Method, device and equipment for evaluating contribution degree of replacement of potential user by electric energy and storage medium

Abstract

The invention relates to the technical field of electric energy substitution, and discloses an electric energy substitution potential user contribution degree evaluation method and device, equipment and a storage medium, which are used for solving the technical problem of insufficient reliability of electric energy substitution potential user contribution degree evaluation. The method comprises the following steps: acquiring view data of multiple types of samples of electric energy replacement potential users under multiple views; inputting view data in the aspects of power consumption behaviors, industry attributes, economic potential and equipment conditions into a pre-trained cross-view affinity and rejection coding model, and outputting hidden codes; inputting the implicit codes generated by the view data into a pre-trained discriminator, and calculating a reliability score of each view data; calculating a multi-view reliability score corresponding to each sample according to the reliability score; inputting the implicit codes into a pre-training classifier to obtain a preliminary classification result; and calculating contribution degree classification of the electric energy replacing potential users by adopting the preliminary classification result and the multi-view reliability score.

Description

Technical Field

[0001] This invention relates to the field of electricity substitution technology, and in particular to a method, apparatus, equipment, and storage medium for evaluating the contribution of potential users to electricity substitution. Background Technology

[0002] In the electricity substitution business, accurately assessing the contribution of potential users is crucial for power grid companies to formulate promotion strategies and optimize resource allocation. Data on potential electricity substitution users is multi-source, capable of characterizing user features from various dimensions such as electricity consumption, economic value, policy compatibility, and operational stability. However, electricity substitution data faces numerous challenges. On the one hand, the data is susceptible to noise interference; errors in enterprise data entry and temporary policy changes can affect the accuracy of the assessment. On the other hand, traditional multi-view classification methods simply integrate multi-view information without robust noise handling mechanisms, failing to provide reliable assessment results and thus failing to meet the requirements of high-reliability decision-making. Single-view reliability classification methods, utilizing only single-dimensional data, cannot fully leverage the complementarity of multi-view data, limiting classification accuracy, and the accuracy of reliability assessment decreases when single-dimensional data is affected by noise. Some multi-view electricity user reliability classification studies lack robustness to data noise, have complex model structures and calculation processes, and are not conducive to practical deployment.

[0003] Traditional multi-view electricity user classification methods integrate multi-view data using early, late, or intermediate fusion strategies to achieve user classification. However, they lack a dedicated mechanism for handling data noise in electricity substitution scenarios, leading to significant bias in evaluation results and compromised reliability when the data is noisy. Single-view reliable electricity user classification methods focus on a single dimension of data, evaluating the reliability of classification results within that dimension using relevant methods. However, they cannot fully integrate the complementarity of multi-view data, and the accuracy of reliability assessment drops sharply when the single-dimensional data is affected by noise. Other multi-view reliable electricity user classification methods attempt to introduce reliability assessment and design dedicated fusion strategies, but they lack robustness to data noise in electricity substitution scenarios, and their complex models are not conducive to practical deployment. Summary of the Invention

[0004] This invention provides a method, apparatus, device, and storage medium for assessing the contribution of potential users of electricity substitution, in order to solve the technical problem of insufficient reliability in assessing the contribution of potential users of electricity substitution.

[0005] This invention provides a method for evaluating the contribution of potential users of electricity substitution, comprising:

[0006] Acquire view data of multiple types of samples from potential users of electricity substitution across multiple views;

[0007] The view data is input into a pre-trained cross-view affinity and repulsion coding model, which outputs a hidden code.

[0008] The hidden code is input into a pre-trained discriminator to calculate a reliability score for each view data.

[0009] Calculate the multi-view reliability score for each sample based on the reliability score;

[0010] The hidden code is input into a pre-trained classifier to obtain preliminary classification results;

[0011] The contribution classification of potential users of electricity substitution is calculated using the preliminary classification results and the multi-view reliability score.

[0012] Optionally, the training process of the cross-view affinity and repulsion coding model includes:

[0013] Acquire training view data of multiple types of training samples under multiple views, and encode the training view data into an initial hidden code;

[0014] The cross-view affinity and exclusion coding model is trained based on the initial implicit coding and the preset cross-view affinity loss function and intra-view exclusion learning loss function.

