Load prediction method and device, computer device and storage medium

Abstract

The application relates to a load prediction method and device, computer equipment and a storage medium, and relates to the technical field of intelligent power utilization. The load prediction method comprises the following steps: obtaining historical daily load data of each power distribution device in a power distribution network and historical daily load data of the power distribution network; at least two different generalized linear models are constructed; the historical daily load data of each power distribution device is used to initialize parameters in each generalized linear model, so that daily load data predicted by each generalized linear model is obtained; according to the historical daily load data of the power distribution network and predicted daily load data of the power distribution network, a target model is selected from each generalized linear model to predict future daily load data of the power distribution network according to current daily load data of each power distribution device in the power distribution network. The method can accurately predict future daily load data of the power distribution network by using the historical daily load data of the power distribution device, so that future load demand of the power distribution system is obtained, and the planning and operation of the power distribution network are effectively supported.

Description

Technical Field

[0001] This application relates to the field of smart electricity technology, and in particular to a load forecasting method, apparatus, computer equipment, and storage medium. Background Technology

[0002] Load forecasting is a fundamental task in ensuring the planning of power systems and is of great importance for their reliable and economical operation. Therefore, accurate load forecasting of power systems is necessary. Currently, related technologies typically utilize a large amount of external data, such as population density and humidity information, to build forecasting models to predict load.

[0003] However, due to the poor quality and low efficiency of external data acquisition, and the wide variety of factors affecting short-term load forecasting, the volatility and randomness are very high, making it difficult to achieve accurate forecasting, and improvements are urgently needed. Summary of the Invention

[0004] Therefore, it is necessary to provide a load forecasting method, apparatus, computer equipment, and storage medium that can achieve accurate load forecasting in response to the above-mentioned technical problems.

[0005] Firstly, this application provides a load forecasting method. The method includes:

[0006] Obtain historical daily load data for each power distribution device in the power distribution network and historical daily load data for the power distribution network itself;

[0007] Construct at least two different generalized linear models; wherein the different generalized linear models include the daily load parameters of different power distribution equipment;

[0008] Historical daily load data of each power distribution device is used to initialize the daily load parameters in each generalized linear model, thereby obtaining the predicted daily load data of the power distribution network predicted by each generalized linear model.

[0009] Based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by various generalized linear models, a target model is selected from the generalized linear models. The target model is used to predict the future daily load data of the distribution network based on the current daily load data of each distribution device in the distribution network.

[0010] In one embodiment, based on historical daily load data of the distribution network and predicted daily load data of the distribution network predicted by various generalized linear models, a target model is selected from the generalized linear models, including:

[0011] For each generalized linear model, the model performance index value of the generalized linear model is determined based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the generalized linear model, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network.

[0012] Based on the model performance index values ​​of each generalized linear model, the target model is selected from among the generalized linear models.

[0013] In one embodiment, the model performance index value of the generalized linear model is determined based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the generalized linear model, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network, including:

[0014] Based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by the generalized linear model, the sum of squared residuals is determined;

[0015] The model performance index value of the generalized linear model is determined based on the sum of squared residuals, the number of model parameters of the generalized linear model, and the total amount of historical daily load data of the distribution network.

[0016] In one embodiment, selecting a target model from the generalized linear models based on their model performance index values ​​includes:

[0017] Candidate models are selected from the generalized linear models based on their model performance index values.

[0018] Based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the candidate model, and the total number of historical daily load data of the distribution network, the error data of the candidate model is determined; the error data includes root mean square error and / or average error.

[0019] Based on the error data of the candidate models, the target model is selected from the candidate models.

[0020] In one embodiment, the historical daily load data of each power distribution device in the power distribution network is the historical daily maximum load data of each power distribution device, and the historical daily load data of the power distribution network is the historical daily maximum load data of the power distribution network.

[0021] Accordingly, historical daily load data of each power distribution device in the power distribution network and historical daily load data of the power distribution network are obtained, including:

[0022] Based on the historical daily raw load data of each power distribution device in the power distribution network, determine the historical daily maximum load data of each power distribution device;

[0023] Based on the same-day rate and the historical daily load data of each power distribution device, determine the historical daily maximum load data of the power distribution network.

[0024] Secondly, this application also provides a load forecasting device. The device includes:

[0025] The data acquisition module is used to acquire historical daily load data of each power distribution device in the power distribution network and historical daily load data of the power distribution network.

[0026] The model building module is used to build at least two different generalized linear models; wherein the different generalized linear models include the daily load parameters of different power distribution equipment;

[0027] The initialization module is used to initialize the daily load parameters in each generalized linear model using the historical daily load data of each power distribution equipment, so as to obtain the predicted daily load data of the power distribution network predicted by each generalized linear model.

[0028] The model selection module is used to select a target model from various generalized linear models based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by various generalized linear models. The target model is used to predict the future daily load data of the distribution network based on the current daily load data of each distribution device in the distribution network.

[0029] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0030] Obtain historical daily load data for each power distribution device in the power distribution network and historical daily load data for the power distribution network itself;

[0031] Construct at least two different generalized linear models; wherein the different generalized linear models include the daily load parameters of different power distribution equipment;

[0032] Historical daily load data of each power distribution device is used to initialize the daily load parameters in each generalized linear model, thereby obtaining the predicted daily load data of the power distribution network predicted by each generalized linear model.

