CN116894163A - Method and device for generating load prediction information for charging and discharging facilities based on information security - Google Patents

Abstract

The embodiment of the invention discloses a charge and discharge facility load prediction information generation method and device based on information security. One embodiment of the method comprises the following steps: collecting a target specimen local data set sequence; initial model training is carried out on the initial charge load prediction information generation model through the target specimen data set; splitting the model files of the trained model file set to obtain a trained model file set; performing primary model file aggregation on the trained model file group to obtain a first aggregated model file; performing secondary model file aggregation on the first aggregated model file set to obtain a second aggregated model file; and generating a model according to the real-time operation data and the charging load prediction information, and generating the charging load prediction information. According to the embodiment, the problem that when a large number of new energy automobiles to be stored exist at the same time, the power grid corresponding to the energy storage station is overloaded, so that unstable voltage is caused, and damage to vehicle charging equipment is possibly caused is avoided.

Description

Method and device for generating load prediction information for charging and discharging facilities based on information security

Technical field

Embodiments of the present disclosure relate to the field of computer technology, and specifically to methods and devices for generating load prediction information for charging and discharging facilities based on information security.

Background technique

With the development and popularization of new energy vehicles, how to improve the energy storage flexibility and convenience of new energy vehicles is particularly important for the rapidly increasing number of new energy vehicles. At present, the usual method for energy storage of new energy vehicles is to set up energy storage sites for new energy vehicles, and store energy for new energy vehicles through vehicle charging equipment set up in the energy storage sites.

However, the inventor found that when the above method is adopted, the following technical problems often exist:

First, when there are a large number of new energy vehicles to be stored at the same time, it will cause the power grid corresponding to the energy storage site to be overloaded, resulting in voltage instability and possible damage to vehicle charging equipment;

Second, due to the large difference in the data volume of the local data group corresponding to each local terminal, when the data volume of the local data group is small, the prediction of the local charging load prediction information generation model based on the local data group training will be accurate. The accuracy is poor. When there is a large difference between the predicted charging load prediction information and the actual charging load, it may cause damage to the vehicle charging equipment;

Third, when the number of network layers of the traditional linearly connected convolutional neural network is deepened, it may cause the problem of feature forgetting, thereby affecting the accuracy of the generated charging load prediction information. When the predicted charging load prediction information is different from the actual charging Large differences between loads may cause damage to the vehicle's charging equipment.

The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art.

Contents of the invention

This Summary is provided to introduce in simplified form concepts that are later described in detail in the Detailed Description. The content of this disclosure is not intended to identify key features or essential features of the claimed technical solutions, nor is it intended to be used to limit the scope of the claimed technical solutions.

Some embodiments of the present disclosure propose methods and devices for generating load prediction information for charging and discharging facilities based on information security to solve one or more of the technical problems mentioned in the background art section above.

In a first aspect, some embodiments of the present disclosure provide a method for generating charging and discharging facility load prediction information based on information security. The method includes: collecting a target local data group sequence, wherein the target local data group sequence is at least one The local data group corresponding to the local terminal, where the terminal type of the local terminal is the vehicle charging equipment type, and the target local data is the operating data generated by the local terminal when charging the vehicle; for each target local data group in the above sequence data group, through the above-mentioned target local data group, perform initial model training on the initial charging load prediction information generation model corresponding to the initial model file, and obtain a post-training model file, wherein the above-mentioned initial model file is a local data group corresponding to the above-mentioned target local data group. The model file sent by the edge server associated with the terminal, in which the initial charging load prediction information generation model is the charging load prediction information generation model to be model trained; the obtained set of trained model files is split into model files to obtain the A set of model file groups, wherein the above-mentioned set of trained model file groups corresponds to at least one edge server; for each post-trained model file group in the above-mentioned set of post-trained model file groups, the edge server corresponding to the above-mentioned post-trained model file group is The server performs a model file aggregation on the above-mentioned trained model file group to obtain a first aggregation model file; performs a second model file aggregation on the obtained first aggregation model file set to obtain a second aggregation model file; for at least the above Each local terminal in a local terminal generates a charging load prediction information corresponding to the local terminal based on the real-time operating data corresponding to the local terminal and the charging load prediction information generation model corresponding to the second aggregated model file.

In the second aspect, some embodiments of the present disclosure provide a device for generating charging and discharging facility load prediction information based on information security. The device includes: a collection unit configured to collect a target local data group sequence, wherein the above target local data The group sequence is a local data group corresponding to at least one local terminal, where the terminal type of the local terminal is the vehicle charging equipment type, and the target local data is the operating data generated by the local terminal when charging the vehicle; the model training unit is configured to Each target local data group in the above-mentioned target local data group sequence performs initial model training on the initial charging load prediction information generation model corresponding to the initial model file through the above-mentioned target local data group, and obtains a post-training model file, wherein the above-mentioned initial The model file is a model file sent by the edge server associated with the local terminal corresponding to the above-mentioned target local data group, in which the initial charging load prediction information generation model is the charging load prediction information generation model to be model trained; the split unit is Configured to perform model file splitting on the obtained post-training model file set to obtain a post-training model file group set, wherein the above-mentioned post-training model file group set corresponds to at least one edge server; the first aggregation unit is configured to perform the above-mentioned Each post-training model file group in the set of post-training model file groups performs a model file aggregation on the above-mentioned post-training model file group through the edge server corresponding to the above-mentioned post-training model file group to obtain a first aggregated model file; The second aggregation unit is configured to perform secondary model file aggregation on the obtained first aggregated model file set to obtain the second aggregated model file; the generating unit is configured to perform a second aggregation of model files for each local terminal in the at least one local terminal. , based on the real-time operating data corresponding to the local terminal and the charging load prediction information generation model corresponding to the second aggregated model file, the charging load prediction information corresponding to the local terminal is generated.

