Data prediction method and device based on multi-angle model complementary integration

Abstract

The invention relates to the technical field of data prediction, in particular to a data prediction method and device based on multi-angle model complementary integration, and the method comprises the steps: analyzing past data, firstly carrying out segmented prediction according to a price interval, and adding the segmented prediction results as features into a subsequent prediction model, meanwhile, for the extreme values of the electricity price, whether the electricity price reaches the extreme values or not is predicted through a normalized probability prediction method, prediction results serve as extra features to be input into the model, and by solving different targets with the unified features, extreme fluctuation in the feature set and electricity price change can be more accurately captured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data prediction, in particular to a data prediction method and device based on multi-angle model complementary integration. BACKGROUND

[0002] Traditional time series methods rely on the time order of historical data for modeling, and are suitable for prediction tasks with strong stability. However, due to the volatility of electricity prices and complex market factors, traditional time series methods often have difficulty capturing nonlinear and periodic changes.

[0003] In recent years, machine learning and deep learning have been widely used in electricity price prediction. For example, using machine learning algorithms to build multiple arrays to form a forest for prediction can better handle nonlinear relationships. For example, using deep learning to build LSTM can perform well in time series prediction, especially in capturing long-term dependencies. However, the accuracy of these methods still has certain limitations when facing complex market fluctuations, extreme price values, etc.

[0004] Some studies propose to improve prediction accuracy by segmenting electricity prices or adding additional features. For example, using price intervals to classify and predict target values, or adding weather, demand, and other external features to enhance the model's expression ability. However, these methods are usually limited to simple segmentation or the introduction of static features, lacking dynamic adjustment and handling of extreme price values. SUMMARY

[0005] Therefore, the purpose of the present application is to provide a data prediction method and device based on multi-angle model complementary integration to overcome the problems existing in the prior art.

[0006] To achieve the above purpose, the present application adopts the following technical solutions: On the one hand, the present application provides a data prediction method based on multi-angle model complementary integration, comprising: Collecting electricity price data at a preset time interval and performing target value normalization processing on the electricity price data; Segmenting the target value normalized electricity price data by a preset price interval; Establishing a segmented prediction model and training the segmented prediction model through the segmentation results of the electricity price data; Judging whether the electricity price will be at these extreme values through a regression probability prediction method and generating corresponding probability features; Combining the data of each data point in the target value normalized electricity price data and the prediction results of the corresponding segmented prediction model, and the corresponding probability features calculated through the regression probability prediction method to generate a new electricity price feature set; obtaining an initial electricity price prediction constant model; training the electricity price prediction constant model by the new electricity price feature set, and iterating for a preset number of rounds, wherein, at each iteration, calculating the residual of the current model, and training a tree to fit the residual; updating the prediction results of the model at each iteration; determining the final electricity price prediction model by weighted sum of all trees generated at each iteration, and generating the final electricity price prediction results.

[0007] Further, the above method, the electricity price data is collected at a preset time interval, and the target value of the electricity price data is normalized, comprising: Collecting the electricity price data every 15 minutes in the past T days; sorting and cleaning all the electricity price data, and normalizing the target value of the electricity price data by a normalization formula; The normalization formula is:

[0008] Wherein, p i is the i-th value in the target value set P, p i ’ is the normalized value of p i . .

[0009] Further, the above method, the electricity price data after target value normalization is segmented by a preset price interval, comprising: Segmenting the electricity price data according to a preset price interval N; Calculating the segment where each data point in the electricity price data is located by a segmentation formula; The segmentation formula is: .

[0010] Further, the above method, the segmented prediction model is:

[0011] Wherein, y i is the data to be predicted, is the mean prediction value of the left subtree, is the mean prediction value of the right subtree, t L and t R are the sample sets of the left and right subtrees, respectively, |t L | and |t R | are the sample sizes of the left and right subtrees, respectively. The calculation formulas of left and right sub-trees are respectively: .

[0012] Further, the prediction result of the segmented prediction model in the above method is: ; wherein, is the prediction value of a certain leaf node, is the sample set contained by the leaf node, is the sum of the true target values of all samples on the leaf node.

[0013] Further, the probability feature in the above method includes: an extremely low value prediction probability and an extremely high value prediction probability; wherein, the extremely low value prediction probability is: ; the extremely high value prediction probability is: ; wherein, is the extremely low value prediction probability, is the extremely high value prediction probability, is the simplified training parameter of the extremely low model, is the simplified training parameter of the extremely high model.

[0014] Further, the electricity price prediction constant model in the above method is: ; wherein, is the prediction value of the initial model, y i is the true target value of the i-th sample in the training data, n is the number of samples in the training set, is the loss function, and θ is the mean value of the training data.