[0033] Based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by various generalized linear models, a target model is selected from the generalized linear models. The target model is used to predict the future daily load data of the distribution network based on the current daily load data of each distribution device in the distribution network.

[0034] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0035] Obtain historical daily load data for each power distribution device in the power distribution network and historical daily load data for the power distribution network itself;

[0036] Construct at least two different generalized linear models; wherein the different generalized linear models include the daily load parameters of different power distribution equipment;

[0037] Historical daily load data of each power distribution device is used to initialize the daily load parameters in each generalized linear model, thereby obtaining the predicted daily load data of the power distribution network predicted by each generalized linear model.

[0038] Based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by various generalized linear models, a target model is selected from the generalized linear models. The target model is used to predict the future daily load data of the distribution network based on the current daily load data of each distribution device in the distribution network.

[0039] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0040] Obtain historical daily load data for each power distribution device in the power distribution network and historical daily load data for the power distribution network itself;

[0041] Construct at least two different generalized linear models; wherein the different generalized linear models include the daily load parameters of different power distribution equipment;

[0042] Historical daily load data of each power distribution device is used to initialize the daily load parameters in each generalized linear model, thereby obtaining the predicted daily load data of the power distribution network predicted by each generalized linear model.

[0043] Based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by various generalized linear models, a target model is selected from the generalized linear models. The target model is used to predict the future daily load data of the distribution network based on the current daily load data of each distribution device in the distribution network.

[0044] The aforementioned load forecasting methods, devices, computer equipment, storage media, and computer program products construct multiple generalized linear models based on historical daily load data of various power distribution equipment in the distribution network. Based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network from each generalized linear model, a target model is selected from these models to predict the future daily load data of the distribution network. This solution does not rely on external data such as population density and humidity; it can construct a model capable of accurately predicting the future daily load data of the distribution network using only the historical daily load data of the power distribution equipment. This provides an optional method for accurately predicting the future load of the distribution network and effectively supports the planning and operation of the distribution network. Attached Figure Description

[0045] Figure 1 This is a diagram illustrating the application environment of the load forecasting method in one embodiment;

[0046] Figure 2 This is a flowchart illustrating a load forecasting method in one embodiment;

[0047] Figure 3 This is a schematic diagram of a power distribution network in one embodiment;

[0048] Figure 4 This is a flowchart illustrating the process of selecting a target model in one embodiment;

[0049] Figure 5 This is a flowchart illustrating the process of selecting a target model in another embodiment;

[0050] Figure 6 This is a schematic diagram of the process for obtaining historical daily maximum load data in one embodiment;

[0051] Figure 7 This is a structural block diagram of a load forecasting device in one embodiment;

[0052] Figure 8 Here is a structural block diagram of the model selection module in one embodiment;

[0053] Figure 9 This is a structural block diagram of the data acquisition module in one embodiment;

[0054] Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0055] 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.

[0056] It should be understood that the term "comprising" as used in the specification and claims of this application indicates the presence of a described feature, integral, step, operation, element, and / or component, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof. It should also be understood that the terminology used in the specification is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in the specification and claims of this application, unless the context clearly indicates otherwise, the singular forms "a," "an," and "the" are intended to include the plural forms.

[0057] The load forecasting method provided in this application can be applied to, for example, Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. The data storage system can store the data that server 104 needs to process, such as historical daily load data of each power distribution device in the power distribution network and the historical daily load of the power distribution network. The data storage system can be integrated on server 104 or placed in the cloud or on other network servers. The load forecasting method provided in this application embodiment can be applied to server 104, terminal 102, or through interaction between terminal 102 and server 104. For example, server 104 can construct multiple generalized linear models based on the historical daily load data of each power distribution device in the power distribution network, and select a target model from the multiple generalized linear models based on the historical daily load data of the power distribution network and the predicted daily load data of the power distribution network predicted by each generalized linear model to predict the future daily load data of the power distribution network; furthermore, server 104 can send the predicted future daily load data of the power distribution network to terminal 102 for display. Terminal 102 can be, but is not limited to, various personal computers, laptops, etc. Server 104 can be implemented using a standalone server or a server cluster composed of multiple servers.

[0058] In one embodiment, such as Figure 2 The diagram illustrates a flowchart of a load forecasting method. This embodiment involves constructing a generalized linear model to predict future daily load data of the distribution network by acquiring historical daily load data of each distribution device and the historical daily load data of the distribution network itself. This embodiment applies this method to... Figure 1 Taking server 104 as an example, this embodiment illustrates the method, which includes the following steps:

[0059] S201, obtain the historical daily load data of each power distribution device in the power distribution network and the historical daily load data of the power distribution network.

[0060] The distribution network mentioned in S201 refers to any power grid with load forecasting demand. Optionally, the distribution network may have multiple power distribution devices for transmitting AC power, such as... Figure 3 The power distribution network shown has multiple distribution transformers.

