Method for generating prediction and analytical model of power data in substation area, prediction method and device

Abstract

The present application discloses a prediction and analytical model generation method for substation power data, a prediction method and a device, and relates to the field of intelligent power distribution technology. The prediction and analytical model generation method for substation power data includes: based on the weight of the neural network model, selecting at least part of the model input parameters from the model input parameters of the neural network model; obtaining an initial element set based on at least part of the model input parameters, and obtaining a data calculation method set; obtaining a target element from the initial element set, and obtaining a target data calculation method based on the data calculation method set; obtaining a display analytical expression based on the target element and the target data calculation method, the display analytical expression represents the predicted value of the power parameter to be predicted obtained based on the target element using the target data calculation method; obtaining a prediction analytical model based on the display analytical expression. The present application improves the prediction accuracy and interpretability of substation power data.

Description

Technical Field

[0001] The present application relates to the field of intelligent power distribution technology, and in particular to a method for generating a prediction and analysis model for substation power data, a prediction method, and a device. Background Art

[0002] Predicting power data such as current, voltage, and load at power substations is a crucial foundational task in distribution network management. Effective prediction methods can effectively improve the intelligence of distribution equipment and enhance substation management efficiency. Currently, methods for predicting power data such as current, voltage, and load at power substations lack sufficient accuracy, are inefficient, and lack interpretability. This lack of interpretability also hinders robustness. Summary of the Invention

[0003] The embodiments of the present application are intended to solve, at least to a certain extent, one of the technical problems in the related art. To this end, the purpose of the embodiments of the present application is to propose a method for generating a predictive analytical model for substation power data, a prediction method, an apparatus, a data processing unit, an electronic device, a distribution network system, a cloud-edge system, a storage medium, and a program product.

[0004] An embodiment of the present application provides a method for generating a prediction and analytical model for substation power data, the prediction and analytical model generation method comprising: selecting at least part of the model input parameters from the model input parameters of the neural network model based on the weights of a neural network model, wherein the neural network model is used to make predictions based on the parameter values ​​of the model input parameters to obtain predicted values ​​of the power parameters to be predicted; obtaining an initial element set based on the at least part of the model input parameters, and acquiring a data calculation method set; obtaining a target element from the initial element set, and obtaining a target data calculation method based on the data calculation method set; obtaining a display analytical expression based on the target element and the target data calculation method, wherein the display analytical expression represents the predicted value of the power parameter to be predicted obtained based on the target element using the target data calculation method; obtaining a prediction analytical model based on the display analytical expression, wherein the prediction analytical model is used to predict the predicted value of the power parameter to be predicted.

[0005] Exemplarily, each model input parameter corresponds to a weight; the weight based on the neural network model selects at least part of the model input parameters from the model input parameters of the neural network model, including: sorting the weights corresponding to each model input parameter, and selecting at least one weight ranked first from the sorted weights; and using the model input parameter corresponding to the selected at least one weight as the selected at least part of the model input parameters.

[0006] Exemplarily, the obtaining of the initial element set based on the at least part of the model input parameters includes: taking the at least part of the model input parameters as the initial element set, wherein the target element includes the target parameter selected from the at least part of the model input parameters, and the display analytical expression represents that the parameter value of the target parameter is processed by the target data calculation method to obtain the predicted value of the power parameter to be predicted; or obtaining at least one constant, and taking the at least part of the model input parameters and the at least one constant as the initial element set, wherein the target element includes the target parameter selected from the at least part of the model input parameters and the target constant selected from the at least one constant, and the display analytical expression represents that the parameter value of the target parameter and the target constant are processed by the target data calculation method to obtain the predicted value of the power parameter to be predicted.

[0007] Exemplarily, the obtaining of the data calculation method set includes: constructing the data calculation method set based on operators and / or functional relationships.

[0008] Exemplarily, the method of obtaining the target element from the initial element set and obtaining the target data calculation method based on the data calculation method set includes: constructing an initial population based on the initial element set and the data calculation method set, wherein the initial population includes at least one individual, and each individual includes at least one element selected from the initial element set and at least one data calculation method selected from the data calculation method set; using the initial population or the next generation population obtained by genetic operation as the current population; calculating the fitness score of each individual in the current population, and judging whether the first current condition satisfies the first end condition; if the first current condition does not meet the first end condition, Satisfy the first end condition, perform genetic operation on the current population based on the fitness score to obtain the next generation population, and judge whether the second current condition satisfies the second end condition; if the second current condition does not satisfy the second end condition, use the next generation population as the current population for the next round of calculation; if the first current condition satisfies the first end condition, use the current population as the last generation population; if the second current condition satisfies the second end condition, use the next generation population as the last generation population, and determine the target individual from the last generation population; use the element in the target individual as the target element, and use the data calculation method in the target individual as the target data calculation method.