[0015] Further, the final electricity price prediction model in the above method is: ; wherein, is the final electricity price prediction result, is the learning rate, f t is the t-th tree, , X i is the original feature, and Z i is the supplementary feature.

[0016] Further, the above method further includes: The data collection days T are updated in real time, and the price interval N is dynamically adjusted according to market fluctuations.

[0017] In another aspect, the application provides a data prediction device based on multi-angle model complementary integration, comprising a processor and a memory, the processor being connected with the memory: The processor is configured to call and execute a program stored in the memory. The memory is configured to store the program, and the program is configured to execute the data prediction method based on multi-angle model complementary integration.

[0018] The application has the following advantages: The application first collects electricity price data at a preset time interval, and performs target value normalization on the electricity price data. The electricity price data after target value normalization is segmented by a preset price interval. Then, a segmented prediction model is established, the segmented prediction model is trained by the segmentation result of the electricity price data, it is judged whether the electricity price will be at these extreme values by a regression probability prediction method, and corresponding probability features are generated. The data of each data point in the electricity price data after target value normalization and the prediction result of the corresponding segmented prediction model, and the corresponding probability features calculated by the regression probability prediction method are combined to generate a new electricity price feature set. Finally, an initial electricity price prediction constant model is obtained, the electricity price prediction constant model is trained by the new electricity price feature set, and the iteration is performed for a preset number of times. In each iteration, the residual error of the current model is calculated, and a tree is trained to fit the residual error. The prediction result of the model is updated during iteration. The final electricity price prediction model is determined by weighted summation of all trees generated during iteration, and the final electricity price prediction result is generated. In the application, past data is analyzed, segmented prediction is first performed according to the price interval, and the results of these segmented predictions are added as features to the subsequent prediction model. At the same time, for the extreme values of the electricity price, the normalized probability prediction method is used to predict whether the electricity price will reach these extreme values, and these prediction results are input as additional features into the model. By solving different targets of unified features, the extreme fluctuations in the feature set and the change of the electricity price can be more accurately captured. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only constitute some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0020] Figure 1is a flow chart provided by one embodiment of a data prediction method based on multi-angle model complementary integration; Figure 2 is a structural schematic diagram provided by one embodiment of a data prediction device based on multi-angle model complementary integration. DETAILED DESCRIPTION

[0021] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described in detail below. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0022] Figure 1 is a flow chart provided by one embodiment of a data prediction method based on multi-angle model complementary integration. Please refer to Figure 1 , the present embodiment can include the following steps: S1, collecting price data at a preset time interval, and performing target value normalization processing on the price data; S2, segmenting the price data after target value normalization processing in a preset price interval; S3, establishing a segmented prediction model, and training the segmented prediction model through the segmentation result of the price data; It should be noted that when constructing the segmented prediction model, it is constructed as recursive division. At each node t, the model selects the optimal division point, which recursively divides the data set into left and right sub-trees, and continues to recursively reach the leaf node to reach the minimum sample number.

[0023] S4, determining whether the price will be at these extreme values through a regression probability prediction method, and generating corresponding probability features; S5, combining the data of each data point in the price data after target value normalization processing, the prediction result of the corresponding segmented prediction model, and the corresponding probability features calculated through the regression probability prediction method to generate a new price feature set; S6, obtaining an initial price prediction constant model; S7, training the price prediction constant model through the new price feature set, and iterating for a preset number of times, wherein the residual error of the current model is calculated at each iteration, and a tree is trained to fit the residual error; S8, updating the prediction result of the model during iteration; S9, determining the final price prediction model through weighted summation of all trees generated during iteration, and generating the final price prediction result.

[0024] It can be understood that the embodiment first collects the electricity price data at a preset time interval, and performs target value normalization processing on the electricity price data, segments the target value normalized electricity price data in a preset price interval, then establishes a segmented prediction model, trains the segmented prediction model through the segmentation results of the electricity price data, judges whether the electricity price will be in these extreme values through the regression probability prediction method, and generates corresponding probability features, to combine the data of each data point in the target value normalized electricity price data and the prediction results of the corresponding segmented prediction model, and the corresponding probability features calculated through the regression probability prediction method to generate a new electricity price feature set, finally, an initial electricity price prediction constant model is obtained, the electricity price prediction constant model is trained through the new electricity price feature set, and is iterated for a preset number of times, wherein the residual error of the current model is calculated at each iteration, and a tree is trained to fit the residual error, the prediction result of the model is updated at iteration, the final electricity price prediction model is determined through weighted summation of all trees generated at iteration, and the final electricity price prediction result is generated. In the embodiment, by analyzing the past data, first, segmented prediction is performed according to the price interval, and the results of the segmented prediction are added as features to the subsequent prediction model, at the same time, for the extreme values of the electricity price, the normalized probability prediction method is used to predict whether the electricity price will reach these extreme values, and these prediction results are input as additional features into the model, by solving different targets of unified features, the extreme fluctuations in the feature set and the change of the electricity price can be more accurately captured.