[0015] Optionally, the step of training the cross-view affinity and repulsion coding model based on the initial implicit coding and the preset cross-view affinity loss function and intra-view repulsion learning loss function includes:

[0016] Calculate the first KL divergence of the same training sample under different views based on the initial hidden code of the same training sample under different views;

[0017] Calculate the cross-view affinity loss function value based on the first KL divergence;

[0018] Calculate the second KL divergence of different samples under the same view based on the initial hidden codes of different training samples under the same view;

[0019] Calculate the in-view repulsion learning loss function value based on the second KL divergence;

[0020] The initial cross-view affinity and rejection coding model is optimized based on the cross-view affinity loss function value and the intra-view rejection learning loss function value to obtain the trained cross-view affinity and rejection coding model.

[0021] Optionally, each type of sample has a corresponding prototype queue; the training process of the discriminator includes:

[0022] The hidden code of each type of sample is used as a negative sample, and the number of such negative samples is recorded.

[0023] The corresponding number of positive samples are obtained from the prototype queue based on the number of negative samples.

[0024] The initial discriminator is trained using the positive and negative samples, and the discriminator loss function value is calculated.

[0025] The initial discriminator is adjusted based on the discriminator loss function value to obtain the adjusted discriminator;

[0026] Calculate the mean of the hidden code, generate new elements in the prototype queue based on the mean, and return to the step of obtaining a corresponding number of positive samples from the prototype queue based on the number of negative samples, until the training of the discriminator is completed.

[0027] The present invention also provides a device for assessing the contribution of potential users of electricity substitution, comprising:

[0028] The view data acquisition module is used to acquire view data of multiple types of samples from potential users of electricity substitution under multiple views;

[0029] The hidden coding output module is used to input the view data into a pre-trained cross-view affinity and repulsion coding model and output the hidden coding.

[0030] The reliability score calculation module is used to input the hidden code into a pre-trained discriminator and calculate the reliability score for each view data.

[0031] The multi-view reliability score calculation module is used to calculate the multi-view reliability score corresponding to each sample based on the reliability score.

[0032] The preliminary classification module is used to input the hidden code into a pre-trained classifier to obtain a preliminary classification result;

[0033] The contribution classification module is used to calculate the contribution classification of the potential users of electricity substitution using the preliminary classification results and the multi-view reliability score.

[0034] Optionally, the training process of the cross-view affinity and repulsion coding model includes:

[0035] The encoding module is used to acquire training view data of multiple training samples under multiple views and encode the training view data into an initial hidden code.

[0036] The cross-view affinity and exclusion coding model training module is used to train the cross-view affinity and exclusion coding model based on the initial hidden coding and the preset cross-view affinity loss function and intra-view exclusion learning loss function.

[0037] Optionally, the cross-view affinity and repulsion coding model training module includes:

[0038] The first KL divergence calculation submodule is used to calculate the first KL divergence of the same training sample under different views based on the initial hidden code of the same training sample under different views.

[0039] A cross-view affinity loss function value calculation submodule is used to calculate the cross-view affinity loss function value based on the first KL divergence;

[0040] The second KL divergence calculation module is used to calculate the second KL divergence of different samples under the same view based on the initial hidden coding of different training samples under the same view.

[0041] The in-view repulsion learning loss function value calculation module is used to calculate the in-view repulsion learning loss function value based on the second KL divergence.

[0042] The optimization module is used to optimize the initial cross-view affinity and rejection coding model based on the cross-view affinity loss function value and the intra-view rejection learning loss function value, so as to obtain the trained cross-view affinity and rejection coding model.

[0043] Optionally, each type of sample has a corresponding prototype queue; the training process of the discriminator includes:

[0044] The negative sample determination module is used to identify the hidden code of each type of sample as a negative sample and record the number of negative samples.

[0045] A positive sample acquisition module is used to acquire a corresponding number of positive samples from the prototype queue based on the number of negative samples;

[0046] The discriminator loss function value calculation module is used to train the initial discriminator using the positive samples and the negative samples, and to calculate the discriminator loss function value;

[0047] An adjustment module is used to adjust the initial discriminator based on the discriminator loss function value to obtain an adjusted discriminator;

[0048] The training module is used to calculate the mean of the hidden code, generate new elements of the prototype queue based on the mean, and return the step of obtaining a corresponding number of positive samples from the prototype queue based on the number of negative samples, until the training of the discriminator is completed.