[0061] Furthermore, the historical daily load data for each power distribution device refers to the daily load data of that device within a certain period prior to the current time (e.g., a month, a quarter, or a year), such as the maximum daily load data of that device within a certain period prior to the current time. Correspondingly, the historical daily load data of the power distribution network refers to the daily load data of the entire power distribution system within a certain period prior to the current time. It is understood that the historical daily load data of each power distribution device involved in this embodiment, as well as the historical daily load data of the power distribution network, are all data representing the actual load during the operation of the power distribution network.

[0062] It should be noted that since short-term load forecasting is generally conducted on an hourly or daily basis, the forecast results are affected by a wide variety of factors, exhibiting strong volatility and randomness. Therefore, in order to achieve accurate forecasting of future daily load data for the distribution network, this embodiment acquires a large amount of historical daily load data (such as historical daily load data within one year) to construct the model.

[0063] Specifically, in this embodiment, historical daily load data of each power distribution device in the power distribution network and historical daily load data of the power distribution network can be obtained from the data storage system.

[0064] S202, construct at least two different generalized linear models.

[0065] Among them, the generalized linear model is a method that uses the linear prediction function of the independent variable as the estimate of the dependent variable, and can be applied to data with nonlinear characteristics by using linear transformation.

[0066] Optionally, the construction of a generalized linear model requires three necessary conditions: the model's density function follows a distribution within an exponential family of distributions with parameter η; it has a connection function g(x); and its expectation function h(x) = E[y|x,θ] = μ = g. -1 (η)=g -1 (x T β). Common connection functions can be categorized into exponential family types such as logarithmic, Poisson, and normal distribution functions. Furthermore, different choices of connection functions will construct different generalized linear models.

[0067] Specifically, after the initial model is built following the selection of the connection function, the model is progressively optimized. This involves gradually removing power distribution equipment with little correlation to the prediction until the optimal performance index of the generalized linear model is achieved. During this process, multiple different generalized linear models can be constructed. In this embodiment, the generalized linear model is one that predicts future daily load data of the power distribution network based on the daily load parameters of the power distribution equipment. Optionally, different generalized linear models may include daily load parameters for different power distribution equipment.

[0068] Understandably, the generalized linear model treats the nonlinear relationships in the data as linear relationships, obscuring the complex relationships between nonlinear data, which facilitates the rapid prediction of future daily load data of the distribution network.

[0069] S203 uses historical daily load data of each power distribution device to initialize the daily load parameters in each generalized linear model, and obtains the predicted daily load data of the power distribution network predicted by each generalized linear model.

[0070] Specifically, after constructing the generalized linear model, for each constructed generalized linear model, the historical daily load data of the power distribution equipment required by the generalized linear model can be selected from the historical daily load data of each power distribution equipment; then, using the selected historical daily load data, the daily load parameters of the power distribution equipment included in the generalized linear model are assigned values, and the predicted daily load data of the power distribution network predicted by the generalized linear model can be obtained.

[0071] For example, a generalized linear model includes the daily load parameters of power distribution equipment A, power distribution equipment B, and power distribution equipment C, and the historical daily load data of power distribution equipment A, power distribution equipment B, and power distribution equipment C are all daily load data within one year prior to the current time. In this case, for each day prior to the current time, the daily load data of power distribution equipment A, power distribution equipment B, and power distribution equipment C on that day can be used to assign values ​​to the daily load parameters of power distribution equipment A, power distribution equipment B, and power distribution equipment C in the generalized linear model, and the generalized linear model can then predict the daily load data of the entire power distribution network on that day.

[0072] S204. Based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by each generalized linear model, select the target model from each generalized linear model.

[0073] In this embodiment, the target model refers to the generalized linear model that best fits the data among all established generalized linear models, and is used to predict the future daily load data of the distribution network based on the current daily load data of each distribution device in the distribution network.

[0074] Optionally, there are many ways to select the target model from the various generalized linear models, and this embodiment does not limit the comparison. For example, one possible approach is to input the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by each generalized linear model into a pre-trained neural network, and then select the target model from the various generalized linear models based on the evaluation scores of each generalized linear model output by the network.

[0075] Another approach is to further process the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by various generalized linear models based on a pre-defined model selection logic, and then determine the target model based on the processing results.

[0076] Furthermore, after determining the target model, the current daily load data of each distribution device in the distribution network can be substituted into the selected target model to predict the future daily load data of the distribution network. Specifically, the current daily load data of the distribution devices required by the target model is selected from the current daily load data of each distribution device, and the selected current daily load data is substituted into the target model to obtain the future daily load data predicted by the target model.

[0077] In this embodiment, multiple generalized linear models are constructed based on the historical daily load data of each power distribution device in the distribution network. Based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network obtained from each generalized linear model, a target model is selected from these models to predict the future daily load data of the distribution network. This approach eliminates the need to rely on external data such as population density and humidity; it constructs a model capable of accurately predicting the future daily load data of the distribution network solely using the historical daily load data of each power distribution device within the network. This avoids inaccurate predictions of the future daily load data of the distribution network caused by poor quality or inefficient acquisition of external data.