In a third aspect, some embodiments of the present disclosure provide an electronic device, including: one or more processors; a storage device on which one or more programs are stored. When one or more programs are processed by one or more The processor executes, causing one or more processors to implement the method described in any implementation manner of the first aspect.

In a fourth aspect, some embodiments of the present disclosure provide a computer-readable medium on which a computer program is stored, wherein when the program is executed by a processor, the method described in any implementation manner of the first aspect is implemented.

The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the charging and discharging facility load prediction information generation method based on information security of some embodiments of the present disclosure, it is avoided that when there are a large number of new energy vehicles to be stored at the same time, causing the storage The power grid corresponding to the power station is overloaded, causing voltage instability and possibly causing damage to vehicle charging equipment. Specifically, the reason for damage to vehicle charging equipment is that the presence of a large number of new energy vehicles to be stored at the same time may cause the grid load corresponding to the energy storage site to increase rapidly, causing the grid corresponding to the energy storage site to be overloaded, which may cause damage to the vehicle. Damage to charging equipment. Based on this, some embodiments of the present disclosure provide a method for generating load prediction information for charging and discharging facilities based on information security. First, a target local data group sequence is collected, wherein the target local data group sequence is a local data group corresponding to at least one local terminal. , where the terminal type of the local terminal is the vehicle charging equipment type, and the target local data is the operating data generated by the local terminal when charging the vehicle. Secondly, for each target local data group in the above target local data group sequence, perform initial model training on the initial charging load prediction information generation model corresponding to the initial model file through the above target local data group, and obtain the trained model file, where , the above-mentioned initial model file is a model file sent by the edge server associated with the local terminal corresponding to the above-mentioned target local data group, wherein the initial charging load prediction information generation model is the charging load prediction information generation model to be model trained. By using each target local data group in the above target local data group sequence as a training sample and performing initial model training on the initial charging load prediction information generation model corresponding to the initial model file, a personalized model corresponding to the target local data group can be generated. local model. At the same time, for each initial charging load prediction information generation model, the entire model training process only involves the target local data group, and the local data does not go out of the domain (that is, different target local data groups are isolated from each other), achieving protection in Model generation under the premise of data privacy. Next, the obtained set of trained model files is split into model files to obtain a set of post-trained model file groups, where the above set of post-trained model file groups corresponds to at least one edge server. In practice, different trained model files often correspond to different edge servers, and the corresponding edge servers need to aggregate model files. Therefore, the trained model files need to be split into model files. Further, for each post-training model file group in the above-mentioned post-training model file group set, perform a model file aggregation on the above-mentioned post-training model file group through the edge server corresponding to the above-mentioned post-training model file group to obtain the first aggregation Post model file. Through model file aggregation, data fusion and sharing can be achieved indirectly when data are isolated from each other, improving the robustness of the model while protecting the privacy of the data. In addition, secondary model file aggregation is performed on the obtained first aggregated model file set to obtain a second aggregated model file. Through hierarchical aggregation, computing costs can be reduced while protecting data privacy. Finally, for each local terminal in the at least one local terminal, a model is generated based on the real-time operating data corresponding to the local terminal and the charging load prediction information corresponding to the second aggregated model file to generate a charging load prediction corresponding to the local terminal. information. In this way, accurate charging load prediction information can be generated, and the charging load can be adjusted by combining the charging load prediction information. This can avoid the voltage caused by the overload of the power grid corresponding to the energy storage site when there are a large number of new energy vehicles to be stored. instability problem, thereby reducing the probability of vehicle charging equipment being damaged.

Description of the drawings

The above and other features, advantages, and aspects of various embodiments of the present disclosure will become more apparent with reference to the following detailed description taken in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numbers refer to the same or similar elements. It should be understood that the drawings are schematic and that elements and elements are not necessarily drawn to scale.

Figure 1 is a flow chart of some embodiments of a method for generating load prediction information for charging and discharging facilities based on information security according to the present disclosure;

Figure 2 is a schematic structural diagram of some embodiments of an information security-based charge and discharge facility load prediction information generation device according to the present disclosure;

3 is a schematic structural diagram of an electronic device suitable for implementing some embodiments of the present disclosure.

Detailed ways

Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the disclosure are shown in the drawings, it should be understood that the disclosure may be embodied in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete. It should be understood that the drawings and embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of the present disclosure.

It should also be noted that, for convenience of description, only the parts related to the invention are shown in the drawings. The embodiments and features of the embodiments in the present disclosure may be combined with each other without conflict.

It should be noted that concepts such as “first” and “second” mentioned in this disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units. Or interdependence.

It should be noted that the modifications of "one" and "plurality" mentioned in this disclosure are illustrative and not restrictive. Those skilled in the art will understand that unless the context clearly indicates otherwise, it should be understood as "one or Multiple”.

The names of messages or information exchanged between multiple devices in the embodiments of the present disclosure are for illustrative purposes only and are not used to limit the scope of these messages or information.

The present disclosure will be described in detail below in conjunction with embodiments with reference to the accompanying drawings.

Referring to FIG. 1 , a process 100 of some embodiments of an information security-based charging and discharging facility load prediction information generation method according to the present disclosure is shown. The information security-based charging and discharging facility load prediction information generation method includes the following steps:

Step 101: Collect the target local data group sequence.

In some embodiments, the execution subject (for example, a computing device) of the information security-based charging and discharging facility load prediction information generation method may collect a target local data group sequence. Wherein, the target local data group sequence is a local data group corresponding to at least one local terminal. Among them, the terminal type of the local terminal is the vehicle charging equipment type. The target local data is the operating data generated by the local terminal when charging the vehicle. In practice, the local terminal may be a vehicle charging device. For example, the local terminal may be a vehicle charging station for vehicle charging of electric vehicles and/or hybrid vehicles.

As an example, the execution subject may obtain the local data group corresponding to at least one local terminal as the target local data group sequence through a wired connection or a wireless connection.

It should be noted that the above wireless connection methods may include but are not limited to 3G/4G/5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connections that are now known or developed in the future. Connection method.

