the offspring weight encoding information according to a preset mutation operation to obtain an updated population; the updated population includes multiple updated weight encoding information; each updated weight encoding information represents the updated initial meteorological weights.

[0118] In this embodiment, the initial meteorological weights are adjusted by selecting, crossing over, and mutating chromosomes. Optionally, firstly, based on the cost function values ​​corresponding to each chromosome, the chromosome with the smaller cost function value is selected from the initial population as the parent chromosome; then, some weights in different parent chromosomes are exchanged to obtain offspring chromosomes; next, at least one weight in the offspring chromosomes is adjusted according to a preset mutation probability, and the adjusted weights are normalized to obtain an updated population. Each chromosome in the updated population represents a set of updated initial meteorological weights.

[0119] It should be noted that in each iteration, the first meteorological data is fused based on the initial meteorological weights represented by each chromosome, resulting in multiple sets of target meteorological features corresponding one-to-one with each chromosome. Then, each set of target meteorological features is input into the power parameter prediction model to obtain multiple sets of predicted power parameters, and the corresponding cost function value is calculated based on each set of predicted power parameters and historical power parameters. Selection, crossover, and mutation are continued based on the recalculated cost function values ​​until preset conditions are met, and the meteorological weights represented by chromosomes whose cost function values ​​meet the preset conditions are determined as the target meteorological weights.

[0120] In this way, while retaining the better combination of meteorological weights, new combinations of meteorological weights can be generated through cross-pollination and variation, thereby gradually reducing the difference between the forecast results and historical power parameters and improving the accuracy of the target meteorological weights.

[0121] Furthermore, in one embodiment, the above-mentioned cross-processing of the parent generation weight coding information according to the preset cross-operation to obtain the offspring weight coding information includes: determining the gene cross-probability corresponding to each reference power region based on the regional attributes of each reference power region and the stability of the meteorological weight corresponding to each reference power region in multiple historical high-quality weight coding information; for each reference power region, cross-processing the genes corresponding to the reference power region in at least two parent generation weight coding information according to the corresponding gene cross-probability to obtain the offspring genes corresponding to the reference power region in the offspring weight coding information; and generating the offspring weight coding information based on the offspring genes corresponding to each reference power region.

[0122] The regional attributes of the reference power area characterize the relationship between the reference power area and the target power area, and may include one or more of the following: electrical distance, tie-line relationship, renewable energy transmission relationship, and load contribution relationship. The stability of the meteorological weights characterizes the degree of change of the meteorological weights corresponding to the same reference power area across multiple historical high-quality weight coding information sets. For example, the stability can be determined based on the variance or range of the corresponding meteorological weights in multiple historical high-quality weight coding information sets; the smaller the variance or range, the higher the stability.

[0123] Optionally, the basic crossover probability can be determined based on regional attributes, and adjusted according to the stability level. For example, a reference power region with a high degree of correlation with the target power region and a high degree of stability in meteorological weights can be given a lower gene crossover probability to retain its superior weights; a reference power region with a low degree of correlation or a large change in meteorological weights can be given a higher gene crossover probability to expand the weight search range.

[0124] For example, suppose the weight encoding information of the two parent generations is "0.2, 0.5, 0.3" and "0.4, 0.1, 0.5" respectively, and the gene crossover probability corresponding to the second reference power region is 0.8. When the generated random number is less than 0.8, the genes corresponding to the second reference power region in the two parents can be swapped, or the offspring genes can be obtained by weighting the two gene values. Then, the offspring genes corresponding to each reference power region are arranged in a preset order and normalized to obtain the offspring weight encoding information.

[0125] For example, the selected parent chromosomes are crossed to generate offspring chromosomes.

[0126] Suppose the two paternal chromosomes are:

[0127]

[0128]

[0129] Offspring can be generated using arithmetic crossover:

[0130]

[0131] in, The result is a random number between 0 and 1. After crossover, the offspring chromosomes are normalized so that their corresponding weights satisfy the following: This method allows for continuous searching among optimal parent weight combinations, making it suitable for continuous variable optimization problems such as meteorological weights.