[0009] Exemplarily, the calculation of the fitness score of each individual in the current group includes: obtaining sample data, wherein the sample data includes the actual parameter value of the parameter and the actual parameter value of the power parameter to be predicted; using the data calculation method in the individual to process the actual parameter value of the parameter to obtain a processing result; and obtaining the fitness score of the individual based on the difference between the processing result and the actual parameter value of the power parameter to be predicted.

[0010] Exemplarily, the genetic operation includes at least one of the following: retaining at least one individual with the highest fitness score;

[0011] Selecting individuals from the current population; copying at least one individual with the highest fitness score; mutating the individuals in the current population; inserting strings into the individuals in the current population; and recombining the individuals in the current population.

[0012] Exemplarily, the power parameter to be predicted includes at least one of the following: current, voltage, and load.

[0013] Another embodiment of the present application provides a method for predicting substation power data, the prediction method comprising: obtaining a prediction analytical model, wherein the prediction analytical model is obtained according to a prediction analytical model generation method for substation power data; inputting the parameter value of the target parameter into the prediction analytical model for prediction, and obtaining a predicted value of the power parameter to be predicted.

[0014] Another embodiment of the present application provides a prediction and analytical model generation device for substation power data, the prediction and analytical model generation device comprising: a selection module for selecting at least part of the model input parameters from the model input parameters of the neural network model based on the weights of the neural network model, wherein the neural network model is used to make predictions based on the parameter values ​​of the model input parameters to obtain predicted values ​​of the power parameters to be predicted; a first acquisition module for obtaining an initial element set based on at least part of the model input parameters, and acquiring a data calculation method set; a second acquisition module for obtaining a target element from the initial element set, and obtaining a target data calculation method based on the data calculation method set; a third acquisition module for obtaining a display analytical expression based on the target element and the target data calculation method, wherein the display analytical expression represents the predicted value of the power parameter to be predicted obtained based on the target element using the target data calculation method; a fourth acquisition module for obtaining a prediction analytical model based on the display analytical expression, wherein the prediction analytical model is used to predict the predicted value of the power parameter to be predicted.

[0015] Another embodiment of the present application provides a prediction device for substation power data, the prediction device comprising: an acquisition module for acquiring a prediction analytical model, wherein the prediction analytical model is obtained based on a prediction analytical model generation device for substation power data; a prediction module for predicting the parameter value of a target parameter, inputting the parameter value into the prediction analytical model for prediction, and obtaining a predicted value of the power parameter to be predicted.

[0016] Another embodiment of the present application provides a data processing unit, which includes a board or a chip, and is used to implement the steps of the method described in any of the above embodiments.

[0017] Another embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method described in any of the above embodiments when executing the computer program.

[0018] Another embodiment of the present application provides a distribution network system, which is used to implement the steps of the method described in any of the above embodiments, or the distribution network system includes the above electronic device.

[0019] Another embodiment of the present application provides a cloud-edge-terminal system, which includes a cloud, an edge, and a terminal. At least one of the cloud, edge, and terminal is used to implement the steps of the method described in any of the above embodiments, or at least one of the cloud, edge, and terminal includes the above-mentioned electronic device.

[0020] Another embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method described in any one of the above embodiments are implemented.

[0021] Another embodiment of the present application provides a computer program product, which includes instructions. When the instructions are executed by a processor of a computer device, the computer device is able to perform the steps of the method of any of the above embodiments.

[0022] In the above embodiment, based on the weights of the neural network model, at least some of the model input parameters are selected from the model input parameters of the neural network model; based on at least some of the model input parameters, an initial element set is obtained, and a data calculation method set is obtained; a target element is obtained from the initial element set, and a target data calculation method is obtained based on the data calculation method set; based on the target element and the target data calculation method, a display analytical expression is obtained, the display analytical expression representing the predicted value of the power parameter to be predicted obtained based on the target element using the target data calculation method; based on the display analytical expression, a prediction analytical model is obtained. This application improves the prediction accuracy and interpretability of substation power data. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A flow chart of a method for generating a predictive analytical model for substation power data provided in an embodiment of the present application;

[0024] Figure 2 A flow chart of a method for predicting substation power data provided by an embodiment of the present application field;

[0025] Figure 3 A schematic diagram of a device for generating a prediction and analytical model for substation power data provided in an embodiment of the present application;

[0026] Figure 4 A schematic diagram of a device for predicting substation power data provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] The following describes in detail embodiments of the present application, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0028] Predicting parameters such as current, voltage, and load at power grid substations is a crucial foundational task in distribution network management. Effective prediction methods can effectively improve the intelligence of distribution equipment and enhance substation management efficiency. Examples of distribution equipment include FTUs and DTUs.

[0029] At present, the prediction accuracy of the current, voltage and load prediction methods in the substation area is not high enough, the efficiency is relatively low, and it is difficult to provide good interpretability. Due to the lack of interpretability, it is difficult to further ensure its robustness.

[0030] For example, the prediction methods for substation current, voltage, and load can be divided into two categories:

[0031] The first type is the analytical expression fitting method. This method uses polynomial fitting, linear / nonlinear regression and other methods to fit an explicit analytical expression based on sample data to represent the changes in current, voltage and load.