It should be noted that the above computing device may be hardware or software. When the computing device is hardware, it can be implemented as a distributed cluster composed of multiple servers or terminal devices, or it can be implemented as a single server or a single terminal device. When the computing device is embodied as software, it can be installed in the hardware device listed above. It may be implemented, for example, as multiple software or software modules for providing distributed services, or as a single software or software module. There are no specific limitations here. It should be understood that the number of computing devices may be any number depending on implementation needs.

Step 102: For each target local data group in the target local data group sequence, perform initial model training on the initial charging load prediction information generation model corresponding to the initial model file through the target local data group to obtain a trained model file.

In some embodiments, for each target local data group in the target local data group sequence, the execution subject can perform initial model training on the initial charging load prediction information generation model corresponding to the initial model file through the target local data group. , get the model file after training. Wherein, the above initial model file is a model file sent by the edge server associated with the local terminal corresponding to the above target local data group. Among them, the initial charging load prediction information generation model is a charging load prediction information generation model to be trained. The charging load prediction information generation model may be a model used to predict real-time charging load information of the local terminal when charging the vehicle. In practice, the above charging load prediction information generation model may be a GAN (Generative Adversarial Network) model.

In practice, the above execution subject can use the target local data group as a sample for model training, conduct unsupervised model training on the initial charging load prediction information generation model corresponding to the initial model file, and obtain a post-training model file.

In some optional implementations of some embodiments, performing initial model training on the initial charging load prediction information generation model corresponding to the initial model file through the target local data group to obtain the trained model file may include the following steps:

The first step is to perform data preprocessing on the above target local data group to obtain the preprocessed target local data group.

In practice, the above execution subject can perform decimal scaling standardization on the target local data group to obtain the preprocessed target local data group.

The second step is to perform model training on the initial charging load prediction information generation model corresponding to the above-mentioned initial model file according to the above-mentioned preprocessed target local data group to generate a candidate model file.

Wherein, the above candidate model file is a model file corresponding to the initial charging load prediction information generation model whose corresponding training times are consistent with the preset training times.

In practice, the above-mentioned execution subject can use the pre-processed target local data group as a training sample to conduct unsupervised model training on the initial charging load prediction information generation model corresponding to the above-mentioned initial model file, and will reach the initial charging after the preset training times. The initial model file corresponding to the load forecast information generation model is used as a candidate model file.

The third step is to perform parameter encryption on the above-mentioned candidate model files to generate the above-mentioned post-training model files.

In practice, the above execution subject may use a symmetric encryption algorithm to encrypt the parameters of the above candidate model file to generate the above trained model file.

Step 103: Split the model files on the obtained set of trained model files to obtain a set of post-trained model file groups.

In some embodiments, the above execution subject can split the obtained set of trained model files into model files to obtain a set of post-trained model file groups. Wherein, the above-mentioned set of trained model file groups corresponds to at least one edge server.

In practice, the above execution subject can use the terminal type of the local terminal to split the obtained set of trained model files into model files to obtain a set of post-trained model file groups.

In some optional implementations of some embodiments, the above execution subject performs model file splitting on the obtained set of trained model files to obtain a set of post-trained model file groups, which may include the following steps:

The first step is to determine the edge server information set corresponding to the above target local data group sequence.

Wherein, the number of edge server information in the edge server information set is greater than a preset value. In practice, the default number is greater than or equal to 2.

The second step is to determine at least one post-training model file corresponding to each edge-side server information in the above-mentioned edge-side server information set in the above-mentioned post-training model file set, as a post-training model file group, to obtain the above-mentioned post-training model file group. gather.

Wherein, the above-mentioned execution subject may use the trained model file corresponding to at least one local terminal that has a communication relationship with the edge server corresponding to the edge server information as a group of trained model files.

Step 104: For each post-training model file group in the post-training model file group set, perform a model file aggregation on the post-training model file group through the edge server corresponding to the post-training model file group to obtain the first aggregated model. document.

In some embodiments, for each post-training model file group in the above-mentioned set of post-training model file groups, the above-mentioned execution subject can perform an operation on the above-mentioned post-training model file group through the edge server corresponding to the above-mentioned post-training model file group. The model files are aggregated to obtain the first aggregated model file.

In practice, the above execution subject may perform Bayesian aggregation on each post-training model file group in the post-training model file group set to obtain the first aggregated model file.

In some optional implementations of some embodiments, the execution subject performs a model file aggregation on the trained model file group through the edge server corresponding to the trained model file group to obtain the first aggregated model file, The following steps can be included:

The first step is to determine the first Bayesian prior probability and the first sample distribution of the above-mentioned post-training model file group based on the above-mentioned post-training model file group.

Among them, the above-mentioned first Bayesian prior probability is:

In practice,/> It is the edge parameter corresponding to the edge server. /> Indicates the serial number. /> Is the first/> The edge parameters corresponding to each edge server. /> Is the number of corresponding local terminals in the model file group after training. /> are local terminal parameters. /> is the serial number. /> Is the first/> local terminal parameters. /> Indicates the 1st to the/> local terminal parameters, which means/> . /> is the first Bayesian prior probability. /> Is the first/> The true posterior probability of the edge parameters corresponding to each edge server. /> is in/> If exists,/> probability of occurrence.

Among them, the above-mentioned first sample distribution is:

In practice,/> is the target local data group. /> is the serial number. /> Is the first/> target local data group. /> It is a traditional neural network model. For example,/> is a convolutional neural network model. /> is the first sample distribution.

The second step is to determine the first log-likelihood function of the edge parameters corresponding to the edge server.

Among them, the above first log-likelihood function is: Among them,/> Is the first/> The approximate posterior probability of the edge parameters corresponding to each edge server. /> is the first log-likelihood function. /> Represents the base 10 logarithmic function. is the KL divergence, where KL divergence measures the difference between the approximate posterior probability and the true posterior probability. Express/> with/> KL divergence between. /> Yes/> Evidence lower bound (ELBO, evidence lower bound).