[0132] In the above embodiments, by setting different gene crossover probabilities for different reference power regions, it is possible to retain stable and effective meteorological weights while focusing on adjusting unstable meteorological weights, thereby improving the targeting of weight search.

[0133] Furthermore, in one embodiment, the above-mentioned mutation processing of the offspring weight coding information according to the preset mutation operation to obtain the updated population includes: determining the mutation probability and mutation amplitude of the corresponding gene of each reference power region based on at least one of the historical frequency of extreme weather events in each reference power region, the contribution degree of abnormal power parameters corresponding to each reference power region, the prediction error corresponding to each reference power region, and the stability of the corresponding gene in high-quality weight coding information over multiple consecutive generations; determining the gene to be mutated from the offspring weight coding information based on the mutation probability of the gene corresponding to each reference power region; randomly perturbing the gene to be mutated based on the mutation amplitude corresponding to the gene to be mutated to obtain the mutated offspring weight coding information; and performing non-negative processing and normalization processing on the gene values ​​in the mutated offspring weight coding information to obtain the updated population.

[0134] Among them, the historical frequency of extreme weather refers to the number of times extreme weather events such as high temperature, cold wave, strong wind or rainstorm have occurred in the reference power area within a historical period; the contribution of abnormal power parameters refers to the degree of influence of extreme weather events in the reference power area on the occurrence of abnormal power parameters in the target power area; the prediction error refers to the difference between the predicted power parameters and the historical power parameters when using meteorological data of the corresponding reference power area for prediction; and the stability refers to the degree of change of the same gene in the high-quality weight coding information of consecutive generations. The smaller the change, the more stable the gene is.

[0135] Optionally, the number of times extreme weather occurred in the reference power area and abnormal power parameters simultaneously appeared in the target power area can be counted, and the ratio between this number and the total number of extreme weather events in the reference power area can be calculated to obtain the corresponding contribution of abnormal power parameters. The larger this ratio, the higher the correlation between the extreme weather in the reference power area and the abnormal power parameters in the target power area.

[0136] Optionally, when the prediction error of the reference power area exceeds a first preset threshold or the stability of the corresponding gene is lower than a second preset threshold, a first mutation probability and a first mutation amplitude can be set to expand the adjustment range of the meteorological weight. When the contribution of abnormal power parameters in the reference power area exceeds a third threshold and the corresponding gene is relatively stable, a second mutation probability and a second mutation amplitude can be set to reduce the disruption to the optimal meteorological weight. Specifically, the first mutation probability is greater than the second mutation probability, and the first mutation amplitude is greater than the second mutation amplitude.

[0137] Optionally, random numbers can be generated based on the mutation probability corresponding to each gene. When the random number satisfies the corresponding mutation condition, the gene is identified as the gene to be mutated, and a random number within the mutation range is added to or subtracted from its original gene value to obtain the mutated gene value. Then, each gene value is processed to be non-negative and normalized so that the sum of each meteorological weight is 1, thereby obtaining the mutated offspring weight coding information. Multiple mutated offspring weight coding information sets constitute the updated population.

[0138] In the above embodiments, by adaptively determining the mutation probability and mutation magnitude based on the historical meteorological conditions, prediction errors and gene stability of different reference power areas, the meteorological weights with poor or unstable prediction effects can be adjusted in a focused manner, while retaining the relatively stable meteorological weights, thereby improving the targeting of weight optimization.

[0139] In one embodiment, the predicted power parameters include predicted power parameters for a preset scenario; calculating the cost function value by comparing the predicted power parameters with historical power parameters includes: calculating the mean absolute error between the predicted power parameters and historical power parameters; calculating the preset scenario error value between the predicted power parameters for the preset scenario and historical power parameters; and determining the cost function value based on the mean absolute error and the preset scenario error value.