[0078] Based on the above embodiments, such as Figure 4 As shown, one embodiment involves selecting a target model from various generalized linear models based on historical daily load data of the distribution network and predicted daily load data of the distribution network predicted by various generalized linear models. The steps include:

[0079] S401, for each generalized linear model, determine the model performance index value of the generalized linear model based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the generalized linear model, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network.

[0080] In this embodiment, the model performance index value may include the Akaike Information Criterion (AIC) score, which is used to measure the goodness of model fit. Specifically, it can be used to weigh the complexity of the estimated model against the goodness of the model fit to the data.

[0081] The number of model parameters refers to the number of parameters included in the generalized linear model; the total number of historical daily load data of the distribution network, also known as the number of observations, is specifically the total number of days for which historical daily load data of the distribution network was collected.

[0082] One possible approach is to input the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the generalized linear model, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network into a pre-trained model evaluation network for each generalized linear model. The model evaluation network then outputs the model performance index value of the generalized linear model.

[0083] Another possible approach is to determine the residual sum of squares for each generalized linear model based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by the generalized linear model; and to determine the model performance index value of the generalized linear model based on the residual sum of squares, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network.

[0084] Among them, the residual sum of squares is a measure of the goodness of fit of a generalized linear model. It is a data processing method that uses a continuous curve to approximate or analogize a set of discrete points on a plane to represent the functional relationship between coordinates.

[0085] Optionally, the residual sum of squares (SSR) can be determined using the following formula 1. Where y represents the historical daily load data of the distribution network. This refers to the predicted daily load data of the distribution network as predicted by the generalized linear model.

[0086]

[0087] After determining the sum of squared residuals, the model performance index value AIC can be determined using the following formula 2. Here, k is the number of model parameters, and n is the total number of historical daily load data points for the distribution network.

[0088] AIC = 2k + nln(SSR / n) (Formula 2)

[0089] Understandably, in this example, calculating the AIC value can more accurately select the optimal model among multiple generalized linear models, avoiding the problem that the predicted future daily load data of the distribution network differs greatly from the actual results due to the model's poor quality.

[0090] S402, Select the target model from the generalized linear models based on the model performance index values ​​of each generalized linear model.

[0091] Optionally, increasing the number of parameters can improve the goodness of fit. AIC encourages good data fit but avoids overfitting. Therefore, the model with the smallest AIC value should be given priority. For example, when choosing from n models, the AIC values ​​of all n models can be calculated at once, and the model with the smallest AIC value can be selected as the target model.

[0092] In this embodiment, by using the model performance index values ​​of multiple generalized linear models, the optimal model can be accurately selected, avoiding inaccurate prediction of future daily load data of the distribution network due to low model fit, and improving the accuracy of prediction.

[0093] In one embodiment, such as Figure 5 As shown, based on the model performance index values ​​of each generalized linear model, a target model is selected from the generalized linear models. This embodiment includes the following steps:

[0094] S501: Based on the model performance index values ​​of each generalized linear model, select candidate models from each generalized linear model.

[0095] In this process, after calculating the model performance index values ​​of each generalized linear model, such as the AIC score of the Akaike Information Criterion, it can be found that the AIC values ​​of several generalized linear models are not significantly different, making it impossible to determine an optimal model. Therefore, these generalized linear models with insignificant differences in AIC values ​​are called candidate models and are further screened.

[0096] S502, based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the candidate model, and the total number of historical daily load data of the distribution network, determine the error data of the candidate model.

[0097] The error data is used to measure the difference between the actual data and the model's predicted data, specifically to measure the accuracy of the model's predicted data. Optionally, the error data in this embodiment may include root mean square error and / or average error.

[0098] Optionally, for each candidate model, the root mean square error (RMSE) can be determined using the following formula 3. Where y i This is historical daily load data for the distribution network. The predicted daily load data of the distribution network is given by the candidate model, where n is the total number of historical daily load data of the distribution network and N is the maximum number of days, such as 365.

[0099]

[0100] Furthermore, for each candidate model, the average error MAPE of that candidate model can be determined using the following formula 4.

[0101]

[0102] S503: Select the target model from the candidate models based on the error data of the candidate models.

[0103] Optionally, when the AIC values ​​of multiple generalized linear models are not significantly different, it is impossible to determine the optimal model. In this case, the values ​​of root mean square error (RMSE) and mean error (MAPE) are used as the final model selection criteria for the next step of screening.

[0104] Optionally, the model with the smallest root mean square error (RMSE) and mean average error (MAPE) among multiple candidate models can be selected as the target model.

[0105] In this embodiment, by incorporating the root mean square error (RMSE) and mean square error (MAPE) as metrics for the model, the target model with the smallest prediction error was selected from multiple candidate models, thereby improving the accuracy of the model's predictions.

[0106] In one embodiment, such as Figure 6 As shown, the historical daily load data of each power distribution device in the distribution network is the historical daily maximum load data of each power distribution device, and the historical daily load data of the distribution network is the historical daily maximum load data of the distribution network; correspondingly, the steps to obtain the historical daily load data of each power distribution device and the historical daily load data of the distribution network include:

[0107] S601, based on the historical daily original load data of each power distribution device in the power distribution network, determine the historical daily maximum load data of each power distribution device.