The third step is to perform approximate processing on the above-mentioned first Bayesian prior probability and the above-mentioned first sample distribution to generate a first approximate Bayesian posterior probability.

In practice, the execution subject uses variational inference to approximate the product of the first Bayesian prior probability and the first sample distribution to obtain the first approximate Bayesian posterior probability.

In practice, the first approximate Bayesian posterior probability obtained using variational inference is: In practice,/> is the assignment symbol. /> is the first approximation Bayesian posterior probability. /> Yes/> variational parameters. /> Is the first/> /> the approximate posterior probability of . /> Yes/> variational parameters. /> Yes/> variational parameters. The variational parameter refers to the independent variable (here the independent variable refers to the /> in the formula ,/> ,/> ) adaptive parameters that remain unchanged during transformation and are used in mathematical analysis. /> Is the first/> /> The approximate posterior probability of ,/> Is the first/> Approximate posterior probabilities of local terminal parameters. /> Yes/> and/> approximate posterior probabilities. /> Is the first/> /> and/> approximate posterior probabilities.

The fourth step is to obtain the first objective function based on the above-mentioned first log-likelihood function and the above-mentioned first approximate Bayesian posterior probability.

Wherein, the above-mentioned first objective function includes: an edge server parameter and a local terminal parameter set, the edge server parameter represents the edge parameter corresponding to the edge server, and the local terminal parameter represents the local terminal parameter corresponding to the edge server.

In practice, the above execution subject uses standard variational reasoning technology to obtain the first objective function.

Among them, the first objective function is: In practice,/> is the first objective function. /> express and/> KL divergence between. /> is the first log-likelihood function. /> Yes/> The log-likelihood function of , the formula is:/> Among them,/> No./> /> The true posterior probability of ,/> Is the first/> The true posterior probability of the local terminal parameters. /> Yes/> Evidence lower bound (ELBO, evidence lower bound). /> Express/> with/> KL divergence between.

In the fifth step, in response to determining that the edge server parameters converge, perform parameter optimization processing on the local terminal parameter set to obtain a first model file.

In practice, in response to determining that the edge server parameters converge, the execution subject performs parameter optimization processing on the local terminal parameter set by minimizing the first objective function to obtain the first model file.

Among them, when the above-mentioned edge server parameters converge, the above-mentioned minimization first objective function is: Among them:/> . /> Express/> .

In the sixth step, in response to determining that the local terminal parameters in the local terminal parameter set have converged, perform parameter optimization processing on the edge server parameters to obtain a second model file.

In practice, in response to determining that the local terminal parameters in the local terminal parameter set have converged, the execution subject performs parameter optimization processing on the edge server parameters by minimizing the first objective function to obtain the second model file.

Wherein, when the local terminal parameters in the above local terminal parameter set converge, the above-mentioned minimization first objective function is: Among them:/> . /> Yes/> Evidence lower bound (ELBO, evidence lower bound). /> Yes/> and/> is the log-likelihood function of the parameters. /> Yes/> with/> KL divergence between.

The seventh step is to determine the above-mentioned first model file and the above-mentioned second model file as the above-mentioned first aggregated model file.

Step 105: Perform secondary model file aggregation on the obtained first aggregated model file set to obtain a second aggregated model file.

In some embodiments, the execution subject may perform secondary model file aggregation on the first aggregated model file set to obtain a second aggregated model file.

In practice, the above execution subject will obtain the first aggregated model file set and perform Bayesian aggregation to obtain the second aggregated model file.

In some optional implementations of some embodiments, performing secondary model file aggregation on the first aggregated model file set to obtain the second aggregated model file may include the following steps:

The first step is to determine the second Bayesian prior probability and the second sample distribution of the first aggregated model file set based on the first aggregated model file set.

Among them, the above second Bayesian prior probability is: In practice,/> Is the cloud aggregation server parameter. /> is the number of edge servers. /> Indicates the 1st to the/> The edge parameters corresponding to the edge servers represent/> . /> is the second Bayesian prior probability. /> It is the true posterior probability of the cloud parameters connecting the cloud aggregation server to each edge server. /> is in/> If exists,/> probability of occurrence.

Among them, the above second sample distribution is:

In practice,/> Yes and No./> The sum of the target local data groups of each local terminal connected to the edge server. /> is a traditional neural network model, for example, /> It can be a convolutional neural network model. is the second sample distribution.

The second step is to determine the second log-likelihood function of the edge-side server parameters corresponding to each edge-side server in the above-mentioned at least one edge-side server, and obtain a second log-likelihood function set.

Among them, the above second log-likelihood function is: In practice,/> is the second log-likelihood function. is the approximate posterior probability of the cloud aggregation server parameters. /> Express/> with/> KL divergence between. /> Yes/> Evidence lower bound (ELBO, evidence lower bound).

The third step is to generate a second approximate Bayesian posterior probability based on the above-mentioned second Bayesian prior probability and the above-mentioned second sample distribution.

In practice, the execution subject uses variational inference to approximate the product of the second Bayesian prior probability and the second sample distribution to obtain a second approximate Bayesian posterior probability.

Therefore, the second approximate Bayesian posterior probability obtained using variational inference is: In practice,/> is the second approximation Bayesian posterior probability. By combining the first-approximation Bayesian posterior probability with the second-approximation Bayesian posterior probability, we can see that the variational parameters by/> ,/> and/> ,/> composition. /> Yes/> variational parameters. /> Yes/> and/> approximate posterior probabilities. /> Is the first/> /> and/> approximate posterior probabilities.

The fourth step is to obtain a second objective function based on the above-mentioned second log-likelihood function set and the above-mentioned second approximate Bayesian posterior probability, where the above-mentioned second objective function includes: cloud aggregation server parameters and target edge server Parameter set, the above-mentioned cloud aggregation server parameters represent cloud parameters between edge servers corresponding to the above-mentioned cloud aggregation server, and the target edge server parameter set represents an edge parameter set of at least one edge server.