[0049] The present invention also provides an electronic device, the device comprising a processor and a memory:

[0050] The memory is used to store program code and transmit the program code to the processor;

[0051] The processor is configured to execute, according to instructions in the program code, the method for assessing the contribution of potential users of electricity substitution as described above.

[0052] The present invention also provides a computer-readable storage medium for storing program code for executing the method for assessing the contribution of potential users of electricity substitution as described in any of the preceding claims.

[0053] As can be seen from the above technical solutions, the present invention has the following advantages: The present invention provides a method for evaluating the contribution of potential users of electricity substitution, and specifically discloses: acquiring view data of multiple types of samples of potential users of electricity substitution under multiple views; inputting the view data into a pre-trained cross-view affinity and exclusion coding model to output hidden codes; inputting the hidden codes into a pre-trained discriminator to calculate the reliability score of each view data; calculating the multi-view reliability score corresponding to each type of sample based on the reliability score; inputting the hidden codes into a pre-trained classifier to obtain preliminary classification results; and using the preliminary classification results and multi-view reliability scores to calculate the contribution classification of potential users of electricity substitution.

[0054] This invention utilizes multi-view affinity and exclusion coding to learn cross-view... Figure 1 High-quality features of consistency and intra-view distinguishability, including cross-view Figure 1 Consistency is reflected in the fact that users with high electricity consumption and high economic value should belong to the high contribution category. Intra-view distinguishability is reflected in the significant differences in characteristics between high, medium and low contribution users in the same view. With the help of a reliable perception discriminator, noise samples in the electricity substitution data are effectively identified and a reliability score is output. Noise samples include enterprises with abnormal data entry. Then, through reliable multi-view classification, the multi-view classification results are weighted and fused based on the reliability score. This improves the robustness to noise and enables an effective assessment of the reliability of the contribution evaluation results, providing more reliable support for the power grid enterprise's electricity substitution business decision-making. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 A flowchart illustrating the steps of a method for evaluating the contribution of potential users of electricity substitution, as provided in an embodiment of the present invention;

[0057] Figure 2 This is a structural block diagram of a device for evaluating the contribution of potential users of electricity substitution, provided in an embodiment of the present invention. Detailed Implementation

[0058] This invention provides a method, apparatus, device, and storage medium for assessing the contribution of potential users of electricity substitution, in order to solve the technical problem of insufficient reliability in assessing the contribution of potential users of electricity substitution.

[0059] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0060] Please see Figure 1 , Figure 1 A flowchart illustrating the steps of a method for evaluating the contribution of potential users of electricity substitution, provided in an embodiment of the present invention.

[0061] The present invention provides a method for evaluating the contribution of potential users of electricity substitution, which may specifically include the following steps:

[0062] Step 101: Obtain view data of multiple types of samples from potential users of electricity substitution under multiple views;

[0063] In this embodiment of the invention, view data of multiple types of samples from potential users of electricity substitution can be obtained under multiple views.

[0064] The sample types can include three categories of user samples with high, medium and low contribution levels, and the view data can include electricity consumption behavior, industry attributes, economic potential and equipment status.

[0065] Step 102: Input the view data into the pre-trained cross-view affinity and repulsion coding model and output the hidden code;

[0066] After obtaining the view data, the view data can be input into a pre-trained cross-view affinity and repulsion coding model to output the hidden code.

[0067] In one example, a cross-view affinity and repulsion coding model may include a cross-view affinity module and an intra-view repulsion learning module. The training process for a cross-view affinity and repulsion coding model may include the following sub-steps:

[0068] S21, acquire training view data of multiple training samples under multiple views, and encode the training view data into the initial hidden code;

[0069] S22, calculate the first KL divergence of the same training sample in different views based on the initial hidden coding and the preset cross-view affinity loss function and intra-view exclusion learning loss function according to the initial hidden coding of the same training sample in different views;

[0070] S23, calculate the cross-view affinity loss function value based on the first KL divergence;

[0071] S24, calculate the second KL divergence of different training samples under the same view based on the initial hidden coding of different training samples under the same view;

[0072] S25, calculate the in-view repulsion learning loss function value based on the second KL divergence;

[0073] S26. Optimize the initial cross-view affinity and exclusion coding model based on the cross-view affinity loss function value and the intra-view exclusion learning loss function value to obtain the trained cross-view affinity and exclusion coding model.