[0140] The mean absolute error (MAE) reflects the overall difference between predicted and historical power parameters across all historical samples. The preset scenario error value reflects the model's prediction differences under preset scenarios such as parameter spikes, negative parameter values, extreme weather, or line congestion. For example, historical moments belonging to preset scenarios can be selected from the historical time period, and the MAE between the predicted and historical power parameters corresponding to these moments can be calculated to obtain the preset scenario error value.

[0141] Then, the mean absolute error and the preset scene error value can be weighted and summed to obtain the cost function value. For example, a first weight can be set for the mean absolute error, a second weight can be set for the preset scene error value, and the weighted sum of the two can be used as the cost function value.

[0142] For example, it can be calculated according to the following formula:

[0143]

[0144] In the above embodiments, by increasing the influence of the preset scenario error in the cost function while evaluating the overall prediction effect, it is possible to avoid only pursuing a small overall error and ignoring special scenarios, so that the optimized meteorological weights have good prediction effects in both ordinary scenarios and preset scenarios.

[0145] Furthermore, in one embodiment, the aforementioned preset scenarios include parameter peak scenarios and line congestion scenarios; a first sub-error value is calculated between the parameter peak scenario and historical power parameters; a second sub-error value is calculated between the line congestion scenario and historical power parameters; a third sub-error value is calculated between the parameter anomaly scenario and historical power parameters; based on the first sub-error value, the second sub-error value, the third sub-error value, and the constraint penalty value, a preset scenario error value is obtained; the constraint penalty value is determined according to the current initial meteorological weight and the initial meteorological weight in the previous iteration process.

[0146] In this embodiment, prediction errors are calculated for parameter peak scenarios, line congestion scenarios, and parameter anomaly scenarios, respectively, and the variation of meteorological weights in adjacent iterations is limited by constraint penalty values.

[0147] The first sub-error value is used to characterize the prediction error under parameter peak scenarios. Specifically, it can be obtained by selecting historical times when the historical power parameters are greater than a preset peak threshold from the historical time period, and calculating the average absolute error between the predicted power parameters and the historical power parameters corresponding to these historical times.

[0148] The second sub-error value is used to characterize the prediction error under line congestion scenarios. Specifically, the historical time when line congestion occurs can be determined based on the line load rate, transmission power, or available transmission capacity, and the average absolute error between the predicted power parameters and the historical power parameters corresponding to these historical times can be calculated to obtain the second sub-error value.

[0149] The third sub-error value is used to characterize the prediction error under abnormal parameter scenarios. Specifically, the historical time when the parameter abnormality occurred can be determined based on whether the historical power parameters exceeded the preset parameter range or whether there was a sudden increase or decrease. The average absolute error between the predicted power parameters and the historical power parameters corresponding to these historical times is then calculated to obtain the third sub-error value.

[0150] The constraint penalty value is used to characterize the degree of change in meteorological weights between the current iteration and the previous iteration. Optionally, the difference between the initial meteorological weights of each power region in the current iteration and the initial meteorological weights in the previous iteration can be calculated separately, and the absolute value or square value of each difference can be summed to obtain the constraint penalty value. The greater the change in weights, the greater the constraint penalty value.

[0151] Then, the first sub-error value, the second sub-error value, the third sub-error value, and the constraint penalty value can be weighted and summed to obtain the preset scenario error value.

[0152] The calculation process for the corresponding cost function value is as follows:

[0153]

[0154] Where MAEall represents the mean absolute error across all validation samples, MAEspike represents the mean absolute error of the sample data corresponding to peak scenarios, MAEnegative represents the mean absolute error of the sample data corresponding to anomalous scenarios, MAEcongestion represents the mean absolute error of the sample data corresponding to line congestion scenarios, and Penalty represents the penalty value for weight constraints. , , , , These are preset or dynamically adjusted weighting coefficients.