[0032] The second category includes non-analytical methods such as neural networks, deep learning, and traditional genetic algorithms. These methods train models based on data and iteratively improve them, ultimately producing a "black box" model that can capture some characteristics of current, voltage, and load changes. However, this model cannot be expressed as an analytical expression with measurable parameters as dependent variables.

[0033] The advantage of the first type of fitting analytical method is that it can more intuitively display the laws of change of substation current, voltage, and load, but the disadvantage is that the prediction accuracy of this type of method is not high enough and the efficiency is relatively low.

[0034] The second type of non-analytical methods, especially deep learning methods, can better mine the features in the data as the scale of high-quality data continues to grow. However, due to their "black box" nature, they are difficult to provide good interpretability, and due to the lack of interpretability, it is difficult to further ensure their robustness.

[0035] In view of this, the present application proposes a method for generating a predictive analytical model for substation power data and a method for predicting substation power data, so as to provide a relatively accurate, efficient and interpretable characterization of substation power data, generate an explicit analytical expression, and predict future substation power data based on it.

[0036] Figure 1 A flowchart of a method for generating a predictive analytical model for substation power data provided in an embodiment of the present application.

[0037] like Figure 1 As shown, the method 100 for generating a prediction and analytical model for substation power data provided in an embodiment of the present application includes, for example, steps S110 - S150 .

[0038] Step S110 , based on the weights of the neural network model, at least some model input parameters are selected from the model input parameters of the neural network model, and the neural network model is used to perform prediction based on the parameter values ​​of the model input parameters to obtain a predicted value of the power parameter to be predicted.

[0039] Exemplarily, the model input parameters of the neural network model may include multiple ones, and the model output parameters are the power parameters to be predicted, which may be one or more. The model input parameters may be independent variables, and the power parameters to be predicted may be dependent variables. The neural network model may be trained in advance, and the sample data used to train the neural network model may be real historical data, and the sample data used to train the model may include the true values ​​of the independent variables and the true values ​​of the dependent variables. When training the model, the neural network model makes predictions based on the parameter values ​​of the model input parameters (the true values ​​of the independent variables) to obtain the predicted values ​​of the power parameters to be predicted (the predicted values ​​of the dependent variables), and adjusts the model weights based on the difference between the predicted values ​​of the predicted power parameters and the true values ​​of the dependent variables in the sample data, so that the model converges, thereby training the neural network model.

[0040] After the neural network model is obtained through training, at least part of the model input parameters are selected from the model input parameters of the neural network model based on the weights of the neural network model.

[0041] Step S120 , obtaining an initial element set based on at least part of the model input parameters, and acquiring a data calculation method set.

[0042] For example, the initial element set is constructed based on at least part of the model input parameters. In addition to at least part of the model input parameters, the initial element set may also include other data.

[0043] The data calculation method set includes multiple calculation methods, and the calculation methods are used to calculate elements in the initial element set.

[0044] Step S130 , obtaining a target element from the initial element set, and obtaining a target data calculation method based on the data calculation method set.

[0045] For example, at least some elements from the initial element combination are determined as target elements, and at least some calculation methods are determined from the data calculation method set as target data calculation methods.

[0046] Step S140 , obtaining a display analytical expression based on the target element and the target data calculation method, wherein the display analytical expression represents a predicted value of the power parameter to be predicted obtained based on the target element using the target data calculation method.

[0047] Exemplarily, the target element includes at least independent variable parameters and may also include other data. The power parameter to be predicted serves as the dependent variable, and the displayed analytical expression may represent the functional relationship between the independent variable and the dependent variable.

[0048] In step S150 , a prediction analytical model is obtained based on the displayed analytical expression, wherein the prediction analytical model is used to predict a predicted value of the power parameter to be predicted.

[0049] For example, the displayed analytical expression can be used directly as the prediction analytical model, or it can be used as the prediction analytical model after relevant optimization adjustments are made to the displayed analytical expression. After obtaining the prediction analytical model, the parameter values ​​of the independent variable parameters related to the power grid (which can be future parameter values) can be input into the prediction analytical model. The prediction analytical model outputs the predicted value of the power parameter to be predicted, thereby realizing the prediction of the future power parameter.

[0050] It can be seen that through the method of the present application, based on the weights of the neural network model, at least some model input parameters are selected from the model input parameters of the neural network model; based on at least some model input parameters, an initial element set is obtained, and a data calculation method set is obtained; a target element is obtained from the initial element set, and a target data calculation method is obtained from the data calculation method set; based on the target element and the target data calculation method, a display analytical expression is obtained, and the display analytical expression represents the predicted value of the power parameter to be predicted obtained based on the target element using the target data calculation method; based on the display analytical expression, a prediction analytical model is obtained. The prediction analytical model is interpretable and accurate in prediction, thereby improving the prediction accuracy and interpretability of the substation power data.