[0108] The historical daily load data for each power distribution device refers to the hourly load data of that device within a certain period prior to the current time (such as a month, a quarter, or a year). For example, it could be the maximum hourly load data of that device within a certain period prior to the current time. Correspondingly, the historical daily maximum load data for each power distribution device refers to the maximum load data among the daily load data of that device within a certain period prior to the current time.

[0109] Furthermore, the historical daily maximum load data for each power distribution device can be determined using the following formula 5. Where X... i This represents the historical daily maximum load of the i-th power distribution equipment. This represents the load data of power distribution equipment i at time j, that is, the power distribution data at the time with the largest load in 24 hours of a day is selected as the maximum daily load data for that day.

[0110]

[0111] Optionally, many power distribution devices may experience issues such as missing load data or significant data errors during the recording of daily load data due to maintenance, malfunctions, or other reasons. Therefore, to determine the historical daily maximum load data of each power distribution device based on its historical daily raw load data, another approach is to process the missing and / or outlier values ​​in the historical daily raw load data of each device; and then determine the historical daily maximum load data of each device based on the processed historical daily raw load data.

[0112] Commonly used methods include sample deletion, univariate imputation, regression imputation, and multiple imputation. Taking multiple imputation as an example, the steps are as follows: generate a set of possible imputations from the original loading data containing missing values, forming a set of multiple complete data; perform statistical analysis on these generated complete data, synthesize the results of each imputation data, generate the final statistical inference, and introduce confidence intervals for the missing values; and test for outliers, i.e., outliers that have been in a zero-load state for a long time or that are too far beyond the load quartile, and replace them using the above-mentioned multiple imputation methods.

[0113] After processing the historical daily raw load data of each power distribution device, the maximum daily load data of each power distribution device is calculated.

[0114] S602, based on the same-day rate and the historical daily original load data of each power distribution equipment, determine the historical daily maximum load data of the power distribution network.

[0115] The same-day rate refers to the probability that different power distribution equipment will work together in the same time period. In this embodiment, the power distribution equipment that works together in different time periods is different.

[0116] Furthermore, because the hourly timeframes for selecting the maximum daily load data for each power distribution device differ, the historical daily maximum load data of the distribution network cannot be simply summed by the maximum daily load data of all power distribution devices for that day. Instead, the distribution data of all devices within the distribution network are summed hourly and multiplied by the simultaneity rate to obtain the hourly load data of the distribution network system. Here, summing hourly refers to summing the load data of power distribution devices operating within the same time period. For example, if only power distribution devices A and C operate simultaneously between 3:00 and 4:00 on a given day, the load data of power distribution devices A and C are added together and multiplied by the simultaneity rate to obtain the load data of the distribution network system during the 3:00-4:00 time period.

[0117] After calculating the hourly load data of the distribution network system, the maximum load data of the distribution network within 24 hours is selected. The selected data is the historical daily maximum load data of the distribution network system.

[0118] In this embodiment, by preprocessing the historical daily raw load data of each power distribution device, the accuracy of model training is improved, thereby improving the accuracy of model prediction.

[0119] In one embodiment, taking the annual load data of each power distribution device in a small power distribution network consisting of 15 power distribution devices in a certain region as an example, the specific implementation includes:

[0120] Because the 15 power distribution devices were randomly selected, their maximum daily load data fluctuates greatly, and the load differences between different devices are also significant. For example, the average maximum daily load of power distribution device 1 and power distribution device 2 is approximately 2500 and 1.5, respectively. This random sampling method reflects the different electricity consumption patterns in different areas in real life. For example, the electricity consumption in residential areas is much lower than that in commercial and industrial areas.

[0121] The historical daily load data of the 15 initialized power distribution devices are input into the generalized linear model, which can be represented as Equation 6. Here, f(x) represents the daily peak load data of the small power distribution network predicted by the generalized linear model for the second day, i represents the device number, and x represents the load value. i This represents the maximum load data of the power distribution equipment on that day. In this embodiment, the function g(x) is connected using log(x) based on the data distribution, and the conditional probability distribution is a Gaussian distribution.

[0122] f(x) = g(x1 + x2 + ... + x) n )(Formula 6)

[0123] Substituting the historical daily load data of the initialized power distribution equipment yields the generalized linear model results. Then, the generalized linear model needs to be further filtered using the AIC value. Specifically, the number of power distribution equipment in the generalized linear model that is not closely related to the predicted peak daily load data of the small power distribution network and has multiple linear correlations needs to be reduced, so that the load prediction of the generalized linear model is more accurate.

[0124] During the model simplification process, four process models were formed, denoted as Model 1 to Model 4, where Model 1 is the initial model, and so on. A backward stepwise regression method was used to successively reduce the number of power distribution equipment until a generalized linear model of power distribution equipment combinations that prevented the AIC value from decreasing further was found. This generalized linear model was then designated as the optimal prediction model.

[0125] Further verification of the generalized linear model reveals that the predicted maximum load data and the actual maximum load data show little difference in basic trend, and the average absolute percentage error is stable between 5% and 6%, indicating that the prediction results are relatively accurate.

[0126] It should be understood that although the steps in the flowcharts of the embodiments described above 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 embodiments described above 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.

