Prediction and assessment method, device and storage medium based on power grid data risk

Abstract

The present application provides a prediction and assessment method, device and storage medium based on power grid data risk, the method comprising: obtaining business operation data and text data; annotating the business operation data with risk categories to obtain a first batch of annotated data, and annotating the text data with sentiment categories to obtain a second batch of annotated data; inputting the first batch of annotated data and the second batch of annotated data into a combined network model for training to obtain a trained risk prediction model, the combined network model being a combined network of CNNs and Attention; inputting the power grid data obtained at the current moment into the risk prediction model to obtain the output corresponding predicted risk status. In the process of analyzing the security of power grid data, this method introduces sentiment indicators and uses the CNNs+Attention neural network to extract features between data, thereby making timely predictions and assessments of the risk status of power grid data.

Description

Technical Field

[0001] The present application relates to the field of information security of electric power enterprises, and specifically to a prediction and assessment method based on power grid data risks, a prediction and assessment device based on power grid data risks, a computer-readable storage medium, and an electronic device. Background Art

[0002] Driven by advances in technologies like artificial intelligence and cloud computing, the digitalization of the power industry is rapidly advancing. However, in practice, power companies face challenges such as inadequate accuracy and timeliness in addressing data security risks, and inconsistent data classification and grading standards across various companies. Summary of the Invention

[0003] The main purpose of this application is to provide a prediction and assessment method based on power grid data risk, a prediction and assessment device based on power grid data risk, a computer-readable storage medium and an electronic device, so as to at least solve the problems of insufficient accuracy of existing power enterprise data security risks and inconsistent risk assessment standards within and outside each enterprise.

[0004] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a prediction and assessment method based on power grid data risk is provided, including: obtaining business operation data and text data, the business operation data being the historical operation data of the power grid digital business operation platform, and the text data being the operation text data of each branch and subsidiary obtained from the internal network of the power grid; performing risk category annotation processing on the business operation data to obtain a first batch of annotation data, and performing sentiment category annotation processing on the text data to obtain a second batch of annotation data; inputting the first batch of annotation data and the second batch of annotation data into a combined network model for training to obtain a trained risk prediction model, the combined network model being a combination network of CNNs and Attention; inputting the power grid data obtained at the current moment into the risk prediction model to obtain the corresponding predicted risk status as output.

[0005] Optionally, obtaining business operation data and text data includes: obtaining initial business data and initial text data, wherein the initial text data is the operation text data of each branch and subsidiary obtained through web crawler technology; preprocessing the initial business data and the initial text data respectively to obtain the business operation data and the text data, wherein the preprocessing includes at least one of the following: data cleaning, data filling, and data standardization, wherein the data cleaning is a process of cleaning out outliers, and the data filling is a process of filling missing values.

[0006] Optionally, the first batch of annotation data and the second batch of annotation data are input into a combined network model for training to obtain a trained risk prediction model, including: obtaining the data storage time of the first batch of annotation data, and setting corresponding data weight values ​​for the first batch of annotation data according to the data storage time to obtain first processed data; extracting corresponding emotional features from the second batch of annotation data based on the LSTM neural network, and classifying and predicting the second batch of annotation data according to the emotional features to obtain corresponding predicted emotional categories; converting the predicted emotional categories into corresponding category values ​​as target emotional features and forming target feature data with the first processed data; inputting the target feature data into a combined network of CNNs and Attention for training to obtain the trained risk prediction model.

[0007] Optionally, according to the data storage time, a corresponding data weight value is set for the first batch of annotation data to obtain the first processed data, including: when the year of the data storage time is the current year, the data weight value Where n is the number of months between the data storage time and the current time; if the year of the data storage time is not the current year, the data weight value is Wherein, b is the number of years between the data storage time and the current time.

[0008] Optionally, the target feature data is input into a combined network of CNNs and Attention for training to obtain the trained risk prediction model, including: using a CNNs network to extract target feature vector information from the target feature data; inputting the target feature vector information into an Attention network for training to obtain the trained risk prediction model.

[0009] Optionally, the combined network model is a model combining CNNs and Attention networks, and the combined network model is based on historical data [A K (t), A K (t-1),…,A K (t-m+1)], predict the next time step The main formula of the Attention network in the actual model training prediction task is: S t+1 =f(S t-1 ,y t-1 , c t ), where St is the state output at time t after decoding, St-1 is the state output at time t-1, yt-1 is the label corresponding to the input data at time t-1, f is an RNN, c trepresents the weighted sum of the hidden states.

[0010] Optionally, the c t The calculation formula is: Among them, t represents the time, h j is the output of the j-th input in the decoder, a tj is a weight parameter.

[0011] According to another aspect of the present application, a prediction and evaluation device based on power grid data risk is provided, including: an acquisition unit for acquiring business operation data and text data, wherein the business operation data is the historical operation data of the power grid digital business operation platform, and the text data is the operation text data of each branch and subsidiary obtained from the internal network of the power grid; a processing unit for performing risk category annotation processing on the business operation data to obtain a first batch of annotation data, and performing sentiment category annotation processing on the text data to obtain a second batch of annotation data; a training unit for inputting the first batch of annotation data and the second batch of annotation data into a combined network model for training to obtain a trained risk prediction model, wherein the combined network model is a combined network of CNNs and Attention; a prediction unit for inputting the power grid data obtained at the current moment into the risk prediction model to obtain the corresponding predicted risk state as output.

[0012] According to another aspect of the present application, a computer-readable storage medium is provided, which includes a stored program, wherein when the program is run, the device where the computer-readable storage medium is located is controlled to execute any one of the predictive assessment methods based on power grid data risks.

[0013] According to another aspect of the present application, an electronic device is provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include a method for executing any one of the predictive assessment methods based on power grid data risks.