[0155] In the above embodiments, by calculating the prediction errors under parameter peak scenarios, line congestion scenarios, and parameter anomaly scenarios separately, the cost function can simultaneously reflect the prediction performance under different special operating scenarios, avoiding the optimization process focusing only on the overall prediction error. Simultaneously, by setting constraint penalty values ​​to limit the variation of meteorological weights in adjacent iterations, drastic fluctuations in weights can be reduced, thereby improving the stability of the target meteorological weights and the accuracy of power parameter prediction under special scenarios.

[0156] In one embodiment, after predicting the target power parameters of the target power region in a future time period, the method further includes: obtaining the extreme weather identifier, the power grid operation risk identifier, and the target weather weight corresponding to each reference power region in the future time period; determining that the future time period meets the preset correction condition when the extreme weather identifier indicates that at least one reference power region has experienced extreme weather, and the target weather weight corresponding to the reference power region experiencing extreme weather is greater than a preset weight threshold; when the future time period meets the preset correction condition, inputting the target power parameters, target weather characteristics, extreme weather identifier, and power grid operation risk identifier into a pre-trained residual correction model to obtain the prediction residual corresponding to the target power parameters; and correcting the target power parameters based on the prediction residual to obtain the corrected target power parameters.

[0157] The extreme weather indicator is used to indicate whether extreme weather events such as high temperatures, cold waves, strong winds, heavy rain, or freezing are likely to occur in various reference power areas within a future time period. The power grid operation risk indicator is used to indicate whether operational risks such as line congestion, sudden changes in renewable energy output, abnormal load, or a decrease in transmission capacity may occur within a future time period.

[0158] When extreme weather occurs in at least one reference power region, and the target weather weight corresponding to that reference power region is greater than the preset weight threshold, it indicates that the region has a significant impact on the target power parameters, and the prediction results of conventional prediction models may have a large deviation. Therefore, it is determined that the future time period meets the preset correction conditions.

[0159] When the preset correction conditions are met, the target power parameters, target meteorological characteristics, extreme weather indicators, and power grid operation risk indicators obtained from the preliminary prediction are input into the residual correction model. The residual correction model then predicts the possible deviations between the preliminary prediction results and the actual power parameters, obtaining the prediction residuals. Subsequently, the prediction residuals are added to or subtracted from the target power parameters to obtain the corrected target power parameters.

[0160] In the above embodiments, residual correction is initiated only when extreme weather occurs in important reference power areas. This can compensate for the prediction bias of conventional prediction models under extreme weather and high-risk operating scenarios, thereby improving the prediction accuracy of target power parameters in special scenarios, while avoiding unnecessary corrections to the prediction results in ordinary scenarios.

[0161] For example, in combination Figure 3 As shown, Figure 3 This is a flowchart of the steps in a power parameter prediction method in one embodiment.

[0162] Step 302: Obtain the first meteorological data of the target power area and each reference power area within the historical time period, and obtain the historical power parameters and reference power parameters corresponding to the target power area.

[0163] Step 304: Perform time alignment, missing value processing, outlier processing, and standardization on the first meteorological data, historical power parameters, and reference power parameters to obtain the processed historical data.

[0164] Step 306: Assign multiple sets of initial meteorological weights to the target power area and each reference power area, and generate weight coding information according to each set of initial meteorological weights. The initial population is composed of multiple weight coding information.

[0165] Step 308: For each weighted encoding information in the initial population, the first meteorological data of the target power area and each reference power area are weighted and fused according to the initial meteorological weight represented by the weighted encoding information to obtain the corresponding target meteorological features.

[0166] Step 310: Input the meteorological characteristics of each target and the corresponding reference power parameters into the pre-trained power parameter prediction model to obtain multiple sets of predicted power parameters for the target power area within the historical time period.

[0167] Step 312: Calculate the cost function value corresponding to each weighted encoding information based on the difference between the predicted power parameters and historical power parameters for each group.