[0074] In the specific implementation, given the content One view, Multi-view data of a sample ,in, Representing the The sample at the th View data under each view. Introducing a multi-view encoder. , data Encoded as implicit encoding . Indicates the first The first view The implicit encoding set of class data. Considering the similarity of samples within the same class, we assume... Follows a Gaussian distribution ,in and Let be the mean and standard deviation of the Gaussian distribution, respectively. Then, from the The hidden encoding of a sample class can be represented as:

[0075]

[0076] in, Indicates the first The first in the class The initial hidden coding of each sample, For the first Class center of the class, Reflecting the Variance and uncertainty of implicit coding.

[0077] To fully utilize the complementary information across multiple views, it is assumed that the latent coding distribution of samples of the same class is consistent across different views. Therefore, this invention proposes a cross-view affinity enhancement module to fully exploit the consistency information across multiple views. Furthermore, within a single view, the latent coding distributions of different sample classes differ. Based on this, an intra-view repulsion learning module is proposed, and joint optimization of these two modules can improve the effectiveness of latent coding.

[0078] Cross-view affinity enhancement module: To make the latent codes of the same type of samples closer to each other in different views, KL divergence (KL divergence) is used as a metric to measure the distance between two distributions. By reducing the KL divergence, the two distributions can be made closer together. The loss function of the cross-view affinity enhancement module is defined as follows:

[0079]

[0080] in, Indicates the number of categories. It is used to measure the distance between two Gaussian distributions. The formula for calculation is as follows:

[0081]

[0082] in, and These represent two Gaussian distributions. and They are respectively and The dimensional components, and They are respectively and The Dimensional components.

[0083] In-view repulsion learning module: To increase the distance between different class latent codes and improve their discriminative power, this invention expands the distance between two different class distributions by increasing their KL divergence. The loss function of the in-view repulsion learning module is defined as follows:

[0084]

[0085] in, and Each represents the same view Next category and categories The latent coding distribution. The loss function of the multi-view encoder is obtained by integrating the cross-view affinity enhancement module and the intra-view repulsion learning module, and the specific formula is as follows:

[0086]

[0087] Step 103: Input the hidden code into the pre-trained discriminator and calculate the reliability score for each view data.

[0088] In this embodiment of the invention, the hidden code can be input into a pre-trained discriminator to calculate the reliability score of each view data.

[0089] In one example, the discriminator training process is as follows:

[0090] S31, use the hidden code of each class of samples as a negative sample, and record the number of negative samples;

[0091] S32, obtain the corresponding number of positive samples from the prototype queue based on the number of negative samples;

[0092] S33, use positive and negative samples to train the initial discriminator and calculate the discriminator loss function value;

[0093] S34, Adjust the initial discriminator according to the discriminator loss function value to obtain the adjusted discriminator;

[0094] S35, calculate the mean of the hidden code, generate new elements in the prototype queue based on the mean, and return the step of obtaining the corresponding number of positive samples from the prototype queue based on the number of negative samples, until the discriminator converges and the training of the discriminator is completed.

[0095] The prototype queue is a dedicated dataset built for potential users of electricity substitution based on their "contribution categories"—a separate queue is maintained for each contribution category (such as high contribution, medium contribution, and low contribution). Each queue corresponds to a contribution level, and the queue stores the "typical characteristics" of users at that level, which are used to determine whether new samples conform to the characteristic patterns of that category.

[0096] In the scenario of assessing the contribution of potential users to electricity substitution, noise contained in multi-view data can significantly affect the assessment results. To effectively assess this type of noise and predict the reliability of the contribution assessment results, a reliable perception discriminator adapted to this scenario was designed. Its core function is to output lower reliability scores for samples with high noise levels and higher reliability scores for samples with low noise levels. Appropriate positive and negative samples are constructed to train the discriminator. This is the key to the implementation of this module. The distribution of the known implicit coding satisfies... The mean of this distribution As a positive sample, the implicit encoding of multi-view data As negative samples, positive samples will receive higher reliability scores, while negative samples will receive lower reliability scores.