[0014] Applying the technical solution of the present application, the above-mentioned prediction and assessment method based on power grid data risk first obtains business operation data and text data; then annotates the business operation data with risk categories to obtain the first batch of annotated data, and annotates the text data with sentiment categories to obtain the second batch of annotated data; then inputs the first batch of annotated data and the second batch of annotated data into a combined network model for training to obtain a trained risk prediction model, where the combined network model is a combined network of CNNs and Attention; finally, the power grid data obtained at the current moment is input into the risk prediction model to obtain the corresponding predicted risk status as output. This method assigns higher weights to new data according to year and month information, introduces text sentiment features into the security features of power grid business data, uses an improved CNNs network to learn deep features, and combines the combination of CNNs + attention to focus more on parts that are useful for improving risk assessment performance, thereby enabling timely prediction and assessment of power grid data risks, timely discovery of power grid business risks, and solving the problems of insufficient accuracy of existing power enterprise data security risks and inconsistent risk assessment standards within and outside the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The drawings that constitute part of this application are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation on this application. In the drawings:

[0016] Figure 1 A hardware structure block diagram of a mobile terminal for executing a prediction and assessment method based on power grid data risk provided in an embodiment of the present application is shown;

[0017] Figure 2 A schematic diagram of a flow chart of a method for predicting and evaluating power grid data risk provided in accordance with an embodiment of the present application is shown;

[0018] Figure 3 A schematic diagram of a CNN model provided according to an embodiment of the present application is shown;

[0019] Figure 4 A schematic flow chart of another method for predicting and evaluating power grid data risk provided in accordance with an embodiment of the present application is shown;

[0020] Figure 5 A structural block diagram of a prediction and assessment device based on power grid data risk provided according to an embodiment of the present application is shown.

[0021] The above drawings include the following reference numerals:

[0022] 102. Processor; 104. Memory; 106. Transmission device; 108. Input / output device. DETAILED DESCRIPTION

[0023] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0024] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0025] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] As introduced in the background technology, the existing technology lacks accuracy and timeliness in data security risks of power enterprises, and data classification and grading still face the problem of inconsistent classification standards within and outside each enterprise. In order to solve the problems of insufficient accuracy in existing data security risks of power enterprises and inconsistent risk assessment standards within and outside each enterprise, the embodiments of the present application provide a predictive assessment method based on power grid data risks, a predictive assessment device based on power grid data risks, a computer-readable storage medium and an electronic device.

[0027] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0028] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 This is a hardware structure block diagram of a mobile terminal for a method of predicting and evaluating power grid data risk according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1Only one is shown) a processor 102 (the processor 102 may include but is not limited to a microprocessor MCU or a programmable logic device FPGA and other processing devices) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input and output device 108 for communication functions. It will be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown.

[0029] Memory 104 can be used to store computer programs, such as application software programs and modules, such as the computer program corresponding to the method for predicting and assessing power grid data risk in an embodiment of the present invention. Processor 102 executes the computer program stored in memory 104 to perform various functional applications and data processing, thereby implementing the aforementioned method. Memory 104 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, memory 104 may further include memory remotely located from processor 102, which can be connected to the mobile terminal via a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof. Transmission device 106 is used to receive or transmit data via a network. Specific examples of such networks may include a wireless network provided by the mobile terminal's telecommunications provider. In one example, transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0030] In this embodiment, a predictive assessment method based on power grid data risk is provided, which runs on a mobile terminal, a computer terminal or a similar computing device. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0031] Figure 2 FIG is a flow chart of a method for predicting and evaluating grid data risk according to an embodiment of the present application. Figure 2 As shown, the method includes the following steps:

[0032] Step S201: Acquire business operation data and text data. The business operation data is historical operation data of the power grid digital business operation platform, and the text data is operation text data of each branch and subsidiary obtained from the internal network of the power grid.

[0033] Specifically, business operation data mainly comes from the power grid digital business operation platform and database storage data, etc. Text data is text data such as the annual operation reports of various branches and subsidiaries obtained from internal websites through web crawler technology. The data includes date, company name, and the company's text title on that date, text content, and other information.

[0034] The steps of obtaining business operation data and text data include:

[0035] Step S2011, obtaining initial business data and initial text data, wherein the initial text data is the operation text data of each branch and subsidiary obtained by web crawler technology;

[0036] Step S2012, preprocessing the above-mentioned initial business data and the above-mentioned initial text data respectively to obtain the above-mentioned business operation data and the above-mentioned text data, wherein the above-mentioned preprocessing includes at least one of the following: data cleaning, data filling, and data standardization, the above-mentioned data cleaning is a process of cleaning out abnormal values, and the above-mentioned data filling is a process of filling missing values.

[0037] Specifically, by introducing text sentiment features into the security features of power grid business data, using the improved CNNs network to learn deep features, and combining CNNs+attention to focus more on the parts that are useful for performance improvement, it is possible to make timely predictions and assessments of power grid data risks.

[0038] Business operation data is collected through web crawlers and databases. Raw data may contain missing values ​​and outliers, which require cleaning to meet requirements. For outliers, measures such as removal or replacement can be taken to address their anomalies. When a variable has 45% or more missing values, it is deemed meaningless and removed. When the missing values ​​are relatively few and the data still have analytical significance, data can be supplemented with value additions.

[0039] In this example, stacking filling is used to fill missing values. The basic filling model includes the combination of mean and random coefficient, median, and mode filling. Stacking filling considers the prediction results of all basic filling models and outputs a final filling value to achieve the best filling effect. The formula for combining mean and random coefficient filling is as follows:

[0040] y=x+x×η,η∈(-0.3,0.3), where x is the mean and η is the filling factor, which is randomly selected between -0.3 and 0.3.

[0041] In addition, since different feature variables have different dimensions, it is necessary to consider standardizing them and adjust the data to a number between 0 and 1. The standardization formula is as follows:

[0042]

[0043] Among them, X i is the data after standardization, x i is the original data.

[0044] Step S202: Annotate the business operation data according to risk categories to obtain first-order annotated data, and annotate the text data according to sentiment categories to obtain second-order annotated data.

[0045] Specifically, the collected business operation data and text data are annotated according to the actual operation of the system. Based on the collected business operation data, the corresponding risk status is marked according to the actual business situation to form the initial training data set and test set.

[0046] The sentiment categories are marked according to the collected text data. There are five categories in total: positive, slightly positive, neutral, slightly negative, and negative. The corresponding sentiment discrimination values ​​are (1, 0.5, 0, -0.5, -1).

[0047] Step S203: Input the first batch of annotation data and the second batch of annotation data into a combined network model for training to obtain a trained risk prediction model. The combined network model is a combined network of CNNs and Attention.

[0048] Specifically, after the text data enters the sentiment classification model branch, the predicted value is added as a feature to the risk prediction model feature to perform power grid data security risk prediction.

[0049] The first batch of annotation data and the second batch of annotation data are input into a combined network model for training to obtain a trained risk prediction model, including the following steps:

[0050] Step S301, obtaining the data storage time of the first batch of annotation data, and setting corresponding data weight values ​​for the first batch of annotation data according to the data storage time, to obtain first processed data;

[0051] The method includes setting corresponding data weight values ​​for the first batch of annotation data according to the data storage time to obtain first processed data, including the following steps:

[0052] Step S3011: When the year of the data storage time is the current year, the data weight value Where n is the number of months between the data storage time and the current time;

[0053] Step S3012: If the year of the data storage time is not the current year, the data weight value Where b is the number of years between the data storage time and the current time.