In practice, the above execution subject uses standard variational reasoning techniques to obtain the second objective function.

Among them, the second objective function is: in: ,/> is the second objective function. Express/> and/> KL divergence between. /> is the second log-likelihood function. Yes/> The log-likelihood function of , where:/> Among them,/> No./> /> The true posterior probability of ,/> Is the first/> The true posterior probability of the edge parameters of the edge server. /> Yes/> Evidence lower bound (ELBO, evidence lower bound). Express/> with/> KL divergence between.

In the fifth step, in response to determining that the cloud aggregation server parameters have converged, perform parameter optimization processing on the target edge server parameter set to obtain a third model file.

In practice, in response to determining that the cloud aggregation server parameters have converged, the execution subject performs parameter optimization processing on the target edge server parameter set by minimizing the first objective function to obtain a third model file.

Among them, the above-mentioned minimization second objective function is: in, Express/> .

In the sixth step, in response to determining that the target edge server parameters in the target edge server parameter set have converged, perform parameter optimization processing on the cloud aggregation server parameters to obtain a fourth model file.

In practice, in response to determining that the target edge server parameters in the target edge server parameter set have converged, the execution subject performs parameter optimization processing on the cloud aggregation server parameters by minimizing the second objective function to obtain a fourth model file.

Among them, the above-mentioned minimization second objective function is: Among them,/> Yes/> The evidence lower bound (ELBO, evidence lowerbound). /> Yes/> and/> is the log-likelihood function of the parameters. /> yes with/> KL divergence between.

The seventh step is to determine the above-mentioned third model file and the above-mentioned fourth model file as the above-mentioned second aggregated model file.

The content in "Some optional implementations in some embodiments" in steps 104 to 105, as an inventive point of this disclosure, solves the second technical problem mentioned in the background art, that is, "because each local terminal The data volume of the corresponding local data group differs greatly. When the data volume of the local data group is small, the prediction accuracy of the local charging load prediction information generation model trained based on the local data group will be poor. When the predicted charging When there is a large difference between the load prediction information and the actual charging load, it may cause damage to the vehicle charging equipment." In practice, due to the large difference in the amount of data in the local data group corresponding to each local terminal, the number of training samples To a certain extent, it determines the prediction accuracy of the local charging load prediction information generation model. When the amount of data in the local data group is small, it will lead to poor prediction accuracy of the local charging load prediction information generation model trained based on the local data group. , when there is a large difference between the predicted charging load prediction information and the actual charging load, it may cause damage to the vehicle charging equipment. Based on this, first of all, this disclosure performs a model file aggregation on the post-training model file group through the edge server corresponding to the post-training model file group for each post-training model file group in the post-training model file group set, and obtains the first An aggregated model file. Model aggregation is performed on the trained model file group corresponding to at least one local terminal of the same edge server. While protecting data privacy, it indirectly realizes the integration and sharing of local data groups corresponding to each local terminal, improving improve the robustness of the model. Finally, the present disclosure performs secondary model file aggregation on the obtained first aggregated model file set to obtain the second aggregated model file. Through hierarchical model aggregation, the trained model files corresponding to each local terminal can be indirectly integrated under the condition of data isolation. In summary, through two model aggregations, the local data groups corresponding to each local terminal are data fused and indirectly shared while protecting data privacy, thereby improving the accuracy of charging load prediction information and avoiding It eliminates the problem of possible damage to vehicle charging equipment due to a large difference between the predicted charging load prediction information and the actual charging load, and ensures the equipment safety of vehicle charging equipment.

Step 106: For each local terminal in at least one local terminal, generate a model based on the real-time operating data corresponding to the local terminal and the charging load prediction information corresponding to the second aggregated model file, and generate the charging load prediction information corresponding to the local terminal.

In some embodiments, the execution subject may generate a model for each local terminal in the at least one local terminal based on the real-time operating data corresponding to the local terminal and the charging load prediction information corresponding to the second aggregated model file, and generate Charging load prediction information corresponding to the above-mentioned local terminal. Among them, the real-time operating data is the corresponding real-time operating data of the local terminal when the terminal is running (for example, during vehicle charging). In practice, the charging load prediction information can represent the predicted charging load of the local terminal in a future period of time.

In practice, the execution subject may input the real-time operating data corresponding to the local terminal into the charging load prediction information generation model corresponding to the second aggregated model file to generate the charging load prediction information corresponding to the local terminal.

Optionally, the charging load prediction information generation model corresponding to the second aggregated model file may include: a data feature extraction model and an information prediction model.

In some optional implementations of some embodiments, the execution subject generates a model based on the real-time operating data corresponding to the local terminal and the charging load prediction information corresponding to the second aggregated model file, and generates the charging load corresponding to the local terminal. Forecasting information can include the following steps:

In the first step, the real-time operating data corresponding to the above-mentioned local terminal is input into the above-mentioned data feature extraction model to generate feature-extracted data information.

Among them, the data feature extraction model can include K serially connected convolutional residual blocks based on the attention mechanism. In practice, the convolutional residual block consists of 2 convolutional layers, 1 residual module and 1 content-based attention mechanism (Content-Based Attention, CBA) layer.

In practice, the above-mentioned execution subject can perform feature extraction on real-time operating data according to the above-mentioned data feature extraction model to generate feature-extracted data information.

In the second step, the data information after feature extraction is subjected to data cleaning processing to obtain the cleaned data information.

In practice, the above-mentioned execution subject can perform outlier elimination processing on the above-mentioned feature-extracted data information to generate cleaned data information.

The third step is to perform information prediction on the above-mentioned cleaned data information through the above-mentioned information prediction model to obtain the charging load prediction information corresponding to the above-mentioned local terminal.

In practice, the execution subject may input the cleaned data information into the information prediction model (for example, a fully connected layer) to generate the charging load prediction information corresponding to the local terminal.