[0097] However, directly using the distribution mean Using positive samples as positive samples is unsuitable because their features change during discriminator training, causing the discriminator to fail to accurately capture their characteristics. To address this, a momentum strategy can be employed to record the mean of positive samples in each iteration, thus providing more stable positive samples with smaller variance. During training, a prototype queue is maintained for each class. ( ,in (Number of categories). Hidden encoding mean It is calculated by averaging the hidden codes of all samples in that category during a given iteration. Assume... For the current prototype queue The last element is used to generate a new element in the queue. :

[0098]

[0099] in, As momentum weights, the newly generated queue elements are then... Insert into the prototype queue The back of the line.

[0100] Given positive samples With negative samples The loss function of the reliable perception discriminator is defined as follows:

[0101]

[0102] in, It uses a fully connected neural network and employs the sigmoid function to predict reliable scores. It's worth noting that in the current batch... Class contains To balance the number of positive and negative samples during discriminator training, a prototype queue of the corresponding category can be used. Random selection One sample is used as a positive sample.

[0103] During training, the discriminator's loss function needs to be continuously monitored. When the loss function value fluctuates sufficiently over multiple training epochs and no longer decreases significantly, it indicates that the discriminator has converged, and training can be considered complete.

[0104] Step 104: Calculate the multi-view reliability score for each sample based on the reliability score;

[0105] After the discriminator is trained, it can be used... Calculating the reliability score for each sample in each view reveals that the reliability score for each sample is different across all views. To obtain the reliability score for each multi-view sample, this invention averages the reliability scores of all views for a sample to obtain the multi-view reliability score. .

[0106]

[0107] Step 105: Input the hidden code into the pre-trained classifier to obtain the preliminary classification result;

[0108] In this embodiment of the invention, the pre-trained classifier can use the cross-entropy loss function to classify the latent code of each sample, thus classifying the latent code of multi-view data. Input pre-trained classifier The classification results are denoted as The loss function for the pre-trained classifier is as follows:

[0109]

[0110] Step 106: Calculate the contribution classification of potential users of electricity substitution using the preliminary classification results and multi-view reliability scores.

[0111] Because the view data of potential customers may contain noise, noise can significantly affect the classification results. However, reliable scores can serve as a valid guide for multi-view fusion, helping to obtain more accurate classification results. When the quality of a view is poor, its reliability score will also decrease, thus reducing the view's impact on the final decision. The contribution classification of potential users of electricity substitution can be achieved in the following way:

[0112]

[0113] It should be noted that, in embodiments of the present invention, the above-mentioned loss functions can also be integrated into a framework, and the entire system can be jointly optimized through the overall loss function, which is as follows:

[0114]

[0115] in, and These are the weight parameters.

[0116] This invention utilizes multi-view affinity and exclusion coding to learn cross-view... Figure 1 High-quality features of consistency and intra-view distinguishability, including cross-view Figure 1Consistency is reflected in the fact that users with high electricity consumption and high economic value should belong to the high contribution category. Intra-view distinguishability is reflected in the significant differences in characteristics between high, medium and low contribution users in the same view. With the help of a reliable perception discriminator, noise samples in the electricity substitution data are effectively identified and a reliability score is output. Noise samples include enterprises with abnormal data entry. Then, through reliable multi-view classification, the multi-view classification results are weighted and fused based on the reliability score. This improves the robustness to noise and enables an effective assessment of the reliability of the contribution evaluation results, providing more reliable support for the power grid enterprise's electricity substitution business decision-making.

[0117] Please see Figure 2 , Figure 2 This is a structural block diagram of a device for evaluating the contribution of potential users of electricity substitution, provided in an embodiment of the present invention.

[0118] This invention provides a device for assessing the contribution of potential users of electricity substitution, comprising:

[0119] The view data acquisition module 201 is used to acquire view data of multiple types of samples of potential users of electric energy substitution under multiple views;

[0120] The latent coding output module 202 is used to input view data into a pre-trained cross-view affinity and repulsion coding model and output latent codes.

[0121] The reliability score calculation module 203 is used to input the hidden code into a pre-trained discriminator and calculate the reliability score for each view data.

[0122] The multi-view reliability score calculation module 204 is used to calculate the multi-view reliability score corresponding to each sample based on the reliability score.