[0054] Specifically, by using deep neural networks and giving new data higher weights according to year and month information during model training, more accurate predictions can be made on the risk status of power grid data.

[0055] During the model training process, different weights are assigned to new and old data according to the year and month information. The above process only applies to business operation data.

[0056] Step S302: extracting corresponding sentiment features from the second annotated data based on the LSTM neural network, and performing classification prediction on the second annotated data according to the sentiment features to obtain a corresponding predicted sentiment category;

[0057] Step S303, converting the predicted emotion category into a corresponding category value as a target emotion feature and combining it with the first processed data to form target feature data;

[0058] In step S304, the target feature data is input into a combined network of CNNs and Attention for training to obtain the trained risk prediction model.

[0059] Specifically, sentiment indicators are introduced while analyzing power grid data, and the improved CNNs+Attention neural network is used to extract features between data, so as to make timely predictions and assessments of the risk status of power grid data and provide early warnings for data security risks.

[0060] The target feature data is input into a combined network of CNNs and Attention for training to obtain the trained risk prediction model, which includes the following steps:

[0061] Step S3041, using a CNN network to extract target feature vector information from the target feature data;

[0062] Step S3042: Input the target feature vector information into the Attention network for training to obtain the trained risk prediction model.

[0063] Specifically, by introducing text sentiment features into the security features of power grid business data, using the improved CNNs network to learn deep features, and combining CNNs+attention to focus more on the parts that are useful for performance improvement, it is possible to timely predict and evaluate power grid data risks, effectively ensure data security, and promptly discover power grid business risks.

[0064] Among them, the above-mentioned combined network model is a model that combines CNNs and Attention network. The above-mentioned combined network model is based on historical data [A K (t), A K (t-1),…,A K (t-m+1)], predict the next time step The main formula of the Attention network in the actual model training prediction task is: S t+1 =f(S t-1 ,y t-1 , c t ), where St is the state output at time t after decoding, St-1 is the state output at time t-1, yt-1 is the label corresponding to the input data at time t-1, f is an RNN, c t represents the weighted sum of the hidden states.

[0065] Specifically, the improved CNNs network is used to learn deep features, which enables timely prediction and assessment of power grid data risks and effectively ensures data security.

[0066] Among them, the above c t The calculation formula is: Among them, t represents the time, h j is the output of the j-th input in the decoder, a tj is a weight parameter.

[0067] Specifically, the combination of CNNs+attention focuses more on the parts that are useful for performance improvement, thereby enabling timely prediction and assessment of power grid data risks, effectively ensuring data security, and timely discovering power grid business risks.

[0068] In the above steps, CNNs feature extraction uses an improved CNN network to extract data risk features. The CNNs network uses an improved convolutional network, and its structure is as follows: Figure 3As shown in the figure, the network structure includes a 32-channel 3×3 convolution layer, a 64-channel 3×3 convolution layer, a 64-channel 1×1 convolution layer, a 192-channel 1×1 convolution layer, a 192-channel 3×3 convolution layer, and a 64-channel 1×1 convolution layer. This model's backbone network can learn deep features. Furthermore, during the learning process, some important but sparse data is lost. Therefore, the concat operation is used to connect the feature tensors before convolution, thereby improving the model's performance.

[0069] The extracted feature vector is input into the prediction model, and the state prediction and its score are output. The input of the prediction model is the feature vector extracted from the collected data, and the output is the state prediction. The input space is X∈R d , the category label space is Y∈[0,1], and there are 2 prediction types. The labeled training sample set is where x i For each piece of collected data, each piece of data contains a lot of information, y i is the state type of each data mark, and n is the number of manually marked samples. The ultimate goal is to make the predicted state y i Infinitely close to the real state y t , so that the accuracy of the prediction evaluation is the highest, that is, y i =softmax(f(x i )), where the softmax function represents the probability of each classification being taken, and f(·) is what needs to be trained to make the output y i Infinitely close to the true value y t function, x i is the input data.

[0070] The prediction model structure is as follows Figure 3 As shown, the prediction model uses a combination of CNNs and Attention networks to have better learning and prediction capabilities, that is, based on historical data [A K (t), A K (t-1),…,A K (t-m+1)], predict the next time step Where k represents the sequence, t represents the time, and m represents the distance between the corresponding time step and the predicted time step.

[0071] The main formula of the Attention network in the actual model training prediction task is as follows:

[0072] S t+1 =f(S t-1 ,y t-1 ,C t );

[0073] Among them, S t is the output of the state at time t after decoding, S t-1 is the state output at time t-1, y t-1 is the expert label corresponding to the input data at time t-1, and f is a nonlinear multi-layer neural network, which consists of multiple layers of LSTM and softmax layers. t Represents the input sequence x i The weighted sum of each element in is weighted according to its importance. The specific calculation formula is as follows:

[0074]

[0075] The parameter t represents the time, h j =f(x j ) is the element x j The encoding.

[0076] a tj Is a weight parameter that reflects the element x j C t The importance of is calculated as follows:

[0077]

[0078] where e tj Reflected in the x t The element x to be encoded in the sequence j With other elements x k The higher the matching degree, the greater the influence of the element. tj The larger the value of e tk Represents the element to be encoded x k (k=1,2,3…) The degree of matching with other elements.

[0079] Step S204: input the power grid data acquired at the current moment into the above-mentioned risk prediction model to obtain the corresponding predicted risk status as output.

[0080] Specifically, the trained prediction model is used to analyze and predict the data security situation. Based on the current status, the current business operation data security status is predicted. The final prediction results are of two types: normal and dangerous.

[0081] In the embodiment, the collected power grid business operation data and text data are first subjected to data normalization preprocessing, and then the preprocessed data are sent to the trained data risk status prediction and evaluation model. The sentiment coefficient features of the collected data are first extracted through the LSTM network, and then added to the business operation data, and finally sent to the prediction and evaluation model. The model predicts whether the current power grid business operation is in a risky state.