[0127] Based on the same inventive concept, this application also provides a load forecasting apparatus for implementing the load forecasting method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more load forecasting apparatus embodiments provided below can be found in the limitations of the load forecasting method described above, and will not be repeated here.

[0128] In one embodiment, such as Figure 7 As shown, a load forecasting device 1 is provided, including: a data acquisition module 10, a model building module 20, an initialization module 30, and a model selection module 40, wherein:

[0129] Data acquisition module 10 is used to acquire historical daily load data of each power distribution device in the power distribution network and historical daily load data of the power distribution network.

[0130] Model building module 20 is used to build at least two different generalized linear models; wherein the different generalized linear models include the daily load parameters of different power distribution equipment;

[0131] Initialization module 30 is used to initialize the daily load parameters in each generalized linear model using the historical daily load data of each power distribution equipment, so as to obtain the predicted daily load data of the power distribution network predicted by each generalized linear model.

[0132] The model selection module 40 is used to select a target model from various generalized linear models based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by various generalized linear models; wherein, the target model is used to predict the future daily load data of the distribution network based on the current daily load data of each power distribution device in the distribution network.

[0133] In one embodiment, such as Figure 8 As shown, above Figure 7 The model selection module 40 includes:

[0134] Model index unit 41 is used to determine the model performance index value of each generalized linear model based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the generalized linear model, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network.

[0135] The model determination unit 42 is used to select the target model from each generalized linear model based on the model performance index value of each generalized linear model.

[0136] Building upon the previous embodiment, in one embodiment, the model index unit 41 is specifically used for:

[0137] The residual sum of squares is determined based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by the generalized linear model.

[0138] The model performance index value of the generalized linear model is determined based on the sum of squared residuals, the number of model parameters of the generalized linear model, and the total amount of historical daily load data of the distribution network.

[0139] Building upon the previous embodiment, in one embodiment, the model determination unit 42 is specifically used for:

[0140] Candidate models are selected from the generalized linear models based on their performance index values.

[0141] The error data of the candidate models are determined based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the candidate models, and the total number of historical daily load data of the distribution network. The error data includes the root mean square error and / or the average error.

[0142] Based on the error data of the candidate models, the target model is selected from the candidate models.

[0143] In one embodiment, such as Figure 9 As shown, Figure 7 The data acquisition module 10 includes:

[0144] The data processing unit 11 is used to determine the historical daily maximum load data of each power distribution device based on the historical daily raw load data of each power distribution device in the power distribution network. Specifically, it processes missing and / or outlier values ​​in the historical daily raw load data of each power distribution device; and determines the historical daily maximum load data of each power distribution device based on the processed historical daily raw load data.

[0145] The data calculation unit 12 is used to determine the historical daily maximum load data of the distribution network based on the same-day rate and the historical daily original load data of each power distribution equipment.

[0146] Each module in the aforementioned load forecasting 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.

[0147] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 10 As shown. The computer device includes a processor, memory, and network interface connected via a system bus. The processor provides computing 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 an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores historical daily load data for each power distribution device in the distribution network and historical daily load data for the distribution network itself. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements a load forecasting method. The display screen can be an LCD screen or an e-ink display screen. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0148] Those skilled in the art will understand that Figure 10 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.

[0149] In one 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 perform the following steps:

[0150] Obtain historical daily load data of each power distribution device in the power distribution network and historical daily load data of the power distribution network.

[0151] Construct at least two different generalized linear models; wherein the different generalized linear models include the daily load parameters of different power distribution equipment.

[0152] Historical daily load data of each power distribution device is used to initialize the daily load parameters in each generalized linear model, thereby obtaining the predicted daily load data of the power distribution network predicted by each generalized linear model.

[0153] Based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by various generalized linear models, a target model is selected from the various generalized linear models; wherein, the target model is used to predict the future daily load data of the distribution network based on the current daily load data of each power distribution device in the distribution network.

[0154] Based on the previous embodiment, in this embodiment, when the processor executes the logic in the computer program to select the target model from various generalized linear models based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by various generalized linear models, the specific steps are as follows:

[0155] For each generalized linear model, the model performance index value of the generalized linear model is determined based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the generalized linear model, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network. Based on the model performance index values ​​of each generalized linear model, the target model is selected from the generalized linear models.

[0156] Based on the previous embodiment, in this embodiment, when the processor executes the logic in the computer program to determine the model performance index value of the generalized linear model based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the generalized linear model, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network, the specific steps are as follows:

[0157] Based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by the generalized linear model, the residual sum of squares is determined; based on the residual sum of squares, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network, the model performance index value of the generalized linear model is determined.

[0158] In one embodiment, when the processor executes the logic in the computer program that selects the target model from various generalized linear models based on the model performance index values ​​of each generalized linear model, the following steps are specifically implemented:

[0159] Based on the model performance index values ​​of each generalized linear model, candidate models are selected from each generalized linear model; based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the candidate models, and the total number of historical daily load data of the distribution network, the error data of the candidate models are determined; the error data includes root mean square error and / or average error; based on the error data of the candidate models, a target model is selected from the candidate models.