[0123] The preliminary classification module 205 is used to input the hidden code into the pre-trained classifier to obtain the preliminary classification result;

[0124] The contribution classification module 206 is used to calculate the contribution classification of potential users of electricity substitution using preliminary classification results and multi-view reliability scores.

[0125] In this embodiment of the invention, the training process of the cross-view affinity and exclusion coding model includes:

[0126] The encoding module is used to acquire training view data of multiple classes of training samples under multiple views and encode the training view data into the initial hidden code.

[0127] The cross-view affinity and exclusion coding model training module is used to train the cross-view affinity and exclusion coding model based on the initial implicit coding and the preset cross-view affinity loss function and intra-view exclusion learning loss function.

[0128] In this embodiment of the invention, the cross-view affinity and exclusion coding model training module includes:

[0129] The first KL divergence calculation submodule is used to calculate the first KL divergence of the same training sample under different views based on the initial hidden code of the same training sample under different views.

[0130] The cross-view affinity loss function value calculation submodule is used to calculate the cross-view affinity loss function value based on the first KL divergence;

[0131] The second KL divergence calculation module is used to calculate the second KL divergence of different samples under the same view based on the initial hidden coding of different training samples under the same view.

[0132] The in-view repulsion learning loss function value calculation module is used to calculate the in-view repulsion learning loss function value based on the second KL divergence.

[0133] The optimization module is used to optimize the initial cross-view affinity and rejection coding model based on the cross-view affinity loss function value and the intra-view rejection learning loss function value, so as to obtain the trained cross-view affinity and rejection coding model.

[0134] In this embodiment of the invention, each type of sample has a corresponding prototype queue; the training process of the discriminator includes:

[0135] The negative sample determination module is used to identify the hidden code of each class of samples as a negative sample and record the number of negative samples.

[0136] The positive sample acquisition module is used to acquire a corresponding number of positive samples from the prototype queue based on the number of negative samples.

[0137] The discriminator loss function value calculation module is used to train the initial discriminator using positive and negative samples and to calculate the discriminator loss function value.

[0138] The adjustment module is used to adjust the initial discriminator based on the discriminator loss function value to obtain the adjusted discriminator.

[0139] The training module is used to calculate the mean of the hidden code, generate new elements in the prototype queue based on the mean, and return the steps to obtain the corresponding number of positive samples from the prototype queue based on the number of negative samples, until the training of the discriminator is completed.

[0140] This invention also provides an electronic device, which includes a processor and a memory:

[0141] The memory is used to store program code and transfer the program code to the processor;

[0142] The processor is used to execute the method for evaluating the contribution of potential users of electricity substitution according to the instructions in the program code.

[0143] This invention also provides a computer-readable storage medium for storing program code for executing the method for evaluating the contribution of potential users of electricity substitution according to this invention.

[0144] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0145] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0146] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0147] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0148] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0149] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0150] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

[0151] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0152] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0153] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An electric energy alternative potential user contribution degree evaluation method, characterized by, The method comprises the following steps: Obtain view data of multiple types of samples of potential users of electric energy replacement under multiple views; Input the view data into a pre-trained cross-view affinity and repulsion encoding model to output hidden encoding; Input the hidden encoding into a pre-trained discriminator to calculate a reliability score of each view data; Calculate a multi-view reliability score corresponding to each sample according to the reliability score; Input the hidden encoding into a pre-trained classifier to obtain a preliminary classification result; Calculate the contribution degree classification of the potential users of electric energy replacement by using the preliminary classification result and the multi-view reliability score.

2. The method of claim 1, wherein, The training process of the cross-view affinity and repulsion encoding model comprises the following steps: Obtain training view data of multiple types of training samples under multiple views, and encode the training view data into initial hidden encoding; Train the cross-view affinity and repulsion encoding model according to the initial hidden encoding and a preset cross-view affinity loss function and an intra-view repulsion learning loss function.

3. The method of claim 2, wherein, The step of training the cross-view affinity and repulsion encoding model according to the initial hidden encoding and the preset cross-view affinity loss function and the intra-view repulsion learning loss function comprises the following steps: Calculate a first KL divergence of the same training sample under different views according to the initial hidden encoding of the same training sample under different views; Calculate a cross-view affinity loss function value according to the first KL divergence; Calculate a second KL divergence of different samples under the same view according to the initial hidden encoding of different training samples under the same view; Calculate an intra-view repulsion learning loss function value according to the second KL divergence; Optimize the initial cross-view affinity and repulsion encoding model according to the cross-view affinity loss function value and the intra-view repulsion learning loss function value to obtain a trained cross-view affinity and repulsion encoding model.