[0082] The above-mentioned prediction and assessment method based on power grid data risk of the present application first obtains business operation data and text data; then performs risk category annotation processing on the business operation data to obtain the first batch of annotated data, and performs sentiment category annotation processing on the text data to obtain the second batch of annotated data; then the first batch of annotated data and the second batch of annotated data are input into the combined network model for training to obtain a trained risk prediction model, and the combined network model is a combination of CNNs and Attention; finally, the power grid data obtained at the current moment is input into the risk prediction model to obtain the output corresponding predicted risk status. This method gives new data a higher weight according to year and month information, introduces text sentiment features into the security features of power grid business data, uses the improved CNNs network to learn deep features, and combines the combination of CNNs+attention to focus more on the parts that are useful for improving risk assessment performance, thereby being able to timely predict and assess power grid data risks, timely discover power grid business risks, and solve the problems of insufficient accuracy of existing power enterprise data security risks and inconsistent risk assessment standards between internal and external enterprises.

[0083] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the implementation process of the prediction and assessment method based on power grid data risk of the present application will be described in detail below with reference to specific embodiments.

[0084] This embodiment relates to a specific prediction and assessment method based on power grid data risk, such as Figure 4 As shown, the following steps are included:

[0085] Step S1: Collect business operation data and text data through web crawlers and databases;

[0086] Step S2: pre-processing the collected business operation data and text data;

[0087] Step S3: annotate the collected business operation data and text data according to the actual system operation situation.

[0088] Step S4: Based on the collected business operation data, the corresponding risk status is marked according to the actual business situation to form the initial training data set and test set. Using the LSTM convolutional network, the sentiment features are extracted from the labeled text data set and the corresponding categories are predicted. Different weights are assigned to the business operation data according to the year and month information.

[0089] Step S5: After the text data enters the sentiment classification model branch, the predicted values ​​are added as features to the risk prediction model to predict power grid data security risks. The sentiment coefficient is added to the labeled business operation data, and the extracted feature vector is input into the prediction model of CNNs and Attention Networks, which outputs the state prediction and its score.

[0090] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0091] According to another embodiment of the present invention, a prediction and evaluation device based on power grid data risk includes a data acquisition module, a data processing module, a data risk feature extraction module, and a prediction and evaluation module, wherein the data acquisition module is used to regularly collect the company's business operation data and text data information; the data processing module is used to normalize the collected data and place the data on the same dimension to facilitate feature extraction and learning; the feature extraction module is used to first use the LSTM network to extract the sentiment coefficient from the corresponding text data, and then add the extracted sentiment coefficient as a feature to the normalized business operation data, and then give the new data a higher weight according to the year and month information, and then extract the data risk information features through the neural network; the prediction and evaluation module is used to input the extracted feature vector into a prediction model that combines CNNs and Attention networks, and output a state prediction and its score.

[0092] Among them, the data acquisition module transmits the collected data information to the data processing module through crawler scripts, database extraction, etc., which can be used to predict and evaluate the data risk status.

[0093] The data processing module mainly normalizes the collected data to facilitate feature extraction of the data.

[0094] The embodiments of the present application also provide another predictive assessment device based on power grid data risk. It should be noted that the predictive assessment device based on power grid data risk of the embodiments of the present application can be used to execute the predictive assessment method based on power grid data risk provided by the embodiments of the present application. The device is used to implement the above-mentioned embodiments and preferred implementation methods, and those that have been explained will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and conceived.

[0095] The following introduces the prediction and assessment device based on power grid data risk provided in an embodiment of the present application.

[0096] Figure 5 Schematic diagram of a prediction and assessment device based on power grid data risk according to an embodiment of the present application. Figure 5 As shown, the device includes: an acquisition unit 10, a processing unit 20, a training unit 30 and a prediction unit 40, the acquisition unit 10 is used to acquire business operation data and text data, the above-mentioned business operation data is the historical operation data of the digital business operation platform of the power grid, and the above-mentioned text data is the operation text data of each branch and subsidiary obtained from the internal network of the above-mentioned power grid; the processing unit 20 is used to perform risk category annotation processing on the above-mentioned business operation data to obtain the first batch of annotation data, and perform sentiment category annotation processing on the above-mentioned text data to obtain the second batch of annotation data; the training unit 30 is used to input the above-mentioned first batch of annotation data and the above-mentioned second batch of annotation data into the combined network model for training to obtain a trained risk prediction model, and the above-mentioned combined network model is a combined network of CNNs and Attention; the prediction unit 40 is used to input the power grid data acquired at the current moment into the above-mentioned risk prediction model to obtain the corresponding predicted risk state of the output.

[0097] The above-mentioned prediction and assessment device based on power grid data risk of the present application includes: an acquisition unit, a processing unit, a training unit, and a prediction unit. The acquisition unit is used to first acquire business operation data and text data; the processing unit is used to perform risk category annotation processing on the business operation data to obtain a first batch of annotated data, and perform sentiment category annotation processing on the text data to obtain a second batch of annotated data; the training unit is used to input the first batch of annotated data and the second batch of annotated data into a combined network model for training to obtain a trained risk prediction model, wherein the combined network model is a combination of CNNs and Attention. The prediction unit is used to input the power grid data acquired at the current moment into the risk prediction model to obtain a corresponding predicted risk status as an output. The device assigns higher weights to new data according to year and month information, introduces text sentiment features into the security features of power grid business data, uses an improved CNNs network to learn deep features, and combines CNNs + Attention to focus more on parts that are useful for improving risk assessment performance, thereby enabling timely prediction and assessment of power grid data risks and timely discovery of power grid business risks, solving the problems of insufficient accuracy of existing power enterprise data security risks and inconsistent risk assessment standards within and outside the enterprise.

[0098] In some optional instances, the acquisition unit includes a first acquisition module and a preprocessing module. The first acquisition module is used to acquire initial business data and initial text data. The initial text data is the operation text data of each branch and subsidiary obtained through web crawler technology. The preprocessing module is used to preprocess the initial business data and the initial text data respectively to obtain the business operation data and the text data. The preprocessing includes at least one of the following: data cleaning, data filling, and data standardization. The data cleaning is the process of cleaning out abnormal values, and the data filling is the process of filling missing values. By introducing text sentiment features into the security features of power grid business data, using the improved CNNs network to learn deep features, and combining the combination of CNNs+attention to focus more on the parts that are useful for performance improvement, it is possible to make timely predictions and assessments of power grid data risks.