[0051] In another example, the model input parameters may include multiple, each model input parameter may correspond to a weight, and the weight corresponding to each model input parameter may be the original weight in the model, or the weight obtained by summing, weighted averaging, etc. the original weights. For example, a neural network model may include a multi-layer structure, each layer including multiple nodes, so the weights between nodes can be regarded as original weights, and the original weights are usually more than the number of model input parameters. Therefore, the weight corresponding to each model input parameter may be an important original weight associated with the model input parameter determined from the multiple original weights of the model as the weight corresponding to the model input parameter, or multiple original weights associated with the model input parameter may be determined from the multiple original weights of the model, and the multiple associated original weights may be summed, weighted averaged, etc. as the weight corresponding to the model input parameter.

[0052] The weights corresponding to each model input parameter are sorted, and at least one weight ranked first is selected from the sorted weights. The model input parameters corresponding to the at least one selected weight are then used as at least part of the selected model input parameters. For example, k weights ranked first are selected from the sorted weights, where k is an integer greater than 0, and the k model input parameters corresponding to the k weights are used as the selected parameters. The k model input parameters can then be added to the initial element set.

[0053] In one example, the power parameter to be predicted may be at least one of current, voltage, and load, and may also include other parameters.

[0054] When the power parameters to be predicted include current, voltage, and load, the model input parameters of the neural network model may include at least one of multiple parameters such as temperature, humidity, date, and time.

[0055] When the power parameter to be predicted is current, the model input parameters of the neural network model may include at least one of multiple parameters such as voltage, load, temperature, humidity, date, time, etc.

[0056] When the power parameter to be predicted is voltage, the model input parameters of the neural network model may include at least one of multiple parameters such as current, load, temperature, humidity, date, time, etc.

[0057] When the power parameter to be predicted is load, the model input parameters of the neural network model may include at least one of multiple parameters such as current, voltage, temperature, humidity, date, time, etc.

[0058] When the power parameters to be predicted are current and voltage, the model input parameters of the neural network model may include at least one of multiple parameters such as load, temperature, humidity, date, time, etc. The same applies to other cases and will not be described in detail here.

[0059] For example, consider the power parameters to be predicted, including current, voltage, and load, and the neural network model input parameters, including temperature, humidity, date, and time. Based on the model weights, k model input parameters, such as temperature, humidity, and time, are selected from the model input parameters, indicating that temperature, humidity, and time have a greater impact on current, voltage, and load.

[0060] In one example, an initial element set can be constructed based on at least some of the model input parameters and at least one constant. For example, in one case, at least some of the model input parameters can be used as the initial element set. In another case, at least one constant can be obtained, and at least some of the model input parameters and at least one constant can be used as the initial element set. In other words, the initial element set can include only k model input parameters, or the initial element set can include one or more constants in addition to the k model input parameters. For example, the initial element set includes temperature, humidity, time, constant 1, constant 2, constant 3, and so on.

[0061] In one example, a data calculation method set can be constructed based on operators and / or function relationships. The data calculation method set can include multiple operators and / or multiple function relationships. The operators can be addition, subtraction, multiplication, division, and other operators, and the function relationships can be elementary functions.

[0062] When determining the display analytical expression, it is necessary to determine a target element from the initial element set. The target element includes a target parameter selected from at least some of the model input parameters, and the display analytical expression represents the predicted value of the power parameter to be predicted obtained by processing the parameter value of the target parameter using a target data calculation method. Alternatively, the target element includes a target parameter selected from at least some of the model input parameters and a target constant selected from at least one constant, and the display analytical expression represents the predicted value of the power parameter to be predicted obtained by processing the parameter value of the target parameter and the target constant using the target data calculation method. A target data calculation method also needs to be obtained from the data calculation method set.

[0063] For example, a target parameter (e.g., temperature, time) and a target constant (e.g., constant 1) can be selected from the initial element set, and a target data calculation method can be selected from the data calculation method set. Based on the target parameter (e.g., temperature, time), the target constant (e.g., constant 1), and the target data calculation method, a display analytical expression can be obtained. The display analytical expression, for example, has the power parameter to be predicted (current, voltage, load) as the dependent variable, the target parameter (e.g., temperature, time) as the independent variable, and the target constant (e.g., constant 1) as the constant of the function.

[0064] For ease of understanding, the analytical expression is y=a*x1 2 -x2+b as an example, y can represent the power parameter to be predicted, and the target data calculation method selected from the data calculation method set includes, for example, the function relationship () 2 , operators *, +, -, x1, x2, for example, represent target parameters (temperature and time), and a, b, for example, are both target constants.

[0065] In one example, the target element may be obtained from the initial element set based on a genetic algorithm, and the target data calculation method may be obtained based on a data calculation method set.

[0066] For example, based on the initial element set and the data calculation method set, an initial group is constructed, the initial group includes at least one individual, and each individual includes at least one element selected from the initial element set and at least one data calculation method selected from the data calculation method set.

[0067] The genetic algorithm includes multiple rounds of calculations. For one round of calculations, the initial population or the next generation population obtained by genetic operations can be used as the current population.