[0025] Preferably, step S1 comprises: collecting the electricity price data every 15 minutes in the past T days; sorting and cleaning all the electricity price data, and performing target value normalization processing on the electricity price data through a normalization formula; wherein the normalization formula is:

[0026] wherein p i is the i th value in the target value set P, p i ’ is the value after normalization calculation of p i .

[0027] It can be understood that the data of every 15 minutes in the past T days is sorted and cleaned to generate a total of 96T data as input, wherein the target value is normalized.

[0028]

[0029] wherein, .

[0030] ​Preferably, step S2 comprises: segmenting the electricity price data according to the preset price interval N; calculating the segment where each data point in the electricity price data is located by a segment formula; wherein the segment formula is: .

[0031] It can be understood that the electricity price is segmented according to the preset price interval N, and the segment where each data point is located is calculated.

[0032] .

[0033] Preferably, the segment prediction model is:

[0034] wherein y i is the data to be predicted, is the mean prediction value of the left sub-tree, is the mean prediction value of the right sub-tree, t L and t R are the sample sets of the left sub-tree and the right sub-tree, respectively, |t L | and |t R | are the sample quantities of the left sub-tree and the right sub-tree, respectively; The calculation formulas of the left and right sub-trees are respectively: .

[0035] Preferably, the prediction result of the segment prediction model is: ; wherein, is the prediction value of a certain leaf node, is the sample set contained by the leaf node, is the sum of the true target values of all samples on the leaf node.

[0036] It can be understood that the implementation of the decision tree is adopted by the machine learning algorithm. The model learning is performed on the electricity price segments, and the prediction result is input as a new feature into the subsequent model. Each leaf node has a specific prediction value.

[0037] Preferably, the probability feature includes: an extremely low value prediction probability and an extremely high value prediction probability; wherein the extremely low value prediction probability is: ; the extremely high value prediction probability is: ; wherein, is the predicted probability of the electricity price being an extreme low value, is the predicted probability of the electricity price being an extreme high value, is the simplified training parameter of the extreme low model, is the simplified training parameter of the extreme high model.

[0038] It can be understood that, for whether the electricity price is an extreme value (the normalized electricity price, whose extreme value is 0 and 1), a regression probability prediction method is used to judge whether the electricity price will be at the extreme value, and corresponding two-class 2-group 4 probability features are generated.

[0039] For extreme low value prediction

[0040] For extreme high value prediction .

[0041] Preferably, the electricity price prediction constant model is: ; wherein, is the predicted value of the initial model, the mean value of the target value set, y i is the true target value of the i th sample in the training data, and n is the number of samples in the training set, is the loss function, which is used to measure the error between the predicted value of the model and the true value, and the MSE is used, and θ is the mean value of the training data.

[0042] Preferably, the final electricity price prediction model is: ; wherein, is the final electricity price prediction result, is the learning rate, f t is the t th tree, , X i is the original feature, Z i is the supplementary feature.

[0043] It can be understood that, the new features obtained through price segmentation and extreme value probability prediction are combined with other features of the original data, input into the model for training, and gradient boosting tree is used to realize. The method of the application, multiple different implemented models, through the explanation and fitting of different angles of the target on the unified feature set, the complementary learning between the models to capture those features or feature relationships that may be ignored by a single model, the model can more flexibly cope with the fluctuation of the electricity price, and improve the processing capability of the extreme value.

[0044] First, initialize a constant model:

[0045] Then iterative update is performed, in the tth iteration, the residual of the current model is calculated

[0046] wherein, is the residual in the (t-1)th iteration, indicating the difference between the predicted value and the real value of the model in the (t-1)th iteration, is the predicted value in the (t-1)th iteration, is the partial derivative of the loss function to the predicted value of the model, indicating the gradient of the error under the current prediction of the model.

[0047] The tth tree is trained to fit the residual

[0048] In each iteration, the prediction result is updated

[0049] wherein is the learning rate.

[0050] Finally, the final prediction result is the weighted sum of all trees .

[0051] Preferably, it further comprises: The data collection days T are updated in real time, and the price interval N is dynamically adjusted according to market fluctuations.

[0052] It can be understood that in actual application, the model can update the reference window T in real time and predict future electricity prices, and dynamically adjust the price segmentation and feature selection according to market fluctuations, so as to maintain high-precision prediction ability.