[0099] In some optional instances, the training unit includes a second acquisition module, a first extraction module, a conversion module, and a first training module. The second acquisition module is used to obtain the data storage time of the first batch of annotation data, and according to the data storage time, set the corresponding data weight value for the first batch of annotation data to obtain the first processed data; the first extraction module is used to extract the corresponding emotional features from the second batch of annotation data based on the LSTM neural network, and classify and predict the second batch of annotation data according to the emotional features to obtain the corresponding predicted emotional category; the conversion module is used to convert the predicted emotional category into the corresponding category value and use it as the target emotional feature to form the target feature data with the first processed data; the first training module is used to input the target feature data into the combined network of CNNs and Attention for training to obtain the above-mentioned risk prediction model after training. While analyzing the power grid data, emotional indicators are introduced, and the improved CNNs+Attention neural network is used to extract features between the data, so as to timely predict and evaluate the risk status of the power grid data and provide early warning for data security risks.

[0100] In this embodiment, the second acquisition module includes a first processing module and a second processing module. The first processing module is used to obtain the weight value of the data when the year of the data storage time is the current year. Wherein, n is the number of months between the data storage time and the current time; the second processing module is used for the data weight value when the year of the data storage time is not the current year. Where b is the number of years between the data storage time and the current time. Using a deep neural network, and assigning higher weights to new data based on year and month information during model training, we can more accurately predict the risk status of power grid data.

[0101] In this embodiment, the first training module includes a second extraction module and a second training module. The second extraction module is used to extract target feature vector information from the target feature data using a CNN network. The second training module is used to input the target feature vector information into an attention network for training, thereby obtaining the trained risk prediction model. By introducing text sentiment features into the security features of power grid business data, using an improved CNN network to learn deep features, and combining CNNs with attention to focus more on aspects that are useful for performance improvement, timely prediction and assessment of power grid data risks can be achieved, effectively ensuring data security and promptly identifying power grid business risks.

[0102] An optional solution is that the combined network model is a model that combines CNNs and Attention networks. The combined network model is based on historical data [A K(t), A K (t-1),…,A K (t-m+1)], predict the next time step The main formula of the Attention network in the actual model training prediction task is: S t+1 =f(S t-1 ,y t-1 , c t ), where St is the state output at time t after decoding, St-1 is the state output at time t-1, yt-1 is the label corresponding to the input data at time t-1, f is an RNN, c t It represents the weighted sum of hidden states. By using the improved CNN network to learn deep features, it is possible to timely predict and evaluate power grid data risks and effectively ensure data security.

[0103] As an optional solution, the above c t The calculation formula is: Among them, t represents the time, h j is the output of the j-th input in the decoder, a tj is a weight parameter. The combination of CNNs and attention focuses more on the parts that are useful for performance improvement, thereby enabling timely prediction and assessment of power grid data risks, effectively ensuring data security, and promptly identifying power grid business risks.

[0104] The aforementioned power grid data risk prediction and assessment device includes a processor and memory. The aforementioned acquisition unit and other components are stored as program units in the memory, and the processor executes the program units stored in the memory to implement the corresponding functions. The aforementioned modules are all located in the same processor; alternatively, the aforementioned modules may be located in different processors in any combination.

[0105] The processor contains a core, which retrieves the corresponding program unit from memory. One or more cores can be configured, and adjusting the core parameters can address the current issues of insufficient accuracy in data security risk assessments for power companies, as well as the lack of uniformity in risk assessment standards across various companies.

[0106] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0107] An embodiment of the present invention provides a computer-readable storage medium, which includes a stored program. When the program is run, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned prediction and assessment method based on power grid data risk.

[0108] Specifically, the prediction and assessment methods based on power grid data risks include:

[0109] Step S201: Acquire business operation data and text data. The business operation data is historical operation data of the power grid digital business operation platform, and the text data is operation text data of each branch and subsidiary obtained from the internal network of the power grid.

[0110] Specifically, business operation data mainly comes from the power grid digital business operation platform and database storage data, etc. Text data is text data such as the annual operation reports of various branches and subsidiaries obtained from internal websites through web crawler technology. The data includes date, company name, and the company's text title on that date, text content, and other information.

[0111] Step S202: Annotate the business operation data according to risk categories to obtain first-order annotated data, and annotate the text data according to sentiment categories to obtain second-order annotated data.

[0112] Specifically, the collected business operation data and text data are annotated according to the actual operation of the system. Based on the collected business operation data, the corresponding risk status is marked according to the actual business situation to form the initial training data set and test set.

[0113] Step S203: Input the first batch of annotation data and the second batch of annotation data into a combined network model for training to obtain a trained risk prediction model. The combined network model is a combined network of CNNs and Attention.

[0114] Specifically, after the text data enters the sentiment classification model branch, the predicted value is added as a feature to the risk prediction model feature to perform power grid data security risk prediction.

[0115] Step S204: input the power grid data acquired at the current moment into the above-mentioned risk prediction model to obtain the corresponding predicted risk status as output.

[0116] Specifically, the trained prediction model is used to analyze and predict the data security situation. Based on the current status, the current business operation data security status is predicted. The final prediction results are of two types: normal and dangerous.

[0117] Optionally, obtaining business operation data and text data includes: obtaining initial business data and initial text data, the above-mentioned initial text data is the operation text data of each branch and subsidiary obtained through web crawler technology; preprocessing the above-mentioned initial business data and the above-mentioned initial text data respectively to obtain the above-mentioned business operation data and the above-mentioned text data, the above-mentioned preprocessing includes at least one of the following: data cleaning, data filling, and data standardization, the above-mentioned data cleaning is the process of cleaning out outliers, and the above-mentioned data filling is the process of filling missing values.

[0118] Optionally, the above-mentioned first batch of annotation data and the above-mentioned second batch of annotation data are input into a combined network model for training to obtain a trained risk prediction model, including: obtaining the data storage time of the above-mentioned first batch of annotation data, and setting corresponding data weight values ​​for the above-mentioned first batch of annotation data according to the above-mentioned data storage time to obtain first processed data; based on the LSTM neural network, extracting corresponding emotional features from the above-mentioned second batch of annotation data, and classifying and predicting the above-mentioned second batch of annotation data according to the above-mentioned emotional features to obtain corresponding predicted emotional categories; converting the above-mentioned predicted emotional categories into corresponding category values ​​as target emotional features and forming target feature data with the above-mentioned first processed data; inputting the above-mentioned target feature data into a combined network of CNNs and Attention for training to obtain the above-mentioned risk prediction model after training.

[0119] Optionally, according to the data storage time, the corresponding data weight value is set for the first batch of annotation data to obtain the first processed data, including: in the case where the year of the data storage time is the current year, the data weight value is Where n is the number of months between the data storage time and the current time; if the year of the data storage time is not the current year, the data weight value is Where b is the number of years between the data storage time and the current time.