[0160] In one embodiment, when the processor executes the logic in the computer program to obtain the historical daily load data of each power distribution device in the power distribution network and the historical daily load data of the power distribution network, the following steps are specifically implemented:

[0161] Based on the historical daily raw load data of each power distribution device in the power distribution network, determine the historical daily maximum load data of each power distribution device; based on the same-day rate and the historical daily raw load data of each power distribution device, determine the historical daily maximum load data of the power distribution network.

[0162] In one embodiment, when the processor executes the logic in the computer program to determine the historical daily maximum load data of each power distribution device based on the historical daily raw load data of each power distribution device in the power distribution network, the following steps are specifically implemented:

[0163] The missing and / or outlier values ​​in the historical daily raw load data of each power distribution device are processed; based on the processed historical daily raw load data, the historical daily maximum load data of each power distribution device is determined.

[0164] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0165] Obtain historical daily load data of each power distribution device in the power distribution network and historical daily load data of the power distribution network.

[0166] Construct at least two different generalized linear models; wherein the different generalized linear models include the daily load parameters of different power distribution equipment.

[0167] Historical daily load data of each power distribution device is used to initialize the daily load parameters in each generalized linear model, thereby obtaining the predicted daily load data of the power distribution network predicted by each generalized linear model.

[0168] Based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by various generalized linear models, a target model is selected from the generalized linear models. The target model is used to predict the future daily load data of the distribution network based on the current daily load data of each distribution device in the distribution network.

[0169] Based on the previous embodiment, when the computer program is executed by the processor to select a target model from various generalized linear models based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by various generalized linear models, the following steps are specifically implemented:

[0170] For each generalized linear model, the model performance index value of the generalized linear model is determined based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the generalized linear model, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network. Based on the model performance index values ​​of each generalized linear model, the target model is selected from the generalized linear models.

[0171] Based on the previous embodiment, when the computer program is executed by the processor to determine the model performance index value of the generalized linear model based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the generalized linear model, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network, the specific steps are as follows:

[0172] Based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by the generalized linear model, the residual sum of squares is determined; based on the residual sum of squares, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network, the model performance index value of the generalized linear model is determined.

[0173] In one embodiment, when a computer program is executed by a processor to select a target model from various generalized linear models based on the model performance index values ​​of each generalized linear model, the following steps are specifically implemented:

[0174] Based on the model performance index values ​​of each generalized linear model, candidate models are selected from each generalized linear model; based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the candidate models, and the total number of historical daily load data of the distribution network, the error data of the candidate models are determined; the error data includes root mean square error and / or average error; based on the error data of the candidate models, the target model is selected from the candidate models.

[0175] In one embodiment, when a computer program is executed by a processor to acquire historical daily load data of each power distribution device in the power distribution network and historical daily load data of the power distribution network, the following steps are specifically implemented:

[0176] Based on the historical daily raw load data of each power distribution device in the power distribution network, determine the historical daily maximum load data of each power distribution device; based on the same-day rate and the historical daily raw load data of each power distribution device, determine the historical daily maximum load data of the power distribution network.

[0177] In one embodiment, when a computer program is executed by a processor to acquire historical daily load data of each power distribution device in the power distribution network and historical daily load data of the power distribution network, the following steps are specifically implemented:

[0178] Based on the historical daily raw load data of each power distribution device in the power distribution network, determine the historical daily maximum load data of each power distribution device; based on the same-day rate and the historical daily raw load data of each power distribution device, determine the historical daily maximum load data of the power distribution network.

[0179] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0180] Obtain historical daily load data of each power distribution device in the power distribution network and historical daily load data of the power distribution network.

[0181] Construct at least two different generalized linear models; wherein the different generalized linear models include the daily load parameters of different power distribution equipment.

[0182] Historical daily load data of each power distribution device is used to initialize the daily load parameters in each generalized linear model, thereby obtaining the predicted daily load data of the power distribution network predicted by each generalized linear model.

[0183] Based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by various generalized linear models, a target model is selected from the generalized linear models. The target model is used to predict the future daily load data of the distribution network based on the current daily load data of each distribution device in the distribution network.

[0184] Based on the previous embodiment, when the computer program is executed by the processor to select a target model from various generalized linear models based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by various generalized linear models, the following steps are specifically implemented:

[0185] For each generalized linear model, the model performance index value of the generalized linear model is determined based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the generalized linear model, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network. Based on the model performance index values ​​of each generalized linear model, the target model is selected from the generalized linear models.

[0186] Based on the previous embodiment, when the computer program is executed by the processor to determine the model performance index value of the generalized linear model based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the generalized linear model, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network, the specific steps are as follows:

[0187] Based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by the generalized linear model, the residual sum of squares is determined; based on the residual sum of squares, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network, the model performance index value of the generalized linear model is determined.

[0188] In one embodiment, when a computer program is executed by a processor to select a target model from various generalized linear models based on the model performance index values ​​of each generalized linear model, the following steps are specifically implemented:

[0189] Based on the model performance index values ​​of each generalized linear model, candidate models are selected from each generalized linear model; based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the candidate models, and the total number of historical daily load data of the distribution network, the error data of the candidate models are determined; the error data includes root mean square error and / or average error; based on the error data of the candidate models, the target model is selected from the candidate models.