[0168] Step 314: Based on the cost function value corresponding to each weight encoding information, select, cross over and mutate the weight encoding information in the initial population to obtain an updated population, and repeat steps 308 to 312 until the preset iteration termination condition is met.

[0169] Step 316: The meteorological weight represented by the weight encoding information where the cost function value satisfies the preset conditions when the preset iteration termination condition is met is determined as the target meteorological weight.

[0170] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0171] Based on the same inventive concept, this application also provides a power parameter prediction device for implementing the power parameter prediction method described above. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations in one or more power parameter prediction device embodiments provided below can be found in the limitations of the power parameter prediction method described above, and will not be repeated here.

[0172] In one exemplary embodiment, such as Figure 4 As shown, a power parameter prediction device is provided, including: module M, module N, and module L, wherein:

[0173] The acquisition module 100 is used to acquire historical power parameters and first meteorological data of the target power area and the reference power area within a historical time period; the reference power area is a power area related to the target power area; the first meteorological data includes multi-dimensional initial meteorological characteristics.

[0174] The feature fusion module 200 is used to assign initial meteorological weights to each reference power area and the target power area, and to fuse each initial meteorological feature according to the initial meteorological weights to obtain the target meteorological feature.

[0175] The cost calculation module 300 is used to predict the predicted power parameters of the target power area within a historical time period based on the target meteorological characteristics, and calculate the cost function value by combining the predicted power parameters with the historical power parameters.

[0176] The iteration module 400 is used to adjust the initial meteorological weights according to the cost function value when the cost function value does not meet the preset conditions, and to re-predict the predicted power parameters of the target power area within the historical time period according to the adjusted initial meteorological weights, until the cost function value meets the preset conditions and the target meteorological weights are obtained.

[0177] The prediction module 500 is used to acquire second meteorological data for a future time period and predict the target power parameters corresponding to the target power area in the future time period based on the target meteorological weight and the second meteorological data.

[0178] In one embodiment, the above-described apparatus further includes a parameter acquisition module, which is used to acquire reference power parameters.

[0179] In one embodiment, the prediction module 500 includes:

[0180] The feature generation unit is used to generate meteorological correlation features based on the target meteorological features and reference power parameters. The meteorological correlation features include one or more of the following: time-series correlation features, scene correlation features, regional correlation features, and power grid operation correlation features.

[0181] The parameter prediction unit is used to predict the power parameters of the target power area within a historical time period based on meteorological correlation characteristics.

[0182] In one embodiment, the feature fusion module 200 includes:

[0183] The first weighting unit is used to determine the initial meteorological weights corresponding to each reference power area and target power area based on the number of reference power areas and target power areas.

[0184] The second weighting unit is used to calculate the historical correlation score between the initial meteorological characteristics of each reference power region and the historical power parameters of the target power region; based on the historical correlation score, the initial meteorological weights corresponding to each reference power region and the target power region are determined.

[0185] The third weighting unit is used to calculate the historical frequency of abnormal power parameters in the target power area when extreme weather occurs in the reference power area and the target power area; based on the historical frequency, the initial meteorological weights corresponding to each reference power area and the target power area are determined.

[0186] The fourth weighting unit is used to randomly generate the initial meteorological weights corresponding to each reference power region and the target power region.

[0187] In one embodiment, the above-mentioned apparatus includes:

[0188] The coding module is used to generate weighted coding information based on the initial meteorological weights, and to generate the initial population based on the weighted coding information.

[0189] In one embodiment, the iteration module 400 includes:

[0190] The first selection unit is used to select parent weight encoding information from the initial population based on the cost function value corresponding to each weight encoding information.

[0191] The crossover unit is used to perform crossover processing on the parent generation weight encoding information according to the preset crossover operation to obtain the child generation weight encoding information.

[0192] The mutation unit is used to mutate the offspring weight coding information according to the preset mutation operation to obtain the updated population; the updated population includes multiple updated weight coding information; each updated weight coding information represents the updated initial meteorological weight.