The above-mentioned content in “in some optional implementations of some embodiments” serves as an inventive point of the present disclosure and solves the third technical problem mentioned in the background art, that is, “when the traditional linearly connected convolutional neural network When the number of network layers is deepened, it may cause the problem of feature forgetting, thereby affecting the accuracy of the generated charging load prediction information. When there is a large difference between the predicted charging load prediction information and the actual charging load, it may cause vehicle charging equipment causing damage." In practice, when the number of network layers of a traditional linearly connected convolutional neural network is deepened, it is easy to cause the gradient to disappear, which may cause the problem of feature forgetting, thereby affecting the accuracy of the generated charging load prediction information. Therefore, when there is a large difference between the predicted charging load prediction information and the actual charging load, it may cause damage to the vehicle charging equipment. Based on this, the present disclosure designs a data feature extraction model and an information prediction model. First, features are extracted from the real-time running data through the data feature extraction model. The residual module included in the data feature extraction model can avoid the problem of performance degradation caused by network deepening and alleviate the problem of gradient disappearance. Then, perform data cleaning processing on the above feature extracted data information to obtain the cleaned data information. By cleaning data, computing consumption can be reduced and data quality improved. In addition, the cleaned data information is input into the above-mentioned information prediction model to generate the charging load prediction information corresponding to the above-mentioned local terminal. In this way, the problem of feature forgetting that may be caused by the deepening of the network layers of the traditional linearly connected convolutional neural network is avoided, and the accuracy of the charging load prediction information is improved.

The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the charging and discharging facility load prediction information generation method based on information security of some embodiments of the present disclosure, it is avoided that when there are a large number of new energy vehicles to be stored at the same time, causing the storage The power grid corresponding to the power station is overloaded, causing voltage instability and possibly causing damage to vehicle charging equipment. Specifically, the reason for damage to vehicle charging equipment is that the presence of a large number of new energy vehicles to be stored at the same time may cause the grid load corresponding to the energy storage site to increase rapidly, causing the grid corresponding to the energy storage site to be overloaded, which may cause damage to the vehicle. Damage to charging equipment. Based on this, some embodiments of the present disclosure provide a method for generating load prediction information for charging and discharging facilities based on information security. First, a target local data group sequence is collected, wherein the target local data group sequence is a local data group corresponding to at least one local terminal. , where the terminal type of the local terminal is the vehicle charging equipment type, and the target local data is the operating data generated by the local terminal when charging the vehicle. Secondly, for each target local data group in the above target local data group sequence, perform initial model training on the initial charging load prediction information generation model corresponding to the initial model file through the above target local data group, and obtain the trained model file, where , the above-mentioned initial model file is a model file sent by the edge server associated with the local terminal corresponding to the above-mentioned target local data group, wherein the initial charging load prediction information generation model is the charging load prediction information generation model to be model trained. By using each target local data group in the above target local data group sequence as a training sample and performing initial model training on the initial charging load prediction information generation model corresponding to the initial model file, a personalized model corresponding to the target local data group can be generated. local model. At the same time, for each initial charging load prediction information generation model, the entire model training process only involves the target local data group, and the local data does not go out of the domain (that is, different target local data groups are isolated from each other), achieving protection in Model generation under the premise of data privacy. Next, the obtained set of trained model files is split into model files to obtain a set of post-trained model file groups, where the above set of post-trained model file groups corresponds to at least one edge server. In practice, different trained model files often correspond to different edge servers, and the corresponding edge servers need to aggregate model files. Therefore, the trained model files need to be split into model files. Further, for each post-training model file group in the above-mentioned post-training model file group set, perform a model file aggregation on the above-mentioned post-training model file group through the edge server corresponding to the above-mentioned post-training model file group to obtain the first aggregation Post model file. Through model file aggregation, data fusion and sharing can be achieved indirectly when data are isolated from each other, improving the robustness of the model while protecting the privacy of the data. In addition, secondary model file aggregation is performed on the obtained first aggregated model file set to obtain a second aggregated model file. Through hierarchical aggregation, computing costs can be reduced while protecting data privacy. Finally, for each local terminal in the at least one local terminal, a model is generated based on the real-time operating data corresponding to the local terminal and the charging load prediction information corresponding to the second aggregated model file to generate a charging load prediction corresponding to the local terminal. information. In this way, accurate charging load prediction information can be generated, and the charging load can be adjusted by combining the charging load prediction information. This can avoid the voltage caused by the overload of the power grid corresponding to the energy storage site when there are a large number of new energy vehicles to be stored. instability problem, thereby reducing the probability of vehicle charging equipment being damaged.

With further reference to Figure 2, as an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of an information security-based charge and discharge facility load prediction information generation device. These device embodiments are similar to those shown in Figure 1 Corresponding to the method embodiment, the device for generating information security-based charging and discharging facility load prediction information can be applied to various electronic devices.

As shown in Figure 2, the charging and discharging facility load prediction information generation device 200 based on information security in some embodiments includes: an acquisition unit 201, a model training unit 202, a splitting unit 203, a first aggregation unit 204, and a second aggregation unit 205. and generation unit 206. The collection unit 201 is configured to collect a target local data group sequence, where the target local data group sequence is a local data group corresponding to at least one local terminal, wherein the terminal type of the local terminal is a vehicle charging equipment type, and the target local data group sequence is a vehicle charging equipment type. The data is operating data generated by the local terminal when charging the vehicle; the model training unit 202 is configured to, for each target local data group in the above-mentioned target local data group sequence, correspond to the initial model file through the above-mentioned target local data group. The initial charging load prediction information generation model is subjected to initial model training to obtain a post-training model file, where the above-mentioned initial model file is a model file sent by the edge server associated with the local terminal corresponding to the above-mentioned target local data group, where the initial charging The load prediction information generation model is a charging load prediction information generation model to be model trained; the splitting unit 203 is configured to perform model file splitting on the obtained set of trained model files to obtain a set of trained model file groups, where, The above-mentioned set of trained model file groups corresponds to at least one edge server; the first aggregation unit 204 is configured to, for each post-trained model file group in the above-mentioned set of post-trained model file groups, through the above-mentioned post-trained model file group corresponding The edge server performs a model file aggregation on the above-mentioned set of trained model files to obtain a first aggregated model file; the second aggregation unit 205 is configured to perform a secondary model file aggregation on the obtained first aggregated model file set, Obtain the second aggregated model file; the generating unit 206 is configured to, for each local terminal in the at least one local terminal, based on the real-time operating data corresponding to the local terminal and the charging load prediction corresponding to the second aggregated model file. An information generation model generates charging load prediction information corresponding to the above-mentioned local terminal.