[0068] Then, the fitness score of each individual in the current population is calculated, and it is determined whether the first current condition satisfies the first end condition. The current first condition may be, for example, the fitness score, and the first end condition may be, for example, a threshold condition. If the first current condition satisfies the first end condition, for example, the fitness score satisfies the threshold condition, such as if the fitness score is greater than a preset threshold. The current first condition may also be other conditions, such as the number of genetic iteration calculations or other conditions.

[0069] If the first current condition does not meet the first termination condition, a genetic operation is performed on the current population based on the fitness score to obtain the next generation population, and then a determination is made as to whether the second current condition meets the second termination condition. The second current condition may be, for example, the number of genetic iterations, and the second termination condition may be, for example, a preset number. For example, if the second current condition meets the second termination condition, the number of genetic iterations exceeds the preset number. If the second current condition does not meet the second termination condition, the next generation population is used as the current population for the next round of calculations.

[0070] If the first current condition meets the first end condition, the current group is taken as the last generation group. If the second current condition meets the second end condition, the next generation group is taken as the last generation group, and the target individual is determined in the last generation group. The element in the target individual is taken as the target element, and the data calculation method in the target individual is taken as the target data calculation method.

[0071] When calculating the fitness score of each individual in the current group, sample data may be first obtained. The sample data includes actual parameter values ​​of the parameters and actual parameter values ​​of the power parameters to be predicted. The sample data may be historical data.

[0072] Then, the data calculation method for each individual is used to process the actual parameter values ​​of the parameters to obtain a processing result. The data calculation method for each individual can be, for example, some functional relationship, in which the actual parameter values ​​of the parameters are input into the function as independent variables to obtain the processing result, which is the dependent variable result. Next, based on the difference between the processing result and the actual parameter value of the power parameter to be predicted, the fitness score of the individual is obtained. The smaller the difference between the processing result and the actual parameter value of the power parameter to be predicted, the higher the fitness score, indicating that the data calculation method of the individual is more accurate and effective.

[0073] In one example, the genetic operation includes at least one of the following:

[0074] Retain at least one individual with the highest fitness score;

[0075] Select individuals from the current population, for example randomly;

[0076] Copy at least one individual with the highest fitness score;

[0077] Mutate individuals in the current group, including changing one or more internal elements and / or calculation methods;

[0078] Perform string insertion operations on individuals in the current group. There are many ways to insert strings, including inserting any fragment of an individual into any position of the individual, inserting any fragment of an individual into any position of another individual, etc.

[0079] Recombining individuals in the current group. The recombining method includes splitting two or more individuals from one point or multiple points and reconnecting and combining them.

[0080] Figure 2 A flow chart of a method for predicting substation power data provided in an embodiment of the present application field.

[0081] like Figure 2 As shown, the prediction method 200 of the substation power data includes, for example, steps S210 - S220 .

[0082] Step S210: Obtain a prediction analysis model.

[0083] Exemplarily, the prediction analytical model may be an explicit analytical expression, and the prediction analytical model is obtained according to the above method.

[0084] In step S220 , the parameter value of the target parameter is input into the prediction analysis model for prediction to obtain the predicted value of the power parameter to be predicted.

[0085] For example, the target parameter may be an independent variable parameter, and the parameter value of the target parameter may be a specific numerical value. For example, the target parameter may be temperature, humidity, date, or time, and the parameter value of the target parameter may be a temperature value, a humidity value, a specific date, or a specific time. The parameter value of the target parameter may be a future value, and the predicted value of the power parameter to be predicted obtained thereby may be a future value.

[0086] This application also proposes a combination of a method for generating a prediction and analytical model for substation power data and a method for predicting substation power data.

[0087] (1) Based on the historical current, voltage, load, temperature, humidity, date, time and other data of the substation, a sample set is generated, and the samples are used to train the neural network model.

[0088] (2) The neural network model is trained by the sample set to obtain a neural network model that represents the relationship between the independent variables (such as temperature, humidity, date, time) and the power data to be predicted (such as substation current, voltage, and load).

[0089] (3) According to the weights of the neural network model, the independent variable set E corresponding to the top k weights is selected from high to low.

[0090] (4) Set the end conditions of the genetic algorithm evolution generation process.

[0091] (5) Create the initial population of the evolution process. Let R be the function symbol set (operator or other elementary function), R is the initial element set mentioned above. Let Q be the terminal symbol set (including variables or constants), where all variables are from the independent variable set E, Q is the data calculation method set mentioned above. The initial population includes at least one individual, each individual is a sequence a1, a2, ..., a n ,b1,b2,...,b m ,a i ∈R,b j ∈Q,m≥n(e-1)+1, where e is the maximum number of parameters processed by operators or other elementary functions in R, such as b1, b2..., b m In the example, the number of variables or constants processed by b1 is 3, the number of variables or constants processed by b2 is 2, and m The number of variables or constants processed is 1, and the maximum mesh number e can be 3.