[0120] Optionally, the target feature data is input into a combined network of CNNs and Attention for training to obtain the trained risk prediction model, including: extracting target feature vector information from the target feature data using a CNNs network; and inputting the target feature vector information into an Attention network for training to obtain the trained risk prediction model.

[0121] Optionally, the combined network model is a model that combines CNNs and Attention networks, and the combined network model is based on historical data [A K (t), A K (t-1),…,A K (t-m+1)], predict the next time step The main formula of the Attention network in the actual model training prediction task is: S t+1 =f(S t-1 ,y t-1 , c t ), where St is the state output at time t after decoding, St-1 is the state output at time t-1, yt-1 is the label corresponding to the input data at time t-1, f is an RNN, c t represents the weighted sum of the hidden states.

[0122] Optionally, the above c t The calculation formula is: Among them, t represents the time, h j is the output of the j-th input in the decoder, a tj is a weight parameter.

[0123] An embodiment of the present invention provides a processor, which is used to run a program, wherein the program executes the above-mentioned prediction and assessment method based on power grid data risks when running.

[0124] Specifically, the prediction and assessment methods based on power grid data risks include:

[0125] Step S201: Acquire business operation data and text data. The business operation data is historical operation data of the power grid digital business operation platform, and the text data is operation text data of each branch and subsidiary obtained from the internal network of the power grid.

[0126] Specifically, business operation data mainly comes from the power grid digital business operation platform and database storage data, etc. Text data is text data such as the annual operation reports of various branches and subsidiaries obtained from internal websites through web crawler technology. The data includes date, company name, and the company's text title on that date, text content, and other information.

[0127] Step S202: Annotate the business operation data according to risk categories to obtain first-order annotated data, and annotate the text data according to sentiment categories to obtain second-order annotated data.

[0128] Specifically, the collected business operation data and text data are annotated according to the actual operation of the system. Based on the collected business operation data, the corresponding risk status is marked according to the actual business situation to form the initial training data set and test set.

[0129] Step S203: Input the first batch of annotation data and the second batch of annotation data into a combined network model for training to obtain a trained risk prediction model. The combined network model is a combined network of CNNs and Attention.

[0130] Specifically, after the text data enters the sentiment classification model branch, the predicted value is added as a feature to the risk prediction model feature to perform power grid data security risk prediction.

[0131] Step S204: input the power grid data acquired at the current moment into the above-mentioned risk prediction model to obtain the corresponding predicted risk status as output.

[0132] Specifically, the trained prediction model is used to analyze and predict the data security situation. Based on the current status, the current business operation data security status is predicted. The final prediction results are of two types: normal and dangerous.

[0133] Optionally, obtaining business operation data and text data includes: obtaining initial business data and initial text data, the above-mentioned initial text data is the operation text data of each branch and subsidiary obtained through web crawler technology; preprocessing the above-mentioned initial business data and the above-mentioned initial text data respectively to obtain the above-mentioned business operation data and the above-mentioned text data, the above-mentioned preprocessing includes at least one of the following: data cleaning, data filling, and data standardization, the above-mentioned data cleaning is the process of cleaning out outliers, and the above-mentioned data filling is the process of filling missing values.

[0134] Optionally, the above-mentioned first batch of annotation data and the above-mentioned second batch of annotation data are input into a combined network model for training to obtain a trained risk prediction model, including: obtaining the data storage time of the above-mentioned first batch of annotation data, and setting corresponding data weight values ​​for the above-mentioned first batch of annotation data according to the above-mentioned data storage time to obtain first processed data; based on the LSTM neural network, extracting corresponding emotional features from the above-mentioned second batch of annotation data, and classifying and predicting the above-mentioned second batch of annotation data according to the above-mentioned emotional features to obtain corresponding predicted emotional categories; converting the above-mentioned predicted emotional categories into corresponding category values ​​as target emotional features and forming target feature data with the above-mentioned first processed data; inputting the above-mentioned target feature data into a combined network of CNNs and Attention for training to obtain the above-mentioned risk prediction model after training.

[0135] Optionally, according to the data storage time, the corresponding data weight value is set for the first batch of annotation data to obtain the first processed data, including: in the case where the year of the data storage time is the current year, the data weight value is Where n is the number of months between the data storage time and the current time; if the year of the data storage time is not the current year, the data weight value is Where b is the number of years between the data storage time and the current time.

[0136] Optionally, the target feature data is input into a combined network of CNNs and Attention for training to obtain the trained risk prediction model, including: extracting target feature vector information from the target feature data using a CNNs network; and inputting the target feature vector information into an Attention network for training to obtain the trained risk prediction model.

[0137] Optionally, the combined network model is a model that combines CNNs and Attention networks. The combined network model is based on historical data [A K (t), A K (t-1),…,A K (t-m+1)], predict the next time step The main formula of the Attention network in the actual model training prediction task is: S t+1 =f(S t-1 ,y t-1 , c t ), where St is the state output at time t after decoding, St-1 is the state output at time t-1, yt-1 is the label corresponding to the input data at time t-1, f is an RNN, c t represents the weighted sum of the hidden states.

[0138] Optionally, the above c t The calculation formula is: Among them, t represents the time, h j is the output of the j-th input in the decoder, a tj is a weight parameter.

[0139] An embodiment of the present invention provides a device, comprising a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, at least the following steps are performed:

[0140] Step S201: Acquire business operation data and text data. The business operation data is historical operation data of the power grid digital business operation platform, and the text data is operation text data of each branch and subsidiary obtained from the internal network of the power grid.

[0141] Step S202: Annotate the business operation data according to risk categories to obtain first-order annotated data, and annotate the text data according to sentiment categories to obtain second-order annotated data.

[0142] Step S203: Input the first batch of annotation data and the second batch of annotation data into a combined network model for training to obtain a trained risk prediction model. The combined network model is a combined network of CNNs and Attention.

[0143] Step S204: input the power grid data acquired at the current moment into the above-mentioned risk prediction model to obtain the corresponding predicted risk status as output.

[0144] The devices in this article can be servers, PCs, PADs, mobile phones, etc.