[0193] In one embodiment, the aforementioned cross unit includes:

[0194] The first probability calculation unit is used to determine the gene crossover probability of each reference power region based on the regional attributes of each reference power region and the stability of the meteorological weights corresponding to each reference power region in multiple historical high-quality weight coding information.

[0195] The gene crossover unit is used to crossover genes corresponding to the reference power regions in at least two parent weight encoding information according to the corresponding gene crossover probability for each reference power region, so as to obtain the offspring genes corresponding to the reference power regions in the offspring weight encoding information.

[0196] The update unit is used to generate offspring weight coding information based on the offspring genes corresponding to each reference power region.

[0197] In one embodiment, the above-mentioned variation unit includes:

[0198] The second probability calculation unit is used to determine the mutation probability and mutation magnitude of the corresponding gene in each reference power region based on at least one of the following: the historical frequency of extreme weather events in each reference power region, the contribution of abnormal power parameters corresponding to each reference power region, the prediction error corresponding to each reference power region, and the stability of the corresponding gene in high-quality weighted coding information over multiple generations.

[0199] The second selection unit is used to determine the gene to be mutated from the offspring weight encoding information based on the mutation probability of the gene corresponding to each reference power region.

[0200] The gene mutation unit is used to randomly perturb the gene to be mutated according to the mutation amplitude of the gene to be mutated, so as to obtain the weighted encoding information of the offspring after mutation.

[0201] The information processing unit is used to perform non-negative processing and normalization on the gene values ​​in the weighted encoding information of the mutated offspring to obtain the updated population.

[0202] In one embodiment, the predicted power parameters include predicted power parameters for a preset scenario, and the cost calculation module 300 includes:

[0203] The average error calculation unit is used to calculate the average absolute error between predicted power parameters and historical power parameters.

[0204] The scenario error calculation unit is used to calculate the preset scenario error value between the predicted power parameters and the historical power parameters of the preset scenario.

[0205] The cost value calculation unit is used to calculate and determine the cost function value based on the mean absolute error and the preset scenario error value.

[0206] In one embodiment, the preset scenarios include parameter spike scenarios, parameter anomaly scenarios, and line congestion scenarios, and the scenario error calculation unit includes:

[0207] The first sub-error calculation unit is used to calculate the first sub-error value between the parameter peak scenario and the historical power parameters.

[0208] The second sub-error calculation unit is used to calculate the second sub-error value between the line congestion scenario and historical power parameters.

[0209] The third sub-error calculation unit is used to calculate the third sub-error value between abnormal parameter scenarios and historical power parameters.

[0210] The scene error value calculation unit is used to obtain the preset scene error value based on the first sub-error value, the second sub-error value, the third sub-error value, and the constraint penalty value; the constraint penalty value is determined based on the current initial meteorological weight and the initial meteorological weight in the previous iteration process.

[0211] In one embodiment, the above-described apparatus further includes a post-processing module, which includes:

[0212] The identifier acquisition unit is used to acquire extreme weather identifiers, power grid operation risk identifiers, and target weather weights for each reference power area for future time periods.

[0213] The condition judgment unit is used to determine whether the future time period meets the preset correction conditions when the extreme weather indicator represents at least one reference power area where extreme weather occurs, and the target weather weight corresponding to the reference power area where extreme weather occurs is greater than the preset weight threshold.

[0214] The residual calculation unit is used to input the target power parameters, target meteorological characteristics, extreme weather indicators, and power grid operation risk indicators into a pre-trained residual correction model to obtain the predicted residuals corresponding to the target power parameters, provided that the preset correction conditions are met in the future time period.

[0215] The correction unit is used to correct the target power parameters based on the prediction residuals to obtain the corrected target power parameters.

[0216] Each module in the aforementioned power parameter prediction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0217] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores historical power parameter data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When executed by the processor, the computer program implements a power parameter prediction method.

[0218] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0219] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method in any of the above embodiments.

[0220] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the method in any of the above embodiments.