[0092] (6) Calculate the fitness score of each individual.

[0093] (7) Check whether the end conditions are met. If not, proceed to the next step, otherwise end.

[0094] (8) Through genetic manipulation, the next generation of population is produced.

[0095] (9) Go to step (6).

[0096] The fitness score of each individual in step (6) reflects the degree to which the individual is likely to reach or approach the optimal solution during the calculation process. It is necessary to design a suitable fitness function to calculate the fitness score of each individual. Different fitness functions can be designed for different problems. For example, the following fitness function can be designed:

[0097]

[0098] Among them, n' is the number of samples, the function value f i，j It represents the function value obtained based on the function expression (data calculation method) corresponding to the i-th individual and the parameter value of the independent variable in the j-th sample (the actual parameter value of the independent variable parameter). The function value f i，j is the result of the above processing; S j It represents the true value (actual parameter value) of the power parameter to be predicted (e.g., current, voltage, load, etc.) actually measured in the jth sample.

[0099] The genetic operations in step (8) may include: retaining the individual with the highest current fitness, selection, replication, mutation (including changing one or more elements and / or calculation methods within it, etc.), string insertion (including inserting any fragment in an individual into any position of the individual, inserting any fragment in an individual into any position of another individual, etc.), and recombination (including splitting two or more individuals from one point or multiple points and reconnecting and combining them, etc.).

[0100] Through genetic algorithms, the individual with the highest fitness score can eventually be determined from the last generation of the population. This individual can include target parameters, target constants, and target data calculation methods (operators, other elementary functions). The target parameters are used as independent variables, and the power parameters to be predicted are used as dependent variables. The target data calculation method is used to calculate the target parameters and target constants to obtain the power parameters to be predicted, thereby obtaining the displayed analytical expression.

[0101] After obtaining the explicit analytical expression through the genetic algorithm, the current, voltage and load can be predicted based on the explicit analytical expression.

[0102] In another example of the present application, the above-mentioned prediction analysis model generation method and / or substation power data prediction method can be implemented as a module or an electronic device. The module or electronic device can read historical data such as substation current, voltage, load, etc., receive future prediction parameters, such as the future time and the type of parameter to be predicted (such as current, voltage, load, etc.), and output the value of the parameter to be predicted at a given future time.

[0103] In another example of the present application, the above-mentioned prediction analysis model generation method and / or substation power data prediction method can be integrated into a data processing unit, which includes a board or chip. The board or chip can be plugged into other devices or systems.

[0104] In another example of the present application, the above-mentioned prediction and analysis model generation method and / or the substation power data prediction method, or the electronic equipment of the module or the whole machine, can be embedded in the distribution network system, and the distribution network system includes, for example, FTU and DTU. Many distribution network devices and systems have the need to predict the substation current, voltage, load, etc. In these distribution network devices and systems, such as FTU and DTU, the electronic equipment equipped with the above-mentioned method, module or whole machine can support these devices and systems to achieve more intelligent functions. For example, in FTU and DTU, the equipment can predict the future values ​​of current, voltage, and load more accurately and efficiently. By embedding the electronic equipment of the above-mentioned method, module or whole machine into the distribution network system, more accurate substation load balancing, power scheduling, high-quality power supply, etc. can be achieved.

[0105] In another example of the present application, the above-mentioned prediction and analysis model generation method and / or substation power data prediction method, or the electronic equipment of the module or the whole machine, can be integrated into the cloud, edge, and terminal of the cloud-edge-end system, or at least one of the cloud, edge, and terminal. The cloud-edge-end system is used to implement distribution network management.

[0106] The above method proposed in this application can generate explicit analytical expressions for parameters such as current, voltage, and load relatively accurately and quickly, with high interpretability. By generating explicit analytical expressions, changes in the current, voltage, load, etc. of the substation area can be characterized. Based on the explicit analytical expressions, future current, voltage, load, etc. can be predicted. The efficiency of generating analytical expressions from data in this application is high, more than 2 orders of magnitude faster than traditional evolutionary algorithms, and has higher prediction accuracy. Traditional evolutionary algorithms include genetic algorithms (GA, Genetic Algorithm) and genetic programming (GP, Genetic Programming).

[0107] Figure 3 Schematic diagram of a device for generating a predictive analytical model for substation power data provided in an embodiment of the present application.

[0108] like Figure 3 As shown, the prediction and analytical model generating device 300 for substation power data includes: a selection module 310 , a first obtaining module 320 , a second obtaining module 330 , a third obtaining module 340 and a fourth obtaining module 350 .

[0109] The selection module 310 is used to select at least some model input parameters from the model input parameters of the neural network model based on the weights of the neural network model, wherein the neural network model is used to make predictions based on the parameter values ​​of the model input parameters to obtain predicted values ​​of the power parameters to be predicted.

[0110] The first obtaining module 320 is used to obtain an initial element set based on at least part of the model input parameters, and acquire a data calculation method set.