[0190] In one embodiment, when a computer program is executed by a processor to acquire historical daily load data of each power distribution device in the power distribution network and historical daily load data of the power distribution network, the following steps are specifically implemented:

[0191] Based on the historical daily raw load data of each power distribution device in the power distribution network, determine the historical daily maximum load data of each power distribution device; based on the same-day rate and the historical daily raw load data of each power distribution device, determine the historical daily maximum load data of the power distribution network.

[0192] In one embodiment, when a computer program is executed by a processor to determine the historical daily maximum load data of each distribution device based on the historical daily raw load data of each distribution device in the distribution network, it also performs the following steps:

[0193] The missing and / or outlier values ​​in the historical daily raw load data of each power distribution device are processed; based on the processed historical daily raw load data, the historical daily maximum load data of each power distribution device is determined.

[0194] It should be noted that the data related to the power distribution network involved in this application (including but not limited to historical daily load data and current daily load data of power distribution equipment, as well as historical daily load data of the power distribution network, etc.) are all data authorized by the user or fully authorized by all parties.

[0195] 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 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, etc., and are not limited to these.

[0196] 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 specification.

[0197] 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 load forecasting method, characterized in that, The method includes: Acquire historical daily load data of each power distribution device in the power distribution network and historical daily load data of the power distribution network; Construct at least two different generalized linear models; wherein the different generalized linear models include the daily load parameters of different power distribution equipment; Historical daily load data of each power distribution device is used to initialize the daily load parameters in each generalized linear model, thereby obtaining the predicted daily load data of the power distribution network predicted by each generalized linear model. For each generalized linear model, the residual sum of squares is determined based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by the generalized linear model. Then, the model performance index value of the generalized linear model is determined based on the residual sum of squares, the number of model parameters of the generalized linear model, and the total number of historical daily load data of the distribution network. The model performance index value includes the Akaike Information Criterion (AIC) score. Based on the model performance index values ​​of each generalized linear model, a target model is selected from each generalized linear model; wherein, the target model is used to predict the future daily load data of the distribution network based on the current daily load data of each power distribution device in the distribution network.

2. The method according to claim 1, characterized in that, The step of selecting the target model from the generalized linear models based on their performance index values ​​includes: Candidate models are selected from the generalized linear models based on their model performance index values. The error data of the candidate model is determined based on the historical daily load data of the distribution network, the predicted daily load data of the distribution network predicted by the candidate model, and the total number of historical daily load data of the distribution network; the error data includes root mean square error and / or average error. Based on the error data of the candidate models, a target model is selected from the candidate models.

3. The method according to claim 2, characterized in that, The step of selecting a target model from the candidate models based on the error data of the candidate models includes: The candidate model with the smallest error data is selected as the target model.

4. The method according to claim 1, characterized in that, When the model performance index value is the AIC score, the step of selecting the target model from each generalized linear model based on the model performance index value of each generalized linear model includes: The generalized linear model corresponding to the minimum AIC score is used as the target model.

5. The method according to any one of claims 1-4, characterized in that, The historical daily load data of each power distribution device in the power distribution network is the historical daily maximum load data of each power distribution device, and the historical daily load data of the power distribution network is the historical daily maximum load data of the power distribution network. Accordingly, acquiring the historical daily load data of each power distribution device in the power distribution network and the historical daily load data of the power distribution network includes: Based on the historical daily raw load data of each power distribution device in the power distribution network, determine the historical daily maximum load data of each power distribution device; Based on the same-day rate and the historical daily original load data of each power distribution equipment, the historical daily maximum load data of the power distribution network is determined.

6. The method according to claim 5, characterized in that, The step of determining the historical daily maximum load data of each power distribution device based on the historical daily raw load data of each power distribution device in the power distribution network includes: Process missing and / or outlier values ​​in the historical daily raw load data of each power distribution device; Based on the processed historical daily raw load data, determine the historical daily maximum load data for each power distribution device.

7. A load forecasting device, characterized in that, The device includes: The data acquisition module is used to acquire the historical daily load data of each power distribution device in the power distribution network and the historical daily load data of the power distribution network. The model building module is used to build at least two different generalized linear models; wherein the different generalized linear models include the daily load parameters of different power distribution equipment; An initialization module is used to initialize the daily load parameters in each generalized linear model using historical daily load data of each power distribution device, so as to obtain the predicted daily load data of the power distribution network predicted by each generalized linear model. The model selection module is used to determine the residual sum of squares for each generalized linear model based on the historical daily load data of the distribution network and the predicted daily load data of the distribution network predicted by the generalized linear model. It then determines the model performance index value of the generalized linear model based on the residual sum of squares, the number of model parameters of the generalized linear model, and the total amount of historical daily load data of the distribution network. Based on the model performance index values ​​of each generalized linear model, a target model is selected from the generalized linear models. The model performance index value includes the Akaike Information Criterion (AIC) score. The target model is used to predict the future daily load data of the distribution network based on the current daily load data of each distribution device in the distribution network.

8. 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 6.

9. 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 6.

10. 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 6.

Images