[0221] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the method in any of the above embodiments.

[0222] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0223] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0224] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for predicting power parameters, characterized in that, The method includes: Historical power parameters and first meteorological data of the target power area and a reference power area within a historical time period are obtained; the reference power area is a power area related to the target power area; the first meteorological data includes multidimensional initial meteorological characteristics. Initial meteorological weights are assigned to each of the reference power regions and the target power regions, and each of the initial meteorological features is fused according to the initial meteorological weights to obtain the target meteorological features; Based on the target meteorological characteristics, predict the predicted power parameters of the target power area within a historical time period, and calculate the cost function value by combining the predicted power parameters with the historical power parameters. When the cost function value does not meet the preset conditions, the initial meteorological weight is adjusted according to the cost function value, and the predicted power parameters of the target power area in the historical time period are re-predicted according to the adjusted initial meteorological weight, until the cost function value meets the preset conditions, and the target meteorological weight is obtained. Acquire second meteorological data for a future time period, and predict the target power parameters corresponding to the target power area for the future time period based on the target meteorological weight and the second meteorological data.

2. The method according to claim 1, characterized in that, The method further includes: Obtain reference power parameters; The prediction of predicted power parameters for the target power area within a historical time period based on the target meteorological characteristics includes: Based on the target meteorological characteristics and the reference power parameters, meteorological correlation characteristics are generated; the meteorological correlation characteristics include one or more of the following: time-series correlation characteristics, scenario correlation characteristics, regional correlation characteristics, and power grid operation correlation characteristics. Based on the meteorological correlation characteristics, predict the power parameters of the target power area within the historical time period.

3. The method according to claim 1, characterized in that, The assignment of initial meteorological weights to each of the reference power areas and the target power areas includes at least one of the following: Based on the number of reference power areas and target power areas, the initial meteorological weights corresponding to each reference power area and target power area are determined; Calculate the historical correlation score between the initial meteorological characteristics of each reference power region and the historical power parameters of the target power region; based on the historical correlation score, determine the initial meteorological weights corresponding to each reference power region and the target power region; Calculate the historical frequency of abnormal power parameters in the target power region when extreme weather occurs in the reference power region and the target power region. Based on the historical frequency, the initial meteorological weights corresponding to each reference power area and the target power area are determined; The initial meteorological weights corresponding to each of the reference power regions and the target power regions are randomly generated.

4. The method according to claim 1, characterized in that, After assigning initial meteorological weights to each of the reference power regions and the target power regions, the method further includes: Based on the initial meteorological weights, weight coding information is generated, and based on the weight coding information, an initial population is generated.

5. The method according to claim 4, characterized in that, The step of adjusting the initial meteorological weights according to the cost function value, and adjusting the initial meteorological weights according to the adjusted initial meteorological weights, includes: Based on the cost function value corresponding to each weight encoding information, parent weight encoding information is selected from the initial population; The parent generation weight encoding information is cross-processed according to a preset cross-operation to obtain the child generation weight encoding information; The offspring weight encoding information is mutated according to a preset mutation operation to obtain an updated population; the updated population includes multiple updated weight encoding information; each updated weight encoding information represents the updated initial meteorological weight.

6. The method according to claim 5, characterized in that, The step of performing cross-processing on the parent generation weight encoding information according to a preset cross-processing operation to obtain the child generation weight encoding information includes: Based on the regional attributes of each reference power region and the stability of the meteorological weights corresponding to each reference power region in multiple historical high-quality weight coding information, the gene crossover probability corresponding to each reference power region is determined. For each of the reference power regions, based on the corresponding gene crossover probability, crossover is performed on the genes corresponding to the reference power regions in at least two of the parent weight coding information to obtain the offspring genes corresponding to the reference power regions in the offspring weight coding information. The offspring weight encoding information is generated based on the offspring genes corresponding to each of the reference power regions.