[0111] The second obtaining module 330 is used to obtain a target element from the initial element set, and obtain a target data calculation method based on the data calculation method set.

[0112] The third obtaining module 340 is used to obtain a display analytical expression based on the target element and the target data calculation method, wherein the display analytical expression represents the predicted value of the power parameter to be predicted obtained based on the target element using the target data calculation method.

[0113] The fourth obtaining module 350 is used to obtain a prediction analytical model based on the displayed analytical expression, wherein the prediction analytical model is used to predict a predicted value of the power parameter to be predicted.

[0114] It can be understood that for a detailed description of the device 300 for generating a prediction and analysis model for substation power data, reference can be made to the description of the method for generating a prediction and analysis model for substation power data above, which will not be repeated here.

[0115] Figure 4 A schematic diagram of a device for predicting substation power data provided in an embodiment of the present application.

[0116] like Figure 4 As shown, the prediction device 400 for substation power data includes: an acquisition module 410 and a prediction module 420 .

[0117] The acquisition module 410 is used to acquire the prediction and analysis model.

[0118] The prediction module 420 is used to predict the parameter value of the target parameter, input it into the prediction analysis model for prediction, and obtain the predicted value of the power parameter to be predicted.

[0119] It can be understood that for a detailed description of the prediction device 400 for substation power data, reference can be made to the description of the prediction method for substation power data above, which will not be repeated here.

[0120] An embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method in any of the above embodiments are implemented.

[0121] An embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method in any one of the above embodiments are implemented.

[0122] One embodiment of the present application provides a computer program product, which includes instructions. When the instructions are executed by a processor of a computer device, the computer device is enabled to perform the steps of the method of any of the above embodiments.

[0123] It should be noted that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device, or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device, or apparatus and execute the instructions), or in conjunction with such instruction execution systems, devices, or apparatuses. For purposes of this application, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by an instruction execution system, device, or apparatus, or in conjunction with such instruction execution systems, devices, or apparatuses. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection portion having one or more wires (electronic device), a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or otherwise processing it in a suitable manner if necessary, and then storing it in a computer memory.

[0124] It should be understood that various parts of the present application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0125] In the description of this application, reference to the terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of this application. In this application, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples.

[0126] In the description of the present application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0127] In addition, the terms "first" and "second" used in the embodiments of the present application are for descriptive purposes only and should not be understood as indicating or implying relative importance, or implicitly indicating the number of technical features indicated in the embodiments. Therefore, the features defined in the embodiments of the present application by terms such as "first" and "second" can explicitly or implicitly indicate that at least one of the features is included in the embodiment. In the description of the present application, the word "multiple" means at least two or two or more, such as two, three, four, etc., unless otherwise clearly and specifically defined in the embodiments.

[0128] In this application, unless otherwise specified or limited in the embodiments, the terms "installed", "connected", "connected", and "fixed" appearing in the embodiments should be understood in a broad sense. For example, the connection can be a fixed connection, a detachable connection, or an integral connection. It can also be a mechanical connection, an electrical connection, etc.; of course, it can also be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication between two elements, or the interaction between two elements. For those skilled in the art, the specific meanings of the above terms in this application can be understood based on the specific implementation.

[0129] In this application, unless otherwise expressly specified or limited, when a first feature is "above" or "below" a second feature, it may mean that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. Furthermore, when a first feature is "above," "above," or "above" a second feature, it may mean that the first feature is directly above or diagonally above the second feature, or simply means that the first feature is at a higher level than the second feature. When a first feature is "below," "below," or "below" a second feature, it may mean that the first feature is directly below or diagonally below the second feature, or simply means that the first feature is at a lower level than the second feature.

[0130] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for generating a prediction and analytical model for substation power data, characterized in that: The prediction and analysis model generation method includes: selecting at least some model input parameters from model input parameters of the neural network model based on weights of the neural network model, wherein the neural network model is used to perform prediction based on parameter values ​​of the model input parameters to obtain a predicted value of the power parameter to be predicted; obtaining an initial element set based on at least part of the model input parameters, and obtaining a data calculation method set; Obtaining a target element from the initial element set, and obtaining a target data calculation method based on the data calculation method set; Based on the target element and the target data calculation method, a display analytical expression is obtained, wherein the display analytical expression represents a predicted value of the power parameter to be predicted obtained based on the target element using the target data calculation method; Based on the displayed analytical expression, a prediction analytical model is obtained, wherein the prediction analytical model is used to predict a predicted value of the power parameter to be predicted; The set of calculation methods for obtaining data includes: Constructing the data calculation method set based on the operator and / or function relationship; Each model input parameter corresponds to a weight; the weight based on the neural network model selects at least some model input parameters from the model input parameters of the neural network model, including: Sort the weights corresponding to each model input parameter, and select at least one weight that ranks first among the sorted weights; The model input parameters corresponding to the selected at least one weight are used as at least part of the selected model input parameters.