[0053] The application also provides a data prediction device based on multi-angle model complementary integration, which is used to realize the above method embodiments. Figure 2 is a structural schematic diagram provided by an embodiment of the data prediction device based on multi-angle model complementary integration. As Figure 2 shown, the data prediction device based on multi-angle model complementary integration of the embodiment comprises a processor 21 and a memory 22, and the processor 21 is connected with the memory 22. Among them, the processor 21 is used to call and execute the program stored in the memory 22; the memory 22 is used to store the program, and the program is at least used to execute the data prediction method based on multi-angle model complementary integration in the above embodiment.

[0054] The specific implementation scheme of the data prediction device based on complementary integration of multi-angle models provided in the embodiments of the present application can refer to the implementation scheme of the data prediction method based on complementary integration of multi-angle models in any of the above embodiments, and will not be repeated here.

[0055] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0056] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are used for descriptive purposes only and should not be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" is at least two.

[0057] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0058] It should be understood that various components of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods may 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 may be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0059] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0060] In addition, each function unit in each embodiment of the present application can be integrated in one processing module, or each unit can be physically present separately, or two or more units can be integrated in one module. The integrated module can be realized in the form of hardware, or in the form of a software function module. When the integrated module is realized in the form of a software function module and sold or used as an independent product, it can also be stored in a computer readable storage medium.

[0061] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0062] In the description of the present specification, the description referring to the terms "one embodiment", "some embodiments", "an example", "a specific example" or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily mean the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0063] Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. A data prediction method based on complementary integration of multi-angle models, characterized in that: include: Collecting electricity price data at preset time intervals and normalizing the electricity price data to a target value; Segmenting the electricity price data after target value normalization according to a preset price range; Establishing a segmented prediction model, and training the segmented prediction model based on the segmented results of the electricity price data; The regression probability prediction method is used to determine whether the electricity price will be at these extreme values ​​and generate corresponding probability features; Generate a new electricity price feature set by combining the data of each data point in the electricity price data after the target value is normalized, the prediction result of the corresponding segmented prediction model, and the corresponding probability feature calculated by the regression probability prediction method; Obtain an initial electricity price prediction constant model; Training the electricity price prediction constant model using the new electricity price feature set and iterating a preset number of rounds, wherein at each iteration, a residual of the current model is calculated and a tree is trained to fit the residual; Update the model's predictions during iteration; The final electricity price prediction model is determined by taking the weighted sum of all trees generated during iteration, and the final electricity price prediction result is generated.

2. The method according to claim 1, characterized in that The collecting of electricity price data at preset time intervals and performing target value normalization processing on the electricity price data include: Collect electricity price data every 15 minutes over the past T days; Arrange and clean all the electricity price data, and perform target value normalization on the electricity price data using a normalization formula; Wherein, the normalization formula is: Among them, p i is the i-th value in the target value set P, p i ’ For p i The value after normalization calculation, .

3. The method according to claim 2, characterized in that The step of segmenting the electricity price data after normalizing the target value with a preset price range includes: Segmenting the electricity price data according to a preset price range N; Calculate the segment where each data point in the electricity price data is located using a segmentation formula; Wherein, the segmentation formula is: 。 4. The method according to claim 3, characterized in that The segmented prediction model is: Among them, y i is the data to be predicted, is the mean predicted value of the left subtree, is the mean predicted value of the right subtree, t L and t R are the sample sets of the left subtree and the right subtree respectively, |t L | and |t R | are the number of samples in the left subtree and the right subtree respectively; The calculation formulas for the left and right subtrees are: 。 5. The method according to claim 4, characterized in that The prediction results of the segmented prediction model are: ； in, is the predicted value of a leaf node, is the sample set contained in the leaf node, is the sum of the true target values ​​of all samples on the leaf nodes.

6. The method according to claim 5, characterized in that The probability features include: extremely low value prediction probability and extremely high value prediction probability; Among them, the extremely low value prediction probability is: ； The extremely high value prediction probability is: ； in, is the extremely low value prediction probability, For extremely high value prediction probability, is the simplified training parameter of the extremely low model, Simplified training parameters for very high models.

7. The method according to claim 6, characterized in that The electricity price prediction constant model is: ； in, is the predicted value of the initial model, y i is the true target value of the i-th sample in the training data, n is the number of samples in the training set, is the loss function, and θ is the mean of the training data.

8. The method according to claim 7, characterized in that The final electricity price prediction model is: ； in, is the final electricity price forecast result, is the learning rate, f t is the t-th tree, , X i is the original feature, Z i For supplementary features.

9. The method according to claim 8, characterized in that Also includes: The data collection days T are updated in real time, and the price range N is dynamically adjusted according to market fluctuations.

10. A data prediction device based on complementary integration of multi-angle models, characterized in that: The device comprises a processor and a memory, wherein the processor is connected to the memory: The processor is configured to call and execute the program stored in the memory; The memory is used to store the program, and the program is at least used to execute the data prediction method based on complementary integration of multi-angle models according to any one of claims 1 to 9.

Images