7. The method according to claim 5, characterized in that, The step of performing mutation processing on the offspring weight encoding information according to a preset mutation operation to obtain an updated population includes: Based on at least one of the following: the historical frequency of extreme weather events in each reference power region, the contribution of abnormal power parameters corresponding to each reference power region, the prediction error corresponding to each reference power region, and the stability of the corresponding gene in high-quality weighted coding information over multiple generations, the mutation probability and mutation magnitude of the corresponding gene in each reference power region are determined. Based on the mutation probability of the corresponding gene in each of the reference power regions, the gene to be mutated is determined from the offspring weight encoding information; Based on the mutation amplitude corresponding to the gene to be mutated, the gene to be mutated is randomly perturbed to obtain the mutated offspring weight encoding information. The gene values ​​in the mutated offspring weight encoding information are subjected to non-negative processing and normalization to obtain the updated population.

8. The method according to claim 1, characterized in that, The predicted power parameters include the predicted power parameters for a preset scenario; The step of calculating the cost function value by combining the predicted power parameters with the historical power parameters includes: Calculate the mean absolute error between the predicted power parameters and the historical power parameters; Calculate the preset scenario error value between the predicted power parameters of the preset scenario and the historical power parameters; The cost function value is determined by calculating based on the mean absolute error and the preset scenario error value.

9. The method according to claim 8, characterized in that, The preset scenarios include parameter peak scenarios, parameter anomaly scenarios, and line congestion scenarios; the calculation of the preset scenario error value between the predicted power parameters of the preset scenarios and the historical power parameters includes: Calculate the first sub-error value between the parameter peak scenario and the historical power parameters; Calculate the second sub-error value between the line congestion scenario and the historical power parameters; Calculate the third sub-error value between the abnormal parameter scenario and the historical power parameters; The preset scenario error value is obtained based on the first sub-error value, the second sub-error value, the third sub-error value, and the constraint penalty value; the constraint penalty value is determined based on the current initial meteorological weight and the initial meteorological weight in the previous iteration process.

10. The method according to claim 1, characterized in that, After predicting the target power parameters for the target power region in a future time period, the method further includes: Obtain the extreme weather indicators, power grid operation risk indicators, and target weather weights for each of the reference power areas for future time periods; If the extreme weather indicator represents that at least one of the reference power areas has experienced extreme weather, and the target weather weight corresponding to the reference power area where the extreme weather occurred is greater than a preset weight threshold, then the future time period is determined to meet the preset correction condition. If the preset correction conditions are met in the future time period, the target power parameters, the target meteorological characteristics, the extreme weather indicators, and the power grid operation risk indicators are input into a pre-trained residual correction model to obtain the prediction residuals corresponding to the target power parameters. The target power parameters are corrected based on the predicted residuals to obtain the corrected target power parameters.

11. A power parameter prediction device, characterized in that, The device includes: The acquisition module is used to acquire historical power parameters and first meteorological data of the target power area and the reference power area within a historical time period; the reference power area is a power area related to the target power area; the first meteorological data includes multidimensional initial meteorological features; The feature fusion module is used to assign initial meteorological weights to each of the reference power areas and the target power areas respectively, and to fuse each of the initial meteorological features according to the initial meteorological weights to obtain the target meteorological features; The cost calculation module is used to predict the predicted power parameters of the target power area within a historical time period based on the target meteorological characteristics, and calculate the cost function value by combining the predicted power parameters with the historical power parameters. An iterative module is used to adjust the initial meteorological weights according to the cost function values ​​when the cost function values ​​do not meet the preset conditions, and to re-predict the predicted power parameters of the target power area within the historical time period according to the adjusted initial meteorological weights, until the cost function values ​​meet the preset conditions, thereby obtaining the target meteorological weights. The prediction module is used to acquire second meteorological data for a future time period, and predict the target power parameters corresponding to the target power area in the future time period based on the target meteorological weight and the second meteorological data.

12. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 10.

13. A 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 steps of the method according to any one of claims 1 to 10.

14. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 10.

Images