2. The method for generating a prediction analysis model according to claim 1, wherein: The obtaining of an initial element set based on at least part of the model input parameters comprises: Using at least part of the model input parameters as the initial element set, wherein the target element includes a target parameter selected from the at least part of the model input parameters, and the explicit analytical expression represents a predicted value of the power parameter to be predicted obtained by processing the parameter value of the target parameter using the target data calculation method; or At least one constant is obtained, and the at least part of the model input parameters and the at least one constant are used as the initial element set, wherein the target element includes a target parameter selected from the at least part of the model input parameters and a target constant selected from the at least one constant, and the explicit analytical expression represents a predicted value of the power parameter to be predicted obtained by processing the parameter value of the target parameter and the target constant using the target data calculation method.

3. The prediction analysis model generation method according to claim 1, characterized in that: The obtaining of the target element from the initial element set and the obtaining of the target data calculation method based on the data calculation method set include: constructing an initial population based on the initial element set and the data calculation method set, wherein the initial population includes at least one individual, and each individual includes at least one element selected from the initial element set and at least one data calculation method selected from the data calculation method set; Taking the initial population or the next generation population obtained by genetic operation as the current population; Calculating the fitness score of each individual in the current population, and determining whether the first current condition satisfies the first end condition; If the first current condition does not satisfy the first end condition, performing a genetic operation on the current population based on the fitness score to obtain a next generation population, and determining whether the second current condition satisfies the second end condition; If the second current condition does not meet the second end condition, the next generation group is used as the current group for the next round of calculation; If the first current condition satisfies a first end condition, the current population is taken as the last generation population; if the second current condition satisfies a second end condition, the next generation population is taken as the last generation population, and a target individual is determined from the last generation population; The element in the target individual is used as the target element, and the data calculation method in the target individual is used as the target data calculation method.

4. The method for generating a prediction analysis model according to claim 3, wherein: Calculating the fitness score of each individual in the current population includes: Acquiring sample data, wherein the sample data includes actual parameter values ​​of the parameters and actual parameter values ​​of the power parameters to be predicted; Using the data calculation method in the individual, the actual parameter value of the processing parameter is processed to obtain the processing result; Based on the difference between the processing result and the actual parameter value of the power parameter to be predicted, the fitness score of the individual is obtained.

5. The prediction analysis model generation method according to claim 3, characterized in that: The genetic manipulation includes at least one of the following: Retain at least one individual with the highest fitness score; selecting an individual from the current population; Copy at least one individual with the highest fitness score; mutating individuals in the current population; Performing interpolation on individuals in the current group; The individuals in the current population are reorganized.

6. The method for generating a prediction analysis model according to any one of claims 1 to 5, wherein: The power parameter to be predicted includes at least one of the following: current, voltage, and load.

7. A method for predicting power data in a substation, characterized in that: The prediction method comprises: Obtaining a prediction and analysis model, wherein the prediction and analysis model is obtained according to the method according to any one of claims 1 to 6; The parameter value of the target parameter is input into the prediction analytical model for prediction to obtain the predicted value of the power parameter to be predicted.

8. A device for generating a prediction and analytical model for substation power data, characterized in that: The prediction and analysis model generation device is used to implement the prediction and analysis model generation method according to any one of claims 1 to 6, and the prediction and analysis model generation device includes: a selection module configured to select at least some model input parameters from model input parameters of the neural network model based on weights of the neural network model, wherein the neural network model is configured to perform prediction based on parameter values ​​of the model input parameters to obtain a predicted value of the power parameter to be predicted; a first obtaining module, configured to obtain an initial element set based on at least part of the model input parameters, and a data calculation method set; a second obtaining module, configured to obtain a target element from the initial element set, and obtain a target data calculation method based on the data calculation method set; a third obtaining module, configured to obtain a display analytical expression based on the target element and the target data calculation method, wherein the display analytical expression represents a predicted value of the power parameter to be predicted obtained based on the target element using the target data calculation method; The fourth obtaining module is used to obtain a prediction analytical model based on the display analytical expression, wherein the prediction analytical model is used to predict and obtain a predicted value of the power parameter to be predicted.

9. A prediction device for power data of a substation, characterized in that: The prediction device comprises: an acquisition module, configured to acquire a prediction and analysis model, wherein the prediction and analysis model is obtained by the apparatus according to claim 8; The prediction module is used to predict the parameter value of the target parameter, input it into the prediction analysis model for prediction, and obtain the predicted value of the power parameter to be predicted.

10. A data processing unit, characterized in that: The data processing unit includes a board or a chip, and the data processing unit is used to implement the steps of any one of claims 1-7.

11. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

12. A distribution network system, characterized in that: The distribution network system is used to implement the steps of the method according to any one of claims 1 to 7, or the distribution network system includes the electronic device according to claim 11.

13. A cloud-edge-end system, characterized in that: The cloud-edge-terminal system includes a cloud, an edge, and a terminal, and at least one of the cloud, edge, and terminal is used to implement the steps of the method described in any one of claims 1 to 7, or at least one of the cloud, edge, and terminal includes the electronic device of claim 11.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images