4. The method of claim 2, wherein, Each type of sample has a corresponding prototype queue; the training process of the discriminator comprises the following steps: Take the hidden encoding of each type of sample as a negative sample, and record the number of the negative sample; Obtain a corresponding number of positive samples from the prototype queue according to the number of the negative sample; Train an initial discriminator by using the positive sample and the negative sample, and calculate a discriminator loss function value; Adjust the initial discriminator according to the discriminator loss function value to obtain an adjusted discriminator; Calculate the mean value of the hidden encoding, generate a new element of the prototype queue according to the mean value, and return to the step of obtaining a corresponding number of positive samples from the prototype queue according to the number of the negative sample until the discriminator converges, and the training of the discriminator is completed.

5. An electric power alternative potential user contribution degree evaluation device characterized by comprising: The method comprises the following steps: A view data acquisition module is configured to obtain view data of multiple types of samples of potential users of electric energy replacement under multiple views; A hidden encoding output module is configured to input the view data into a pre-trained cross-view affinity and repulsion encoding model to output hidden encoding; A reliability score calculation module is configured to input the hidden encoding into a pre-trained discriminator to calculate a reliability score of each view data; A multi-view reliability score calculation module is configured to calculate a multi-view reliability score corresponding to each sample according to the reliability score; A preliminary classification module is configured to input the hidden encoding into a pre-trained classifier to obtain a preliminary classification result; The contribution degree classification module is configured to calculate the contribution degree of the potential user of the electric energy substitution by using the preliminary classification result and the multi-view reliability score.

6. The apparatus of claim 5, wherein, The training process of the cross-view affinity and repulsion encoding model comprises the following steps: An encoding module is configured to acquire training view data of multi-class training samples in multiple views and encode the training view data into initial hidden encodings. A cross-view affinity and repulsion encoding model training module is configured to train a cross-view affinity and repulsion encoding model according to the initial hidden encodings and a preset cross-view affinity loss function and an intra-view repulsion learning loss function.

7. The apparatus of claim 5, wherein, The cross-view affinity and repulsion encoding model training module comprises: A first KL divergence calculation submodule is configured to calculate a first KL divergence of the same training sample in different views according to the initial hidden encodings of the same training sample in different views. A cross-view affinity loss function value calculation submodule is configured to calculate a cross-view affinity loss function value according to the first KL divergence. A second KL divergence calculation submodule is configured to calculate a second KL divergence of different samples in the same view according to the initial hidden encodings of different training samples in the same view. An intra-view repulsion learning loss function value calculation submodule is configured to calculate an intra-view repulsion learning loss function value according to the second KL divergence. An optimization module is configured to optimize the initial cross-view affinity and repulsion encoding model according to the cross-view affinity loss function value and the intra-view repulsion learning loss function value to obtain a trained cross-view affinity and repulsion encoding model.

8. The apparatus of claim 6, wherein, Each class of samples has a corresponding prototype queue; and the training process of the discriminator comprises the following steps: A negative sample determination module is configured to take the hidden encoding of each class of samples as a negative sample and record the number of the negative samples. A positive sample acquisition module is configured to acquire a corresponding number of positive samples from the prototype queue according to the number of the negative samples. A discriminator loss function value calculation module is configured to train an initial discriminator by using the positive samples and the negative samples and calculate a discriminator loss function value. An adjustment module is configured to adjust the initial discriminator according to the discriminator loss function value to obtain an adjusted discriminator. A training module is configured to calculate the mean value of the hidden encodings, generate a new element of the prototype queue according to the mean value, and return to the step of acquiring a corresponding number of positive samples from the prototype queue according to the number of the negative samples until the training of the discriminator is completed.

9. An electronic device, comprising: The device comprises a processor and a memory: The memory is configured to store program code and transmit the program code to the processor. The processor is configured to execute the method for evaluating the contribution degree of the potential user of the electric energy substitution according to the instructions in the program code.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium is configured to store program code, and the program code is configured to execute the method for evaluating the contribution degree of the potential user of the electric energy substitution.

Images