It can be understood that the units recorded in the information security-based charging and discharging facility load prediction information generating device 200 correspond to each step in the method described with reference to FIG. 1 . Therefore, the operations, features and beneficial effects described above for the method are also applicable to the charging and discharging facility load prediction information generation device 200 based on information security and the units included therein, and will not be described again here.

Referring now to FIG. 3 , a schematic structural diagram of an electronic device (eg, computing device) 300 suitable for implementing some embodiments of the present disclosure is shown. The electronic device shown in FIG. 3 is only an example and should not impose any limitations on the functions and scope of use of the embodiments of the present disclosure.

As shown in FIG. 3 , the electronic device 300 may include a processing device (eg, central processing unit, graphics processor, etc.) 301 , which may be loaded into the random access memory 303 according to a program stored in the read-only memory 302 or from the storage device 308 program to perform various appropriate actions and processes. In the random access memory 303, various programs and data required for the operation of the electronic device 300 are also stored. The processing device 301, the read-only memory 302 and the random access memory 303 are connected to each other via a bus 304. Input/output interface 305 is also connected to bus 304.

Generally, the following devices may be connected to the I/O interface 305: input devices 306 including, for example, a touch screen, touch pad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, etc.; including, for example, a liquid crystal display (LCD), speaker, vibration An output device 307 such as a computer; a storage device 308 including a magnetic tape, a hard disk, etc.; and a communication device 309. The communication device 309 may allow the electronic device 300 to communicate wirelessly or wiredly with other devices to exchange data. Although FIG. 3 illustrates electronic device 300 with various means, it should be understood that implementation or availability of all illustrated means is not required. More or fewer means may alternatively be implemented or provided. Each block shown in Figure 3 may represent one device, or may represent multiple devices as needed.

In particular, according to some embodiments of the present disclosure, the processes described above with reference to the flowcharts may be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product including a computer program carried on a computer-readable medium, the computer program containing program code for performing the method illustrated in the flowchart. In some such embodiments, the computer program may be downloaded and installed from the network via communication device 309, or from storage device 308, or from read-only memory 302. When the computer program is executed by the processing device 301, the above-described functions defined in the methods of some embodiments of the present disclosure are performed.

It should be noted that the computer-readable medium recorded in some embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium, or any combination of the above two. The computer-readable storage medium may be, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any combination thereof. More specific examples of computer readable storage media may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard drive, random access memory (RAM), read only memory (ROM), removable Programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, a computer-readable storage medium may be any tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. In some embodiments of the present disclosure, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code therein. Such propagated data signals may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device . Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wire, optical fiber cable, RF (radio frequency), etc., or any suitable combination of the foregoing.

In some embodiments, the client and server can communicate using any currently known or future developed network protocol such as HTTP (Hyper Text Transfer Protocol), and can communicate with digital data in any form or medium. Communications (e.g., communications networks) interconnections. Examples of communications networks include local area networks ("LAN"), wide area networks ("WAN"), the Internet (e.g., the Internet), and end-to-end networks (e.g., ad hoc end-to-end networks), as well as any currently known or developed in the future network of.

The above-mentioned computer-readable medium may be included in the above-mentioned electronic device; it may also exist independently without being assembled into the electronic device. The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device: collects a target local data group sequence, wherein the target local data group sequence is at least one The local data group corresponding to the local terminal, where the terminal type of the local terminal is the vehicle charging equipment type, and the target local data is the operating data generated by the local terminal when charging the vehicle; for each target local data group in the above sequence data group, through the above-mentioned target local data group, perform initial model training on the initial charging load prediction information generation model corresponding to the initial model file, and obtain a post-training model file, wherein the above-mentioned initial model file is a local data group corresponding to the above-mentioned target local data group. The model file sent by the edge server associated with the terminal, in which the initial charging load prediction information generation model is the charging load prediction information generation model to be model trained; the obtained set of trained model files is split into model files to obtain the A set of model file groups, wherein the above-mentioned set of trained model file groups corresponds to at least one edge server; for each post-trained model file group in the above-mentioned set of post-trained model file groups, the edge server corresponding to the above-mentioned post-trained model file group is The server performs a model file aggregation on the above-mentioned trained model file group to obtain a first aggregation model file; performs a second model file aggregation on the obtained first aggregation model file set to obtain a second aggregation model file; for at least the above Each local terminal in a local terminal generates a charging load prediction information corresponding to the local terminal based on the real-time operating data corresponding to the local terminal and the charging load prediction information generation model corresponding to the second aggregated model file.

Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages, including object-oriented programming languages—such as Java, Smalltalk, C++, or a combination thereof, Also included are conventional procedural programming languages—such as the "C" language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In situations involving remote computers, the remote computer can be connected to the user's computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer, such as an Internet service provider. connected via the Internet).

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operations of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, segment, or portion of code that contains one or more logic functions that implement the specified executable instructions. It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown one after another may actually execute substantially in parallel, or they may sometimes execute in the reverse order, depending on the functionality involved. It will also be noted that each block of the block diagram and/or flowchart illustration, and combinations of blocks in the block diagram and/or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or operations. , or can be implemented using a combination of specialized hardware and computer instructions.