[0145] Optionally, obtaining business operation data and text data includes: obtaining initial business data and initial text data, the above-mentioned initial text data is the operation text data of each branch and subsidiary obtained through web crawler technology; preprocessing the above-mentioned initial business data and the above-mentioned initial text data respectively to obtain the above-mentioned business operation data and the above-mentioned text data, the above-mentioned preprocessing includes at least one of the following: data cleaning, data filling, and data standardization, the above-mentioned data cleaning is the process of cleaning out outliers, and the above-mentioned data filling is the process of filling missing values.

[0146] Optionally, the above-mentioned first batch of annotation data and the above-mentioned second batch of annotation data are input into a combined network model for training to obtain a trained risk prediction model, including: obtaining the data storage time of the above-mentioned first batch of annotation data, and setting corresponding data weight values ​​for the above-mentioned first batch of annotation data according to the above-mentioned data storage time to obtain first processed data; based on the LSTM neural network, extracting corresponding emotional features from the above-mentioned second batch of annotation data, and classifying and predicting the above-mentioned second batch of annotation data according to the above-mentioned emotional features to obtain corresponding predicted emotional categories; converting the above-mentioned predicted emotional categories into corresponding category values ​​as target emotional features and forming target feature data with the above-mentioned first processed data; inputting the above-mentioned target feature data into a combined network of CNNs and Attention for training to obtain the above-mentioned risk prediction model after training.

[0147] Optionally, according to the data storage time, the corresponding data weight value is set for the first batch of annotation data to obtain the first processed data, including: in the case where the year of the data storage time is the current year, the data weight value is Where n is the number of months between the data storage time and the current time; if the year of the data storage time is not the current year, the data weight value is Where b is the number of years between the data storage time and the current time.

[0148] Optionally, the target feature data is input into a combined network of CNNs and Attention for training to obtain the trained risk prediction model, including: extracting target feature vector information from the target feature data using a CNNs network; and inputting the target feature vector information into an Attention network for training to obtain the trained risk prediction model.

[0149] Optionally, the combined network model is a model that combines CNNs and Attention networks, and the combined network model is based on historical data [A K (t), A K (t-1),…,A K (t-m+1)], predict the next time step The main formula of the Attention network in the actual model training prediction task is: S t+1 =f(S t-1 ,y t-1 , c t ), where St is the state output at time t after decoding, St-1 is the state output at time t-1, yt-1 is the label corresponding to the input data at time t-1, f is an RNN, c t represents the weighted sum of the hidden states.

[0150] Optionally, the above c t The calculation formula is: Among them, t represents the time, h j is the output of the j-th input in the decoder, a tj is a weight parameter.

[0151] The present application also provides a computer program product, which, when executed on a data processing device, is adapted to execute a program for initializing at least the following method steps:

[0152] Step S201: Acquire business operation data and text data. The business operation data is historical operation data of the power grid digital business operation platform, and the text data is operation text data of each branch and subsidiary obtained from the internal network of the power grid.

[0153] Step S202: Annotate the business operation data according to risk categories to obtain first-order annotated data, and annotate the text data according to sentiment categories to obtain second-order annotated data.

[0154] Step S203: Input the first batch of annotation data and the second batch of annotation data into a combined network model for training to obtain a trained risk prediction model. The combined network model is a combined network of CNNs and Attention.

[0155] Step S204: input the power grid data acquired at the current moment into the above-mentioned risk prediction model to obtain the corresponding predicted risk status as output.

[0156] Optionally, obtaining business operation data and text data includes: obtaining initial business data and initial text data, the above-mentioned initial text data is the operation text data of each branch and subsidiary obtained through web crawler technology; preprocessing the above-mentioned initial business data and the above-mentioned initial text data respectively to obtain the above-mentioned business operation data and the above-mentioned text data, the above-mentioned preprocessing includes at least one of the following: data cleaning, data filling, and data standardization, the above-mentioned data cleaning is the process of cleaning out outliers, and the above-mentioned data filling is the process of filling missing values.

[0157] Optionally, the above-mentioned first batch of annotation data and the above-mentioned second batch of annotation data are input into a combined network model for training to obtain a trained risk prediction model, including: obtaining the data storage time of the above-mentioned first batch of annotation data, and setting corresponding data weight values ​​for the above-mentioned first batch of annotation data according to the above-mentioned data storage time to obtain first processed data; based on the LSTM neural network, extracting corresponding emotional features from the above-mentioned second batch of annotation data, and classifying and predicting the above-mentioned second batch of annotation data according to the above-mentioned emotional features to obtain corresponding predicted emotional categories; converting the above-mentioned predicted emotional categories into corresponding category values ​​as target emotional features and forming target feature data with the above-mentioned first processed data; inputting the above-mentioned target feature data into a combined network of CNNs and Attention for training to obtain the above-mentioned risk prediction model after training.

[0158] Optionally, according to the data storage time, the corresponding data weight value is set for the first batch of annotation data to obtain the first processed data, including: in the case where the year of the data storage time is the current year, the data weight value is Where n is the number of months between the data storage time and the current time; if the year of the data storage time is not the current year, the data weight value is Where b is the number of years between the data storage time and the current time.

[0159] Optionally, the target feature data is input into a combined network of CNNs and Attention for training to obtain the trained risk prediction model, including: extracting target feature vector information from the target feature data using a CNNs network; and inputting the target feature vector information into an Attention network for training to obtain the trained risk prediction model.

[0160] Optionally, the combined network model is a model that combines CNNs and Attention networks, and the combined network model is based on historical data [A K (t), A K (t-1),…,A K (t-m+1)], predict the next time step The main formula of the Attention network in the actual model training prediction task is: S t+1 =f(S t-1 ,y t-1 , c t ), where St is the state output at time t after decoding, St-1 is the state output at time t-1, yt-1 is the label corresponding to the input data at time t-1, f is an RNN, c t represents the weighted sum of the hidden states.

[0161] Optionally, the above c t The calculation formula is: Among them, t represents the time, h j is the output of the j-th input in the decoder, a tj is a weight parameter.