The units described in some embodiments of the present disclosure may be implemented in software or hardware. The described unit may also be provided in a processor. For example, it may be described as follows: a processor includes an acquisition unit, a model training unit, a splitting unit, a first aggregation unit, a second aggregation unit and a generation unit. The names of these units do not constitute a limitation on the unit itself under certain circumstances. For example, the collection unit may also be described as "a unit that collects the target local data group sequence."

The functions described above herein may be performed, at least in part, by one or more hardware logic components. For example, and without limitation, exemplary types of hardware logic components that may be used include: Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), Application Specific Standard Products (ASSPs), Systems on Chips (SOCs), Complex Programmable Logical device (CPLD) and so on.

The above description is only an illustration of some preferred embodiments of the present disclosure and the technical principles applied. Persons skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to technical solutions composed of specific combinations of the above technical features, and should also cover the above-mentioned technical solutions without departing from the above-mentioned inventive concept. Other technical solutions formed by any combination of technical features or their equivalent features. For example, a technical solution is formed by replacing the above features with technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to).

Claims (7)

1. A charge and discharge facility load prediction information generation method based on information security comprises the following steps:

collecting a target sample local data set sequence, wherein the target sample local data set sequence is a local data set corresponding to at least one local terminal, the terminal type of the local terminal is a vehicle charging equipment type, and the target sample local data is operation data generated when the local terminal charges a vehicle;

for each target sample local data set in the target sample local data set sequence, performing initial model training on an initial charge load prediction information generation model corresponding to an initial model file through the target sample local data set to obtain a trained model file, wherein the initial model file is a model file sent by an edge server associated with a local terminal corresponding to the target sample local data set, and the initial charge load prediction information generation model is a charge load prediction information generation model to be subjected to model training;

splitting the model files of the obtained training model file set to obtain a training model file set, wherein the training model file set corresponds to at least one edge server;

For each trained model file group in the trained model file group set, performing primary model file aggregation on the trained model file group through an edge server corresponding to the trained model file group to obtain a first aggregated model file;

performing secondary model file aggregation on the obtained first aggregated model file set to obtain a second aggregated model file;

and generating a model for each local terminal in the at least one local terminal according to the real-time operation data corresponding to the local terminal and the charging load prediction information corresponding to the second aggregated model file, and generating the charging load prediction information corresponding to the local terminal.

2. The method according to claim 1, wherein the performing initial model training on the initial charge load prediction information generation model corresponding to the initial model file through the target local data set to obtain a trained model file includes:

performing data preprocessing on the target specimen data set to obtain a preprocessed target specimen data set;

model training is carried out on an initial charge load prediction information generation model corresponding to the initial model file according to the preprocessed target specimen data set so as to generate a candidate model file, wherein the candidate model file is a model file corresponding to the initial charge load prediction information generation model, and the corresponding training times of the model file are consistent with the preset training times;

And carrying out parameter encryption on the candidate model file to generate the trained model file.

3. The method of claim 2, wherein the splitting the model file from the obtained trained model file set to obtain a trained model file set includes:

determining an edge server information set corresponding to the target local data set sequence, wherein the number of the edge server information in the edge server information set is larger than a preset value;

and determining at least one trained model file corresponding to each piece of edge server information in the trained model file set as a trained model file set, and obtaining the trained model file set.

4. The method of claim 3, wherein the performing, by the edge server corresponding to the trained model file group, model file aggregation on the trained model file group once to obtain a first aggregated model file includes:

determining a first Bayesian prior probability and a first sample distribution of the trained model file group according to the trained model file group;

Determining a first log likelihood function of an edge parameter corresponding to the edge server;

performing approximate processing on the first Bayesian prior probability and the first sample distribution to generate a first approximate Bayesian posterior probability;

obtaining a first objective function based on the first log-likelihood function and the first approximate Bayesian posterior probability, wherein the first objective function comprises: the system comprises an edge server parameter and a local terminal parameter set, wherein the edge server parameter represents an edge parameter corresponding to the edge server, and the local terminal parameter represents a parameter of a local terminal corresponding to the edge server;

in response to determining that the edge server parameter converges, performing parameter optimization processing on a local terminal parameter set to obtain a first model file;

responding to the determination of the convergence of the local terminal parameters in the local terminal parameter set, and carrying out parameter optimization processing on the edge server parameters to obtain a second model file;

and determining the first model file and the second model file as the first aggregated model file.

5. An information security-based charge and discharge facility load prediction information generation device, comprising:

The system comprises an acquisition unit, a storage unit and a control unit, wherein the acquisition unit is configured to acquire a target local data set sequence, the target local data set sequence is a local data set corresponding to at least one local terminal, the terminal type of the local terminal is a vehicle charging equipment type, and the target local data is operation data generated when the local terminal charges a vehicle;

the model training unit is configured to perform initial model training on an initial charge load prediction information generation model corresponding to an initial model file through each target local data set in the target local data set sequence to obtain a trained model file, wherein the initial model file is a model file sent by an edge server associated with a local terminal corresponding to the target local data set, and the initial charge load prediction information generation model is a charge load prediction information generation model to be subjected to model training;

the splitting unit is configured to split the model files of the obtained training model file set to obtain a training model file set, wherein the training model file set corresponds to at least one edge server;

The first aggregation unit is configured to aggregate the model files once through the edge server corresponding to the trained model file group for each trained model file group in the trained model file group set to obtain a first aggregated model file;

the second aggregation unit is configured to perform secondary model file aggregation on the first aggregated model file set to obtain a second aggregated model file;

and the generating unit is configured to generate a charging load prediction information corresponding to each local terminal in the at least one local terminal according to the real-time operation data corresponding to the local terminal and the charging load prediction information generation model corresponding to the second aggregated model file.

6. An electronic device, comprising:

one or more processors;

a storage device having one or more programs stored thereon;

when executed by the one or more processors, causes the one or more processors to implement the method of any of claims 1 to 4.

7. A computer readable medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements the method of any of claims 1 to 4.