[0162] Obviously, those skilled in the art will appreciate that the various modules or steps of the present invention described above can be implemented using a general-purpose computing device, can be centralized on a single computing device, or can be distributed across a network of multiple computing devices. They can be implemented using program code executable by the computing device, and thus, can be stored in a storage device and executed by the computing device. In some cases, the steps shown or described herein can be performed in a different order than that shown, or can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0163] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0164] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0165] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0166] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0167] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0168] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0169] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0170] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0171] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0172] 1) The above-mentioned prediction and assessment method based on power grid data risk of the present application first obtains business operation data and text data; then performs risk category annotation processing on the business operation data to obtain the first batch of annotated data, and performs sentiment category annotation processing on the text data to obtain the second batch of annotated data; then the first batch of annotated data and the second batch of annotated data are input into the combined network model for training to obtain a trained risk prediction model, and the combined network model is a combination of CNNs and Attention; finally, the power grid data obtained at the current moment is input into the risk prediction model to obtain the output corresponding predicted risk status. This method gives new data a higher weight according to the year and month information, introduces text sentiment features into the security features of power grid business data, uses the improved CNNs network to learn deep features, and combines the combination of CNNs+attention to focus more on the parts that are useful for improving risk assessment performance, thereby being able to timely predict and assess power grid data risks, timely discover power grid business risks, and solve the problems of insufficient accuracy of existing power enterprise data security risks and inconsistent risk assessment standards within and outside the enterprises.

[0173] 2) The above-mentioned prediction and assessment device based on power grid data risk of the present application includes: an acquisition unit, a processing unit, a training unit and a prediction unit. The acquisition unit is used to first acquire business operation data and text data; the processing unit is used to perform risk category annotation processing on the business operation data to obtain a first batch of annotated data, and perform sentiment category annotation processing on the text data to obtain a second batch of annotated data; the training unit is used to input the first batch of annotated data and the second batch of annotated data into a combined network model for training to obtain a trained risk prediction model, wherein the combined network model is a combination of CNNs and Attention; the prediction unit is used to input the power grid data acquired at the current moment into the risk prediction model to obtain a corresponding predicted risk status as an output. The device assigns higher weights to new data according to year and month information, introduces text sentiment features into the security features of power grid business data, uses an improved CNNs network to learn deep features, and combines CNNs + attention to focus more on parts that are useful for improving risk assessment performance, thereby being able to timely predict and assess power grid data risks and timely discover power grid business risks, solving the problems of insufficient accuracy of existing power enterprise data security risks and inconsistent risk assessment standards within and outside the enterprises.

[0174] The above description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A prediction and assessment method based on power grid data risk, characterized in that: include: Acquire business operation data and text data, wherein the business operation data is historical operation data of the power grid digital business operation platform, and the text data is operation text data of each branch and subsidiary obtained from the internal network of the power grid; Performing risk category annotation processing on the business operation data to obtain first-stage annotation data, and performing sentiment category annotation processing on the text data to obtain second-stage annotation data; Inputting the first batch of annotation data and the second batch of annotation data into a combined network model for training to obtain a trained risk prediction model, wherein the combined network model is a combined network of CNNs and Attention; Inputting the power grid data acquired at the current moment into the risk prediction model to obtain the corresponding predicted risk status as output; Inputting the first batch of annotation data and the second batch of annotation data into a combined network model for training to obtain a trained risk prediction model, including: Obtaining a data storage time of the first batch of annotation data, and setting a corresponding data weight value for the first batch of annotation data according to the data storage time to obtain first processed data; Extracting corresponding sentiment features from the second annotation data based on the LSTM neural network, and performing classification prediction on the second annotation data according to the sentiment features to obtain corresponding predicted sentiment categories; Converting the predicted emotion category into a corresponding category value as a target emotion feature and forming target feature data together with the first processed data; The target feature data is input into a combined network of CNNs and Attention for training to obtain the trained risk prediction model.

2. The prediction and evaluation method according to claim 1, characterized in that: Obtain business operation data and text data, including: Acquire initial business data and initial text data, wherein the initial text data is the operational text data of each branch and subsidiary obtained through web crawler technology; The initial business data and the initial text data are preprocessed respectively to obtain the business operation data and the text data, wherein the preprocessing includes at least one of the following: data cleaning, data filling, and data standardization, wherein the data cleaning is a process of cleaning out abnormal values, and the data filling is a process of filling missing values.

3. The prediction and evaluation method according to claim 1, wherein: According to the data storage time, a corresponding data weight value is set for the first batch of annotation data to obtain first processed data, including: When the year of the data storage time is the current year, the data weight value , where n is the number of months between the data storage time and the current time; If the year of the data storage time is not the current year, the data weight value , where b is the number of years between the data storage time and the current time.

4. The prediction and evaluation method according to claim 1, wherein: The target feature data is input into a combined network of CNNs and Attention for training to obtain the trained risk prediction model, including: Using a CNNs network to extract target feature vector information from the target feature data; The target feature vector information is input into the Attention network for training to obtain the trained risk prediction model.

5. The prediction and evaluation method according to claim 1, characterized in that: The combined network model is a model that combines CNNs and Attention networks. The combined network model is based on historical data. , predict the next time step The formula of the Attention network in the actual model training prediction task is: , where St is the state output at time t after decoding, St-1 is the state output at time t-1, yt-1 is the label corresponding to the input data at time t-1, and f is an RNN. represents the weighted sum of the hidden states.

6. The prediction and evaluation method according to claim 5, characterized in that: described The calculation formula is: , where t represents the time, is the output of the j-th input in the decoder, is a weight parameter.

7. A prediction and assessment device based on power grid data risk, characterized in that: include: An acquisition unit, configured to acquire business operation data and text data, wherein the business operation data is historical operation data of the power grid digital business operation platform, and the text data is operation text data of each branch and subsidiary obtained from the internal network of the power grid; a processing unit configured to perform risk category annotation processing on the business operation data to obtain first annotated data, and perform sentiment category annotation processing on the text data to obtain second annotated data; A training unit, configured to input the first batch of annotation data and the second batch of annotation data into a combined network model for training to obtain a trained risk prediction model, wherein the combined network model is a combined network of CNNs and Attention; A prediction unit, configured to input the current grid data into the risk prediction model to obtain a corresponding predicted risk state as output; The training unit includes a second acquisition module, a first extraction module, a conversion module and a first training module. The second acquisition module is used to obtain the data storage time of the first batch of annotation data, and according to the data storage time, set the corresponding data weight value for the first batch of annotation data to obtain the first processed data; the first extraction module is used to extract the corresponding emotional features from the second batch of annotation data based on the LSTM neural network, and classify and predict the second batch of annotation data according to the emotional features to obtain the corresponding predicted emotional category; the conversion module is used to convert the predicted emotional category into the corresponding category value as the target emotional feature and form the target feature data with the first processed data; The first training module is used to input the target feature data into the combined network of CNNs and Attention for training to obtain the trained risk prediction model.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the prediction and assessment method based on power grid data risk according to any one of claims 1 to 6.

9. An electronic device, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a method for executing the predictive assessment method based on power grid data risk as described in any one of claims 1 to 6.

Images